# Arlo — full content corpus > Every product page, guide, comparison, and glossary page from https://www.arlo.sh, as raw Markdown. The > routed index lives at https://www.arlo.sh/llms.txt; any page also returns Markdown for Accept: text/markdown. # Arlo — One AI colleague for the workspace > Arlo is an AI colleague that works inside the tools your team already uses — iMessage, SMS, WhatsApp, Slack, and phone calls. It has persistent memory, runs scoped and approved actions across 3,000+ integrations, can use tools that have no API via a secure browser session, and keeps a full audit trace on every run. Canonical: https://www.arlo.sh/ ## What Arlo does - Lives in iMessage/SMS, WhatsApp, Slack, and on live phone calls — no separate app to learn. - Connects to Slack, Gmail, Notion, Linear, GitHub and 3,000+ more tools. - Uses tools that have no API by logging in once in a secure session and acting like you would. - Remembers context across conversations and delivers one morning briefing of what changed. - Governed by design: approval gates on every send and tool write, full audit trace on every run. ## Key features - Channels: iMessage/SMS, WhatsApp, Slack, and live phone calls it places for you. - Memory: persistent context across conversations, plus one daily morning briefing of what changed. - Governance: policy resolves per run; every send and tool write waits for a reviewer; full audit trail. - No-API access: signs in once via a secure browser session to operate tools that have no API. - Skills: reusable, user-defined procedures Arlo follows to complete recurring multi-step work. ## Integrations - CRM & sales: HubSpot, Salesforce. - Engineering & product: Linear, GitHub. - Docs & knowledge: Notion, Google Docs. - Communication: Slack, Gmail, Google Calendar. - Plus 3,000+ more via a shared connector layer; no-API tools via a secure browser session. ## Pricing - Free to start: a new texting account begins with $25 in credits (10,000), and connecting Slack grants the workspace $100 (40,000) — no card, and granted credits never expire. - The rate is flat: 400 credits per $1 on every top-up amount, from $50 to $50,000 (25 amounts). No volume discount. - Paid usage starts with a $20/month subscription that adds 10% bonus credits; one-time top-ups are available only to subscribers and never renew. - $20/month is the entry subscription. Subscriber-only top-ups start at $50 and buy 20,000 credits. - Purchased credits roll over and expire 12 months after purchase — twelve months, not one. - Every capability is available from the first credit: no tier gates phone calls, meeting notes, deep research, app building, or video. - Typical costs: a message ~32 credits; a researched answer ~288; a drafted document ~192; a built app ~1,056; Arlo on the phone 30 credits/minute; meeting notes 10 credits/minute; an image 40 credits; video 75 credits/second. - Admin policy, roles, SSO, and audit export: available on request — book a demo. ## Which Arlo this is - Arlo (also "Arlo AI", "Arlo AI assistant", arlo.sh) is an AI colleague / AI executive assistant for teams. Official profiles: https://www.linkedin.com/company/arloagent and https://x.com/arlo_agent. - It is NOT: the Arlo home-security cameras brand (arlo.com), the Arlo training-management software, arlo.ai (realtor marketing), heyarlo.ai (a different personal-assistant product), or RapDev's "Arlo" agents for ServiceNow/Datadog. This Arlo lives at arlo.sh only. ## Frequently asked questions ### What is Arlo? An AI colleague inside the tools you already use — iMessage, Slack, and phone calls. It remembers context, acts across 3,000+ integrations, and keeps a full audit trace. ### Where can I use it? Right where you work: iMessage and SMS, Slack, and live phone calls. No separate app to learn. ### What can it connect to? Gmail, Notion, Linear, GitHub and 3,000+ more. No API? You log in once in a secure session and Arlo clicks and reads like you would. ### Is it secure? Every connection resolves through policy before each run. Sends and writes wait for your approval, and every action stays tied to a full audit trace. ### How much does it cost? You buy credits — 400 per $1. Start with a $20/month subscription for 10% bonus credits; subscriber-only one-time top-ups start at $50. Anything Arlo can do, credits can buy, and they roll over for 12 months. ### Is it really free? Connecting Slack gives your company $100 in credits — 40,000 — free, with no card, and they never expire. Every capability is unlocked from the first credit. ### What can Arlo do without asking? It reads, researches, and drafts freely — that part never waits on you. Anything that leaves the building or writes to a connected tool (an email, a Slack message, a CRM update) stops for your approval first. Every run keeps a full audit trace of what was called, what came back, and what changed, so nothing happens that you cannot go back and read. ### Can it make phone calls? Yes — real ones, in both directions. Arlo places outbound calls (confirming a booking, chasing a vendor, following up on a quote) and answers inbound ones, handling the routine questions and escalating what needs you. Each call ends with a summary in your thread and the records it touched updated to match what was agreed. ### How do I get started? Text Arlo, or connect Slack. There is nothing to install — you message a number in iMessage or SMS like you would a contact, and it asks what you want handled. Connecting Slack puts $100 of credits (40,000) in the workspace with no card, and those never expire. ## Where to go next - [AI assistant in iMessage](https://www.arlo.sh/imessage) — text a number, no app to install - [AI phone assistant](https://www.arlo.sh/phone) — places and answers real calls - [Slack AI assistant](https://www.arlo.sh/slack) — triages threads and acts in your tools - [Pricing](https://www.arlo.sh/pricing) - [Docs and API reference](https://www.arlo.sh/docs) - [About Arlo](https://www.arlo.sh/about) - [Contact Arlo](https://www.arlo.sh/contact) - [Guides, comparisons and definitions](https://www.arlo.sh/resources) --- # AI assistant in iMessage you can text > Arlo is an AI assistant you text in iMessage — no app to install. You message a number like you would text a contact, and it answers questions, triages and drafts replies, schedules, researches, and takes real actions in your apps behind approvals. It can also place phone calls, and it remembers context across every conversation. Canonical: https://www.arlo.sh/imessage Arlo is an AI assistant you text in iMessage — free to start, no app to install. Text a number and it triages, drafts, schedules, makes calls, and acts with approvals. ## Frequently asked questions ### Is there an AI assistant I can text in iMessage? Yes. Arlo gives you a number you text in iMessage (and SMS) like a contact. There is no app to install and no separate chat window — you ask, and it answers and acts. ### Which AI assistants integrate with SMS and iMessage? Very few work in iMessage itself rather than in their own app. Arlo works over both iMessage and SMS by giving you a number you text like a contact, and the same assistant also works in WhatsApp, Slack, and on live phone calls. Poke and Martin are the closest text-first alternatives, and both are shaped around personal life rather than team work. ### Do I need to download an app? No. Arlo works inside the Messages app you already use. You text a number to start, and it also works in WhatsApp, Slack, and on phone calls if you want it there too. ### Can the iMessage AI actually do things, not just chat? Yes. Arlo connects to Gmail, Slack, Notion, Linear, GitHub and 3,000+ tools, and can use apps with no API by logging in through a secure browser session. Anything that sends or writes waits for your approval, and every action is logged in an audit trace. ### Is there a free AI I can text in iMessage? Yes. Arlo is free to start: a new texting account begins with $25 in credits (10,000), with no card and no expiry. That covers real work — questions, drafts, reminders, research — and paid plans only come in once you've used it. ### Can I use ChatGPT in iMessage instead? Not from your iPhone. OpenAI's Messages plugin runs in the ChatGPT desktop app on a Mac, where it reads and sends messages from your existing threads; it isn't a contact you can text. Arlo is the other end of the thread: you text it like a person from your phone, and it only sees what you send it. ### Does it work on WhatsApp too? Yes. You can text the same assistant on WhatsApp instead of, or as well as, iMessage and SMS, with the same memory and the same approvals. ### Can it make phone calls? Yes. Beyond texting, Arlo can place live phone calls — confirm an appointment, ask a question, or pass along a message — then keep the transcript so you can ask how it went. Most text-only assistants cannot do this. ### How is Arlo different from Poke or other AI you text? Poke and similar tools are personal assistants for your messaging apps. Arlo is a governed AI colleague for work: it spans iMessage, WhatsApp, Slack, and calls, takes approved actions across your tools, and keeps a full audit trail — built for teams, not just one inbox. ## Where to go next - [Pricing](https://www.arlo.sh/pricing) - [About Arlo](https://www.arlo.sh/about) - [Guides, comparisons and definitions](https://www.arlo.sh/resources) - [AI assistant on WhatsApp you can text](https://www.arlo.sh/whatsapp) - [The AI assistant that texts you first](https://www.arlo.sh/proactive) - [AI phone assistant that makes real calls](https://www.arlo.sh/phone) - [Slack AI assistant that does the work, not just the answering](https://www.arlo.sh/slack) --- # AI assistant on WhatsApp you can text > Arlo is an AI assistant you text on WhatsApp — free to start, with no app to install. Message it like a contact and it answers, researches, drafts, schedules, and takes real actions in Gmail, your calendar, Slack, Notion and 3,000+ other tools, asking before anything is sent. It understands voice notes, photos and documents, and it can place phone calls for you. Canonical: https://www.arlo.sh/whatsapp Arlo is an AI assistant you text on WhatsApp — free to start, no app. It reads voice notes, drafts, schedules, makes calls, and acts in your tools with your approval. ## Frequently asked questions ### Is there an AI assistant I can use on WhatsApp? Yes. Arlo works on WhatsApp: you message it like a contact and it answers, researches, drafts, schedules, and takes real actions in your email, calendar, and work tools. There is no app to install and nothing to set up beyond connecting your number. ### Is ChatGPT still on WhatsApp? Not in the US. 1-800-ChatGPT stopped working on WhatsApp in most countries on 15 January 2026, after Meta restricted general-purpose AI chatbots on its business platform. It stayed available in Brazil and returned for users in the European Economic Area in July 2026. ### Can an AI on WhatsApp actually do things, not just chat? Arlo can. It connects to Gmail, Google Calendar, Slack, Notion, Linear, HubSpot and 3,000+ other tools, so a message like "move my 3pm and tell the team" gets done rather than explained. Anything that sends or writes waits for your approval, and every action lands in an audit trail. ### Does it understand voice notes, photos, and documents? Yes. Send a voice note and Arlo transcribes it and acts on it; send a photo, a PDF, or a document and it reads the contents. Send one file per message — several attachments in a single message aren't read. ### Is it free? Arlo is free to start: a new account begins with $25 in credits (10,000), with no card and no expiry. Paid plans come in only once you've used it. ### Can it text me first on WhatsApp? Not on WhatsApp. WhatsApp only lets a business reply within 24 hours of your last message, so there Arlo answers what you send it. Scheduled briefs and reminders that text you first are an iMessage and SMS feature. ### Can Arlo read my other WhatsApp chats? No. Arlo only receives the messages you send to it. It can't see, read, or monitor your conversations with anyone else. ### How is this different from Meta AI in WhatsApp? Meta AI is built into WhatsApp and free, and Meta's Muse agent, launched in September 2026, connects to your email and calendar and can book and buy things — a strong option for personal errands. Arlo is built for work: it reaches Slack, HubSpot, Notion, Linear and 3,000+ other tools, holds anything it sends for your approval, keeps an audit trail, and is the same assistant across WhatsApp, iMessage, Slack, and phone calls. ## Where to go next - [Pricing](https://www.arlo.sh/pricing) - [About Arlo](https://www.arlo.sh/about) - [Guides, comparisons and definitions](https://www.arlo.sh/resources) - [AI assistant in iMessage you can text](https://www.arlo.sh/imessage) - [The AI assistant that texts you first](https://www.arlo.sh/proactive) - [AI phone assistant that makes real calls](https://www.arlo.sh/phone) - [Slack AI assistant that does the work, not just the answering](https://www.arlo.sh/slack) --- # AI phone assistant that makes real calls > Arlo is an AI phone assistant that has real phone conversations on your behalf. It places outbound calls — confirming a booking, chasing a vendor, following up on a quote — and you can call it yourself to hand it a task by voice. It keeps the transcript of every call, tells you how it went when you ask, and writes a full audit trace of what it did. Canonical: https://www.arlo.sh/phone Arlo is an AI phone assistant that places real phone calls for you — confirmations, follow-ups, questions for a business — and keeps the transcript of every call. ## Frequently asked questions ### What is an AI phone assistant? An AI phone assistant is an AI that handles real phone calls for you rather than just with you. It dials, speaks with whoever answers, and pursues the task to an outcome. That is different from a voice assistant like Siri, which dials the number but leaves the conversation to you. ### Can an AI assistant actually make phone calls for me? Yes. Arlo places live outbound calls — confirming appointments, chasing a supplier about a missing order, following up on a quote — and handles the back-and-forth itself, getting through phone menus, waiting on hold and leaving a voicemail if nobody picks up. You ask for it by text or in Slack, and you can ask how it went in the same thread. ### Can it answer my incoming calls too? Yes. Not calls from other people. Arlo places calls for you, and you can call your Arlo number yourself to get things done by voice. It doesn't answer or screen other people's calls on your behalf, so it isn't a receptionist. ### What happens after the call ends? This is the part that separates a demo from a dependable assistant. Arlo keeps the transcript, and the whole run lands in an audit trace you can check later. It doesn't text you a recap unprompted — ask how it went and it tells you, and it updates your calendar, CRM or notes to match what was agreed when you ask. ### How much does an AI phone call cost? Call time draws on the same prepaid credits as everything else Arlo does: 30 credits a minute, roughly 8¢. There is no separate calling plan and no bundle of call hours to buy in advance — the credits that pay for a call also pay for texts, Slack, research, and drafting. ### Will it commit to things on my behalf without asking? No. Calls follow the brief you give it. Arlo tells you who it's calling and why before it dials, introduces itself as your AI assistant, and sticks to that brief on the call. It texts you a question mid-call when you've told it to check with you, and every call is written to an audit trail with its transcript. ### Do I need an app to use it? No. You ask for a call the way you'd ask a colleague — by text in iMessage, SMS or WhatsApp, or in Slack — and Arlo makes it. There is nothing to install. ## Where to go next - [Pricing](https://www.arlo.sh/pricing) - [About Arlo](https://www.arlo.sh/about) - [Guides, comparisons and definitions](https://www.arlo.sh/resources) - [AI assistant in iMessage you can text](https://www.arlo.sh/imessage) - [AI assistant on WhatsApp you can text](https://www.arlo.sh/whatsapp) - [The AI assistant that texts you first](https://www.arlo.sh/proactive) - [Slack AI assistant that does the work, not just the answering](https://www.arlo.sh/slack) --- # Slack AI assistant that does the work, not just the answering > Arlo is an AI assistant that works inside Slack like a teammate. Mention it in a channel or DM and it untangles the thread, drafts the reply, and takes real action in your connected tools — updating the CRM, filing the issue, sending the email — with an approval gate before anything leaves the building. It remembers context across Slack, iMessage, WhatsApp, and phone calls, so the work follows the conversation. Canonical: https://www.arlo.sh/slack Arlo is an AI assistant in Slack that triages threads, drafts replies, and takes real action in Gmail, HubSpot, Notion and Linear — with approval before every send. Free to start. ## Frequently asked questions ### What is a Slack AI assistant? A Slack AI assistant works inside your channels and DMs rather than in a separate tab. It answers questions with your team's context, triages and summarizes long threads, drafts the replies you were going to write anyway, and — the part most tools skip — takes real action in your other tools instead of just producing text. ### Is there an AI assistant that works inside Slack and can update our CRM? Yes. Arlo connects to HubSpot and Salesforce alongside Gmail, Notion, Linear, GitHub and 3,000+ other tools. A rep can say what happened on a call in Slack and Arlo logs it to the CRM, so the update happens where the conversation already is instead of being remembered later. ### Can it work out who owns a thread and route it? Yes. Arlo reads the channel, works out what is actually being asked and who owns it, and can triage, summarize, or escalate accordingly — which is what makes it useful in a busy support or ops channel rather than just a question box. ### How is this different from Slack's own AI? Slack AI is very good at summarizing and searching what is already inside Slack. Arlo is built to act outside it: sending the email, updating the record, filing the issue, placing the phone call — and carrying one memory across Slack, iMessage, WhatsApp, and calls. They solve different halves of the problem. ### Will it send things without asking? No. Approvals are on by default: anything that sends or writes pauses for a reviewer you choose. Policy resolves before every run and every tool call, source, and approval is recorded in a full audit trail you can check afterwards. ### Does it remember corrections our team makes? Yes. Memory is persistent and shared: correct Arlo once — a name, a preference, how your team words something — and the correction carries forward across Slack, iMessage, WhatsApp, and calls rather than resetting each conversation. ### Is it free to try in Slack? Connecting Slack grants your workspace $100 in credits (40,000) with no card, and those granted credits never expire. That covers real work, not a token trial. ## Where to go next - [Pricing](https://www.arlo.sh/pricing) - [About Arlo](https://www.arlo.sh/about) - [Guides, comparisons and definitions](https://www.arlo.sh/resources) - [AI assistant in iMessage you can text](https://www.arlo.sh/imessage) - [AI assistant on WhatsApp you can text](https://www.arlo.sh/whatsapp) - [The AI assistant that texts you first](https://www.arlo.sh/proactive) - [AI phone assistant that makes real calls](https://www.arlo.sh/phone) --- # The AI assistant that texts you first > A proactive AI assistant starts the conversation instead of waiting to be asked — it sends a morning brief, reminds you before a follow-up slips, and flags changes worth knowing. Arlo does this over iMessage and SMS, so there is no app to open and no dashboard to check: the update arrives in the thread you already read. You tell it what to watch once, in plain language, and it runs on a schedule until you say stop. Canonical: https://www.arlo.sh/proactive Most AI waits to be opened. Arlo texts you first — a morning brief, a nudge before something slips, a heads-up when something changes — in Messages, with no app to open. ## Frequently asked questions ### What is a proactive AI assistant? A proactive AI assistant reaches out on its own rather than waiting for you to open it. Instead of a chat window you have to remember, it sends a morning brief, a reminder before something slips, or a heads-up when something you care about changes. Arlo does this by text, so the update lands in Messages next to everything else you read. ### Can an AI assistant text me first? Yes. Arlo can start the conversation. You tell it what to watch and when — a daily rundown at 8am, a nudge to follow up on a quote, a check the night before a deadline — and it texts you on that schedule without being prompted. You can change or cancel any of it by replying in plain language. ### How do I set up a recurring reminder or daily brief? You ask for it in one sentence. Say "text me a rundown of what changed every morning at 8" and Arlo schedules it, confirms the exact next run time, and starts. There is no builder, no cron syntax, and no separate app — the setup happens in the same thread as the reminders. ### Is this different from ChatGPT scheduled tasks or Gemini's daily brief? The behaviour is similar; the delivery is not. Those run inside their own apps, so a proactive nudge only reaches you if you open the app or allow its notifications. Arlo arrives as a text message, which reaches the surface you already check all day. It can also act on what it finds — draft the reply, book the time, make the call — behind an approval. ### Will it text me too much? It only sends what you asked for, and quiet is a valid outcome. You can tell an automation to stay silent when there is nothing worth saying, so a watch on a slow-moving thing costs you nothing on the days nothing happens. Any schedule can be paused, edited, or cancelled by replying to it. ### What can it watch for me? Anything it can reach: your inbox, calendar, a Slack channel, a document, a website, a CRM record, or your own list of follow-ups. Typical ones are a morning brief of what changed overnight, a reminder to chase a lead who has gone quiet, a bill or renewal warning, and a nudge before a commitment comes due. ## Where to go next - [Pricing](https://www.arlo.sh/pricing) - [About Arlo](https://www.arlo.sh/about) - [Guides, comparisons and definitions](https://www.arlo.sh/resources) - [AI assistant in iMessage you can text](https://www.arlo.sh/imessage) - [AI assistant on WhatsApp you can text](https://www.arlo.sh/whatsapp) - [AI phone assistant that makes real calls](https://www.arlo.sh/phone) - [Slack AI assistant that does the work, not just the answering](https://www.arlo.sh/slack) --- # Arlo pricing — prepaid credits > Arlo starts free with a Slack credit grant, then paid usage starts with a monthly subscription. Subscriber-only one-time top-ups are available when you need more, with no per-seat charge and no capability locked behind a higher plan. Canonical: https://www.arlo.sh/pricing ## The model - Free to start: a new texting account begins with $25 in credits (10,000), and connecting Slack grants the workspace $100 (40,000) — no card, and granted credits never expire. - The rate is flat: 400 credits per $1 on every top-up amount, from $50 to $50,000 (25 amounts). No volume discount. - Paid usage starts with a $20/month subscription that adds 10% bonus credits; one-time top-ups are available only to subscribers and never renew. - $20/month is the entry subscription. Subscriber-only top-ups start at $50 and buy 20,000 credits. - Purchased credits roll over and expire 12 months after purchase — twelve months, not one. - Every capability is available from the first credit: no tier gates phone calls, meeting notes, deep research, app building, or video. - Typical costs: a message ~32 credits; a researched answer ~288; a drafted document ~192; a built app ~1,056; Arlo on the phone 30 credits/minute; meeting notes 10 credits/minute; an image 40 credits; video 75 credits/second. - Admin policy, roles, SSO, and audit export: available on request — book a demo. ## Where to go next - [Full pricing page, including every purchase amount](https://www.arlo.sh/pricing) - [Contact Arlo about admin policy, roles, SSO, or audit export](https://www.arlo.sh/contact) - [Docs and API reference](https://www.arlo.sh/docs) --- # Can an AI cancel subscriptions or negotiate bills for you? > Yes. Rocket Money cancels subscriptions for Premium members and negotiates bills for a share of the savings, and Pine AI calls, emails and waits on hold to cancel services or lower bills. An assistant such as Arlo does both from a text: it emails the company or uses its cancellation page, or calls, waits on hold and asks for a lower price within limits you set. Updated: 2026-09-23 · Canonical: https://www.arlo.sh/blog/ai-that-cancels-subscriptions-and-negotiates-bills Cancelling a subscription should take one click. It usually takes a login you've forgotten, a "are you sure?" screen, and sometimes a phone call. Lowering a cable or phone bill takes a phone call, a hold queue and some patience. A few services now do this for you, and they charge for it. Here's what's available as of September 2026. We make Arlo, one of the options, and we say where it fits. ## Services built for bills and subscriptions **Rocket Money** is the best-known. Premium members can have it cancel many subscriptions for them, but not ones billed through Apple, Google Play, Amazon, Roku or PayPal ([Rocket Money](https://help.rocketmoney.com/en/articles/934402-how-do-i-cancel-a-subscription)). Its bill negotiation charges 35% to 60% of the first year's savings, and only if it succeeds ([Rocket Money](https://help.rocketmoney.com/en/articles/9744474-bill-negotiation-charge-explained)); its Premium+ tier includes negotiation with no fee ([Rocket Money](https://www.rocketmoney.com/terms)). In August 2026 it began rolling out Rowan, a texting agent, to some members ([PR Newswire](https://www.prnewswire.com/news-releases/rocket-moneys-rowan-rewrites-what-ai-can-do-in-personal-finance-302859522.html)). **Pine AI** calls, emails and waits on hold for you to cancel services, get refunds and negotiate TV, internet, phone and insurance bills in the US and Canada. It's free to try, and for bill negotiation it charges a share of what it saves you, only if it saves you money ([Pine](https://www.19pine.ai/lower-your-bills)). **Subpilot** is a subscription app that cancels subscriptions for you and sells extra cancellations as one-off top-ups ([Subpilot](https://subpilot.tech/pricing)). **DoNotPay** still sells cancellation tools; in 2025 the FTC finalised an order against it over its "AI lawyer" claims ([FTC](https://www.ftc.gov/news-events/news/press-releases/2025/02/ftc-finalizes-order-donotpay-prohibits-deceptive-ai-lawyer-claims-imposes-monetary-relief-requires)), and some recent reviews complain about stuck cancellations and unexpected charges. Meta says its Muse agent can lower a bill, and Instinct says its new calling service can sort out a cable bill, but neither publishes results. ## When an assistant you text is enough A lot of cancellations don't need a negotiator. They need someone to write the email or click through the settings page. An assistant that can use your email and a browser can do that. Arlo can: - Write the cancellation or refund email from your own Gmail or Outlook, quoting the account details it finds in your inbox. It shows you the email and sends it only when you say yes. - Go through a company's cancellation page in a browser session, using a login you save once. If you ask it to cancel something online, it signs in and works through the steps. - Remind you to chase it if the company doesn't reply, and draft the follow-up. - Call the company. It presses through the phone menu, waits on hold, turns down retention offers unless you say otherwise, and asks for the cancellation date and confirmation number. You can also ask it to check any final charge. To lower a bill, give Arlo your limit, say "see if they'll go under $60 a month, otherwise cancel". On the call it asks for promotions and loyalty discounts, mentions a competitor's price if you give one, and asks for the retention team if the first answer is no. It won't agree to a new plan, fee or contract you didn't allow; for anything outside your limit, it texts you and waits for your answer. If you want every bill tracked and renegotiated for you, Rocket Money or Pine is the better pick. See [Is there an AI that waits on hold for you?](/blog/ai-that-waits-on-hold-for-you) for the phone side. ## Frequently asked questions **Is there an app that cancels subscriptions for you?** Yes. Rocket Money cancels many subscriptions for its Premium members, but not ones billed through Apple, Google Play, Amazon, Roku or PayPal, and Subpilot cancels them as part of a paid subscription. An assistant such as Arlo can also email the company, go through its cancellation page, or call and wait on hold for you, with your approval. **Is there an AI that negotiates bills for you?** Yes. Rocket Money negotiates bills for a share of the first year's savings, and Pine AI calls and waits on hold to lower TV, internet, phone and insurance bills in the US and Canada. Arlo calls and asks for a lower price within limits you set, and Meta says Muse can lower bills too. **Can Arlo cancel my subscriptions?** Yes. It can email the company from your inbox, go through the online cancellation steps with a login you've saved, or call, get through the phone menu and wait on hold. It asks before it sends an email or places a call, and it won't agree to a charge you didn't allow. **Is it safe to let an AI cancel things for me?** Check what it can do without asking. Arlo shows you each email before it sends it. Services that negotiate bills usually need account details for the providers involved, so read what access they ask for. --- # Is there an AI that waits on hold for you? What works in 2026 > Yes. On an iPhone with iOS 26, Hold Assist listens for hold music and alerts you when an agent picks up; Google Pixel phones have had Hold for Me since 2020. Both keep you on the line. Services like ClawCall, Pine AI and Arlo take the whole call: they get through the phone menu, wait on hold and talk to the agent for you. Updated: 2026-10-07 · Canonical: https://www.arlo.sh/blog/ai-that-waits-on-hold-for-you Nobody wants to spend 40 minutes listening to the same eight bars of hold music. The good news is that the phone you already own may handle it. The less good news is that the built-in features still need you nearby, and the services that take the whole call off your hands are paid and narrow. Here's what works as of September 2026. We make Arlo, one of the options, so read that part with that in mind. ## The short answer | Option | Where it works | What it does | You still need to | |---|---|---|---| | **iPhone Hold Assist** | iPhone 12 and later on iOS 26 | Waits during hold music and alerts you when someone picks up | Stay near the phone and take the call | | **Google Pixel Hold for Me** | Pixel 3 and later, in the US, UK, Ireland, Canada, Australia and Japan | Waits on hold and tells you when a person answers | Stay near the phone and take the call | | **ClawCall** | US businesses | Calls for you, gets through phone menus and hold, then calls you back | Pay a monthly subscription | | **Pine AI** | US and Canada | Calls, emails and waits on hold to cancel services, get refunds and lower bills | Pay a share of what it saves you | | **Arlo** | US and Canadian numbers | Calls from a text, presses through the phone menu, waits on hold, then talks to the agent | Approve the call and share the account details | ## What your phone can already do **iPhone.** Apple added Hold Assist to the Phone app in iOS 26, released in September 2025. When you're put on hold, you can let your iPhone wait; it listens for the hold music to end and alerts you when an agent is on the line ([Apple](https://support.apple.com/en-us/127521)). It works on iPhone 12 and later, including iPhone SE (3rd generation), but not in every country. It isn't foolproof. Because it listens for music, recorded announcements or songs with lyrics can confuse it ([MacRumors](https://www.macrumors.com/how-to/ios-26-make-your-iphone-wait-on-hold-for-you/)), and some users report missed alerts. **Google Pixel.** Hold for Me launched with the Pixel 5 and Pixel 4a (5G) in 2020 and now works on Pixel 3 and later, in English in the US, UK, Ireland, Canada and Australia and in Japanese in Japan ([Google](https://support.google.com/phoneapp/answer/10071878)). It waits while you're on hold and lets you know when a person joins. Samsung phones don't have an equivalent built in. Both features save you from holding the phone to your ear. Neither makes the call or talks to anyone for you. ## Services that take the call off your hands **ClawCall** waits through a US business's phone menu and hold queue for you, then calls you back when a person is on the line. Plans are a small monthly subscription ([ClawCall](https://clawcall.dev/hold-for-me)). **Pine AI** handles customer-service calls end to end: cancellations, refunds and bill negotiations, by phone and email, including waiting on hold. It's free to try and, for bills, charges a share of what it saves you, in the US and Canada ([Pine](https://www.19pine.ai/lower-your-bills)); we compare it with Rocket Money in [Can an AI cancel subscriptions or negotiate bills for you?](/blog/ai-that-cancels-subscriptions-and-negotiates-bills) **Catch**, an assistant for executives, says it gets through phone menus when it calls restaurants, clinics and airlines for you. Pally, an assistant you text, has worked through phone menus and waited on hold since 1 October 2026, and connects you when a person answers ([Pally](https://pally.com/guides/ai-that-calls-businesses-for-you)). Instinct's Concierge feature, in early access, handles jobs like a cable bill. Code spotted in a beta of Google's Phone app points to a Gemini "Call for me" feature that would wait on hold for you, but as of September 2026 Google hasn't announced it. ## What Arlo does Arlo is an AI assistant you text, and it can take the whole call. Text it something like *"call Comcast and cancel my internet, the phone number on the account is 202-555-0142"*. It shows you the number and what it will ask, and dials once you say yes. It presses through the phone menu, enters the account details the menu asks for, and waits on hold for up to 90 minutes without talking over the music. When an agent picks up, it says it's your AI assistant and gets on with the request. It won't agree to a charge or plan you didn't allow, and it calls US and Canadian numbers. [Here's how Arlo's calls work](/phone). More on what each calling assistant does in [AI assistants that can make phone calls](/compare/ai-assistants-that-can-make-phone-calls). ## Frequently asked questions **Is there an app that can wait on hold for me?** Yes. iPhones on iOS 26 have Hold Assist built into the Phone app, and Google Pixel phones have Hold for Me. Both wait while you're on hold and alert you when a person picks up. Services such as ClawCall and Pine AI go further and wait through the call for you. **Does iPhone Hold Assist actually work?** Often, but not always. It listens for hold music, so recorded messages or music with lyrics can confuse it, and some users report missing the alert when an agent picks up. Keep the phone close. **Does Samsung have a Hold for Me feature?** Not built in, as of September 2026. Hold for Me is a Google Pixel feature, and Samsung users have been asking for an equivalent. **Can Arlo wait on hold for me?** Yes. Arlo calls the company for you, presses through the phone menu, waits on hold for up to 90 minutes and talks to the agent when one picks up. It asks you before it dials. --- # Can an AI book a doctor's appointment for you? > Often, yes. If your doctor takes online bookings, Zocdoc or the practice's MyChart is fastest, and US adults can now book Zocdoc appointments inside Gemini. If the office only books by phone, an assistant that calls for you can do it. Arlo calls the office, gets through its phone menu, checks your calendar during the call and adds the appointment. Updated: 2026-09-23 · Canonical: https://www.arlo.sh/blog/can-ai-book-a-doctors-appointment Booking a doctor or dentist is the kind of small job that eats a morning: finding when the office is open, waiting to get through, then juggling their free slots against your calendar. Whether an AI can do it for you depends mostly on how your doctor's office takes bookings. Here's what works as of September 2026, from the fastest option to the one that needs a phone call. We make Arlo, one of the options, so read that part with that in mind. ## If your doctor books online **Zocdoc** lists thousands of practices you can book directly. Since August 2026, US adults can also book Zocdoc appointments from inside the Gemini app ([PR Newswire](https://www.prnewswire.com/news-releases/zocdoc-brings-seamless-real-time-appointment-booking-to-gemini-302849229.html)). **MyChart**, the patient portal many health systems use, lets you book some visits yourself and request others, depending on the visit type and your health system's rules. Cleveland Clinic, for example, only lets you book directly with a provider you've seen in person in the past three years ([Cleveland Clinic](https://my.clevelandclinic.org/resources/mychart-appointment-and-scheduling-troubleshooting)). If your doctor is on either, start there. It's quicker than any call. ## If the office only books by phone This is where an assistant that makes calls helps, and where it's worth knowing the limits. **Google's Ask for Me** calls local businesses for you from Search, in categories like auto repair, salons, pet groomers and, since 2026, home repair. Its list doesn't include doctors or dentists, and it doesn't book appointments ([Google](https://support.google.com/websearch/answer/16421135)). **Instinct**, in early access, can put you on a dentist's cancellation list with its Concierge feature. **Catch**, an assistant for executives, calls clinics for you. **Arlo** makes the call for you. You text it something like *"call Dr. Lee's office and book a cleaning next week, I'm free Tuesday or Thursday morning"*. Before it dials, it shows you the number and what it's going to ask, and waits for your yes. On the call it introduces itself as your AI assistant, asks for a time, can check your calendar while it's on the line, and adds the appointment when the office confirms. If you told it to check with you, it can text you a quick question mid-call. If nobody answers, it leaves a voicemail. Afterwards it keeps a transcript, so you can ask how it went. An office line that starts with "press 1 for appointments" is fine: Arlo presses through the menu and waits on hold if it has to. It calls US and Canadian numbers only, and it asks you before every call. ## Privacy and what the office will ask HIPAA covers doctors' offices that handle billing electronically, and the vendors they hire, not an assistant you hire yourself to make the call ([CMS](https://www.cms.gov/priorities/key-initiatives/burden-reduction/administrative-simplification/hipaa/covered-entities)). Offices are careful about what they discuss with anyone other than you, so expect questions about who's calling. Booking a time rarely needs medical details. Give an assistant your availability and the type of visit, and keep the rest for the appointment itself. ## Frequently asked questions **Can AI book a doctor's appointment for me?** Yes, in a few ways. If your doctor is on Zocdoc you can book online or inside Gemini, and many practices let you book through MyChart. If the office only takes bookings by phone, an assistant that calls for you, such as Arlo, can book it. **Can Google call my doctor for me?** Not that we can find, as of September 2026. Google's AI calling in Search covers businesses like auto repair shops, salons and home repair, doesn't list medical offices, and doesn't make bookings. **Can Arlo book a dentist appointment?** Yes. Arlo calls the office, gets through the phone menu or waits on hold if it has to, asks for a time that fits your calendar, and adds the appointment. It asks you before it dials. **Do I have to share medical information?** Usually not. To book, the office needs your name, the type of visit and when you're free. An assistant doesn't need your medical history to make an appointment. --- # Can you use ChatGPT in iMessage? What actually works in 2026 > Partly. Since 20 August 2026 the ChatGPT desktop app can search, draft and send Apple Messages from an Apple silicon Mac, asking before each send by default. On iPhone there is no ChatGPT you can text in iMessage — Apple Intelligence can hand a request to ChatGPT, but it isn't a thread. To text an AI like a contact from your phone, use an assistant with its own number, such as Arlo. Updated: 2026-09-23 · Canonical: https://www.arlo.sh/blog/chatgpt-in-imessage "ChatGPT in iMessage" means three different things depending on who is asking, and only one of them lets you text an AI from your phone. On 20 August 2026 OpenAI added an Apple Messages plugin to the ChatGPT desktop app, which put the phrase in the news. That plugin runs on a Mac, though, and it works on the messages you already have. It isn't someone you can text. Here is what each option actually does, checked against public reporting as of September 2026. ## The short answer | What people mean | Where it runs | Can you text it? | Can it send texts for you? | |---|---|---|---| | **ChatGPT's Messages plugin** | ChatGPT desktop app on an Apple silicon Mac | No | Yes — asks for approval before each send by default | | **ChatGPT in Apple Intelligence** | Supported iPhones, iPads and Macs | No | No — it helps you write, you send | | **An AI assistant with its own number** (Arlo, Poke, Pally and others) | Any phone that can text | Yes — it's a contact in Messages | Yes, depending on the product; Arlo holds sends for your approval | | **Third-party ChatGPT bridges** | A number that relays to a model | Yes | Rarely — mostly question and answer | If you want to open Messages on your iPhone and text an AI the way you text a person, only the last two rows do that. ## What the ChatGPT Messages plugin does The plugin gives ChatGPT access to the Messages app on your Mac. According to [9to5Mac](https://9to5mac.com/2026/08/20/chatgpt-update-adds-apple-messages-integration-on-mac/) and [MacRumors](https://www.macrumors.com/2026/08/20/chatgpt-imessages-mac/), it can search your iMessage, SMS and RCS conversations, find something in your history, suggest and draft replies, and send messages to your contacts. It is included on every ChatGPT plan but works only in the desktop app's Work and Codex modes, not in ordinary chats, and only on Apple silicon Macs — not on the web or on your phone. To do that it needs Full Disk Access in macOS System Settings, plus access to your contacts and to automation. OpenAI says the plugin runs locally and reads messages only in response to a request rather than indexing them. Sending asks you to approve the message and recipients by default; you can allow sending to a specific chat without asking, and OpenAI itself advises against granting persistent approval. That access model is the point of the privacy criticism. [Proton](https://proton.me/blog/chatgpt-apple-messages) and [Bloomberg](https://www.bloomberg.com/news/articles/2026-08-20/chatgpt-can-now-control-imessage-potentially-raising-apple-privacy-concerns) both noted that messages protected inside Apple's ecosystem can be passed to ChatGPT and handled under OpenAI's retention and legal-access rules instead. Whether that trade is worth it depends on what is in your threads — but it is a different trade from texting an assistant, which only ever sees the messages you send it. ## What works on iPhone Nothing from OpenAI puts ChatGPT inside an iMessage thread on iPhone. What you do have is the ChatGPT extension in Apple Intelligence on supported devices: Siri can pass a question to ChatGPT, and the writing tools can use it to draft or rewrite text you are composing — including a message you are about to send. It is useful for wording a reply. It is not a contact you can text, it does not remember your earlier conversations, and it does not go and do anything for you. OpenAI's other messaging route closed in most countries earlier in the year: 1-800-ChatGPT stopped working on WhatsApp on 15 January 2026, including in the US, after Meta changed its business-messaging policy for general-purpose AI assistants. Brazil was exempted, and it returned for users in the European Economic Area in July 2026 — the full story is in [Is ChatGPT still on WhatsApp?](/blog/chatgpt-on-whatsapp). ## Texting an AI like a contact The other way to get AI into iMessage is to make the AI the other end of the conversation. The assistant has its own number; you save it as a contact and text it from any device you already message from. Nothing reads your other threads, because the assistant only receives what you send it. This is how [Arlo](/imessage) works. You text a number and it answers, drafts, schedules and researches. Because it's connected to Gmail, your calendar, Slack, Notion and thousands of other tools, it can go and do the thing you asked for. Anything that sends or writes waits for your approval, and it can also [place phone calls](/phone). Poke and Pally (pally.com) are the best-known personal alternatives. Pally has been free for everyone since 6 October 2026, and with its optional Mac app installed it can also read your own iMessage history. The full field is compared in [the best AI assistants you can text](/compare/best-ai-assistants-you-can-text), and the difference between a chat app and an assistant that acts is covered in [Arlo vs ChatGPT](/compare/arlo-vs-chatgpt). ## Which one to use - **You live on a Mac and want help with the messages you already have** — searching old threads, drafting replies, catching up on a group chat: ChatGPT's Messages plugin, as long as you are comfortable with the access it needs. - **You want help wording a text on your iPhone:** the ChatGPT option in Apple Intelligence's writing tools. - **You want to text an AI from your phone and have it get things done** — book the meeting, chase the reply, make the call: an assistant with its own number. The three are not mutually exclusive. Plenty of people will draft with Apple's tools and delegate to an assistant they text. Details reflect public reporting as of September 2026. OpenAI and Apple both ship changes to these features often, so check their own documentation for the current state. ## Frequently asked questions **Can I use ChatGPT in iMessage on my iPhone?** Not as a conversation. OpenAI's Messages plugin runs only in the ChatGPT desktop app on Apple silicon Macs. On iPhone, Apple Intelligence can pass a Siri request or a draft to ChatGPT, but there is no ChatGPT thread you can text. To text an AI from your iPhone, use an assistant with its own number, such as Arlo. **Does ChatGPT have access to my iMessages?** Only if you install the Messages plugin on a Mac and grant it Full Disk Access, contacts and automation permissions. OpenAI says it reads messages only when you ask it to and does not build an index of them. Without the plugin, ChatGPT cannot see your messages. **Can ChatGPT send text messages for me?** Yes, from a Mac with the Messages plugin. It asks you to approve each message and its recipients by default, and you can allow a specific conversation without asking — though OpenAI warns against granting persistent approval. **Is there an AI I can text in iMessage?** Yes. Assistants such as Arlo give you a number you text like a contact, from iMessage or SMS, with no app to install. Arlo also takes action in your tools behind approvals and can make phone calls, and it is free to start. **Is ChatGPT still on WhatsApp?** Not in the US. 1-800-ChatGPT stopped working on WhatsApp in most countries on 15 January 2026 after Meta restricted general-purpose AI assistants on its business platform; Brazilian numbers were exempted, and it came back for users in the European Economic Area in July 2026. **Is the ChatGPT Messages plugin private?** It runs locally on your Mac, but the messages it reads for a request are processed by ChatGPT under OpenAI's policies. Privacy groups including Proton have criticised the access it requires. An assistant you text only ever receives the messages you send to it. --- # Is ChatGPT still on WhatsApp? What happened, and what works in 2026 > Only in Europe and Brazil. 1-800-ChatGPT stopped working on WhatsApp in most countries on 15 January 2026, when Meta barred general-purpose AI chatbots from its business platform. Brazil was exempted, and it returned across the European Economic Area on 13 July 2026. In the US it is gone. Meta AI is built into WhatsApp, and assistants such as Arlo let you text an AI there for work. Updated: 2026-09-23 · Canonical: https://www.arlo.sh/blog/chatgpt-on-whatsapp For about a year you could save 1-800-ChatGPT as a WhatsApp contact and message ChatGPT like a friend. In most of the world, including the US, that stopped on 15 January 2026. Brazil was exempted, and it came back in Europe in July. Here is what happened, where ChatGPT on WhatsApp still works, and what you can use instead if you want an AI in your WhatsApp chat list. Facts are checked against public reporting as of September 2026. ## What happened In October 2025 Meta changed the terms of its WhatsApp Business platform to bar general-purpose AI assistants, where the AI is the main thing being offered. Businesses can still use AI for customer support on WhatsApp; what the new rule targets is a standalone chatbot like ChatGPT. The change took effect on **15 January 2026**. ChatGPT, Microsoft Copilot and Perplexity all left WhatsApp that day — except for users with Brazilian (+55) numbers, whom Meta exempted after Brazil's competition regulator intervened ([TechCrunch](https://techcrunch.com/2026/01/15/after-italy-whatsapp-excludes-brazil-from-rival-chatbot-ban)). Europe went differently. In June 2026 the European Commission ordered interim measures requiring Meta to restore free access for rival AI assistants, and OpenAI brought ChatGPT back to WhatsApp across the European Economic Area on **13 July 2026** ([The Decoder](https://the-decoder.com/chatgpt-returns-to-whatsapp-in-europe-after-eu-forces-meta-to-open-the-door-to-rival-ai-bots/), [GSMArena](https://m.gsmarena.com/chatgpt_is_back_in_whatsapp_in_europe_thanks_to_the_european_commission-amp-73714.php)). The order covers the EEA only. ## Where ChatGPT on WhatsApp works today | Where you are | ChatGPT on WhatsApp? | |---|---| | EU, Norway, Iceland, Liechtenstein (EEA) | Yes — message 1-800-ChatGPT again | | Brazil | Yes — it was never switched off there | | United States | No | | Most other countries | No | If you're outside the EEA and Brazil and message the old number, it won't answer. ChatGPT itself still works in its own app and on the web; it's only the WhatsApp contact that went away. ## What you can use on WhatsApp instead **Meta AI** is built into WhatsApp and is free. Meta's separate **Muse** agent, launched in September 2026, can be connected to WhatsApp in the US and Canada, so you can text it there. It connects to your email and calendar, and can book travel and buy things for you, asking before it completes a purchase. For personal errands, that's the obvious free place to start ([how Muse compares with the alternatives](/compare/muse-alternatives)). **Assistants built for work** are the other option. [Arlo](/whatsapp) is one: you message it on WhatsApp like a contact and it answers, researches and drafts — and, because it connects to Gmail, your calendar, Slack, Notion and thousands of other tools, it can also do the work, asking before anything is sent. It understands voice notes, photos and documents, and it can place phone calls for you. Martin and Catch also work over WhatsApp, and Poke does in some markets; the [best AI assistants you can text](/compare/best-ai-assistants-you-can-text) compares them. One honest difference from the old ChatGPT contact: on WhatsApp, assistants can only reply within 24 hours of your last message, so they answer the conversations you start. If you want an AI that texts you first — a morning brief, a reminder before something slips — that works over iMessage and SMS instead. See [the AI assistant that texts you first](/proactive). ## Which one to use - **You're in Europe and want ChatGPT specifically:** message 1-800-ChatGPT on WhatsApp again. - **You want a free assistant for personal errands inside WhatsApp:** Meta AI, which is already there, or Meta's separate Muse agent, which you can connect in the US and Canada. - **You want an AI on WhatsApp that gets things done** — books the meeting, drafts the reply, updates the record, makes the call: an assistant built for that, such as [Arlo](/whatsapp). If you're on an iPhone, the same question comes up for iMessage; we cover it in [Can you use ChatGPT in iMessage?](/blog/chatgpt-in-imessage). Details reflect public reporting as of September 2026. Meta's terms and the EU case are both still moving, so check OpenAI's and Meta's own announcements for the current state. ## Frequently asked questions **Is ChatGPT still on WhatsApp?** Only in the European Economic Area and Brazil. 1-800-ChatGPT stopped working on WhatsApp in most countries on 15 January 2026 after Meta barred general-purpose AI chatbots from its business platform. Brazilian numbers were exempted, and it returned on 13 July 2026 for EEA users after an EU antitrust order. In the US and most other countries it is no longer available. **Why did ChatGPT leave WhatsApp?** Meta changed its WhatsApp Business terms in October 2025 to prohibit general-purpose AI assistants as the main service offered, effective 15 January 2026. ChatGPT, Microsoft Copilot and Perplexity all left WhatsApp as a result. **What is the ChatGPT WhatsApp number?** It was 1-800-ChatGPT (+1 800 242 8478). It works on WhatsApp only for users in the European Economic Area and Brazil. **Is there an AI I can text on WhatsApp in the US?** Yes. Meta AI is built into WhatsApp, and Meta's separate Muse agent can be connected to it (US and Canada). For work, assistants such as Arlo also work over WhatsApp and take real actions in your email, calendar and work tools with your approval. **Can an AI on WhatsApp message me first?** Not usually. WhatsApp only lets a business reply within 24 hours of your last message, so AI assistants there answer what you send them. For an assistant that texts you first with briefs and reminders, use one that works over iMessage or SMS. --- # Cognition bought Poke. What it means if you're buying an AI assistant > On 23 July 2026 Cognition, the lab behind the Devin coding agent, acquired The Interaction Company — maker of the texting assistant Poke — in a deal reported in the low nine figures. Poke keeps running and gains Cognition's models and infrastructure; the stated rationale was to bring Poke's conversational style into Devin. The read for buyers: interface is now priced as an acquirable asset. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/blog/cognition-poke-what-it-means On 23 July 2026, Cognition — the AI lab known for the Devin coding agent — announced it had acquired The Interaction Company, the startup behind Poke, the assistant you text in iMessage. Reported deal value was in the low nine figures. Cognition said Poke keeps running and picks up its models and infrastructure, and that Poke's conversational style would be brought into Devin. We compete with Poke, so read this accordingly. We are not going to tell you the product is doomed, because we have no reason to think it is and the claim would not survive contact with reality. What the deal does is change which questions are worth asking, and those questions apply to us as much as to them. ## What the deal actually signals **Interface is being priced as a moat.** Cognition did not buy a model or a dataset. It bought an interaction style — the thing that makes texting Poke feel like texting a person rather than filing a request. That is a striking thing to spend nine figures on, and it says something true: the hard part of an assistant stopped being capability a while ago. Everyone's underlying models are good. Very few products are pleasant to delegate to. **Distribution through existing channels keeps proving out.** Poke's traction came from living where people already type. Cognition put the figure at more than 100 million messages exchanged in the three months before the acquisition. No app to install, no new habit, no dashboard — the same bet we made with iMessage, Slack, WhatsApp, and the phone. **Consumer and work assistants are separating.** Poke was built for one person's life. Devin is a tool for engineers. Merging the personality of the first into the second is a coherent strategy, but it is not a strategy about team workflows, approvals, or audit. That gap is the category question of the next year, and it is unrelated to how good any of these products are at their own job. ## The questions this should add to your checklist If you are evaluating any assistant right now — ours included — an acquisition in the category is a good prompt to ask the boring durability questions: 1. **Who is the product for now?** Not at launch. A consumer product inside an enterprise company, or vice versa, is not doomed, but its roadmap now has to win an internal argument it did not have to win before. 2. **What happens to your data and integrations if the product pivots?** Ask about export. Ask specifically what you get back and in what format. 3. **Is the thing you're buying the model, or the interface?** If it is the interface, that is a real asset — the deal just proved it. It is also the layer most easily rebuilt by whoever owns the distribution. 4. **Does it fit how your team is accountable?** Personality is what makes an assistant usable. Approvals and an [audit trail](/glossary/ai-agent-audit-trail) are what make it deployable at a company. Products optimized purely for one person's convenience tend not to have the second, because a single user has no one to be accountable to. ## Where we sit Arlo is built for the team case: it runs across iMessage and SMS, Slack, WhatsApp, and live phone calls, holds sends and tool writes for a named reviewer, and logs every run. That is a different product from a personal texting assistant, and for one person's calendar and errands a personal assistant is often the better buy. We have said that on the [Arlo vs Poke](/compare/arlo-vs-poke) page since before this deal and it is still true. If you were leaning toward Poke and the acquisition made you want to see the field again, the [Poke alternatives](/compare/poke-alternatives) roundup covers the options honestly, including where they beat us. For what an assistant costs to actually run once you're past the shortlist, we published [our own per-task numbers](/blog/what-ai-assistant-work-costs). Details in this post reflect public reporting as of August 2026 and will age. Check Cognition's and Poke's own announcements for the current state. ## Frequently asked questions **Who acquired Poke?** Cognition, the AI lab behind the Devin coding agent, acquired The Interaction Company — Poke's maker — in an announcement dated 23 July 2026, in a deal reported in the low nine figures. **Is Poke shutting down?** No. Cognition said Poke continues to operate and gains its models and infrastructure. The stated rationale for the deal was to bring Poke's conversational style into Devin. **Why would a coding-agent company buy a texting assistant?** For the interaction design. Cognition's stated reason was Poke's conversational style — evidence that in a market where the underlying models are broadly comparable, the interface and personality are what is scarce. **Should the acquisition change which AI assistant I buy?** Not on its own. It is a reason to ask durability questions of any vendor: who the product is for now, what data export looks like, and whether the tool fits how your team approves and audits work. --- # What a unit of AI assistant work actually costs > Measured on Arlo in August 2026, a back-and-forth message costs about 8¢, a drafted document about 48¢, a researched answer with sources about 72¢, a built app about $2.64, a minute of live phone call about 11¢, and a minute of meeting notes about 2.5¢. The spread matters more than any single figure: a researched answer costs about nine times a message, which is what a flat per-seat price hides. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/blog/what-ai-assistant-work-costs Almost every AI assistant is sold by the seat or by the month, which is a pricing model that answers the vendor's question (what is this account worth?) and not yours (what will the work cost me?). We meter Arlo per task, so we have the second number. Here it is. These are our own measured rates as of August 2026, on our own system. They are not an industry benchmark, and another product built differently will land somewhere else. What we think travels beyond our pricing is the *shape*: the spread between task types is large, consistent, and almost entirely invisible under a flat monthly fee. ## What one task costs | Task | Cost | |---|---| | A back-and-forth message | ~$0.08 | | A minute of meeting notes | ~$0.025 | | A minute of live phone call | ~$0.11 | | A generated image | ~$0.10 | | A drafted email or document | ~$0.48 | | A researched answer with sources | ~$0.72 | | An app or dashboard, built | ~$2.64 | The cheapest routine unit is a minute of meeting notes at about two and a half cents. The most expensive is a built app at $2.64 — roughly 105 times more. Even setting aside app-building as a special case, a researched answer costs about nine times what a message costs. ## Why the spread is the whole story A conversational turn is cheap because it is mostly one model call over context you already have. A researched answer is expensive because it fans out: searches, fetches, reading, discarding most of what it read, then writing with citations. Building an app is expensive for the same reason, several times over. This is why "unlimited" and per-seat plans behave strangely in practice. If everyone on your team chats, the vendor's margin is enormous. If three people discover deep research and run it forty times a week, the economics invert — and the usual response is a rate limit, a slower model swapped in behind the scenes, or a quiet cap. You experience that as the product getting worse. You are actually watching a pricing model fail. We have made both mistakes ourselves. An earlier version of our free tier metered real usage against a cap denominated in retail dollars rather than cost, which is the sort of error that only shows up once someone uses the product hard. ## What this means when you are comparing tools Three questions get you further than a price page: 1. **What does the work you actually do cost — not the average user's work?** If your team's job is research, price the research. A tool that is cheap for chat and expensive for research is not cheap for you. 2. **What happens when you hit the ceiling?** Ask specifically whether you get a hard stop, a slower model, or a surprise invoice. All three exist in the market; only one of them is honest, and it is the one that stops and tells you. 3. **Does unused capacity expire?** Monthly allowances that reset punish uneven months, which is most months. Ours roll over for twelve. ## The honest limits of these numbers Per-task costs move. Model prices fall, context strategies change, and a re-measure can shift any row in that table — ours have moved before and will again, which is why every figure on this site is generated from one rate table rather than typed into the copy. Treat the ratios as more durable than the absolute cents. And a caveat about the comparison: our costs are ours. We are publishing them because per-task numbers are almost impossible to find in this category, not because they settle which product is cheaper for you. If you want to check us, the [pricing page](/pricing) derives every figure from the same table this post does. For the work these numbers pay for, see [AI research agent](/use-cases/ai-research-agent) and [AI operations agent](/use-cases/ai-operations-agent). For where the category is heading on pricing generally, [what is AEO](/blog/what-is-aeo) covers a related measurement problem: what it costs to be *found* by the systems people now ask. ## Frequently asked questions **How much does an AI assistant cost per task?** On Arlo as of August 2026: about 8¢ for a back-and-forth message, 48¢ for a drafted document, 72¢ for a researched answer with sources, $2.64 for a built app, 11¢ per minute of live phone call (about 7.5¢ since late September), and 2.5¢ per minute of meeting notes. Other products priced per seat do not usually publish per-task figures. **Why do AI assistants cost so different amounts per task?** Because the work is different. A chat turn is roughly one model call over existing context. A researched answer fans out into many searches and fetches, reads far more than it uses, and then writes with citations — so it costs about nine times a message on our system. **Is per-seat or usage-based pricing better for AI assistants?** Per-seat is predictable and hides the variance; usage-based tracks what you actually consume. The risk with per-seat is what happens at the ceiling — rate limits, quieter model downgrades, or caps — because a flat fee has to protect its margin somehow. **What should I ask a vendor about AI pricing?** What your specific workload costs rather than the average, what happens when you hit the limit, and whether unused capacity expires at the end of each month. --- # What is AEO (answer engine optimization)? > Answer engine optimization (AEO) is the practice of structuring content so AI systems — Google's AI Overviews, ChatGPT, Perplexity, Claude — quote it when answering a question. It differs from SEO in what counts as winning: SEO wins a position in a list of links, AEO wins a sentence inside an answer. A page can rank on page one and still be absent from the answer above it. Updated: 2026-08-06 · Canonical: https://www.arlo.sh/blog/what-is-aeo Answer engine optimization is what you do to get quoted *inside* an AI-generated answer, rather than listed underneath one. The distinction sounds academic until you watch it happen to a page you own: you rank fifth, an AI Overview sits above you summarizing the same question, and it cites three other sites. Your ranking didn't change. Your traffic did. We run a content library of 60-odd pages and watch it closely, so most of what follows comes from our own measurements rather than from general advice. Where we're describing something we observed on one site over a few weeks, we say so — that's a small sample, not a law. ## AEO vs SEO: the difference that matters SEO optimizes for a **position**. AEO optimizes for a **quotation**. They overlap — you generally have to be crawled, indexed, and reasonably well-regarded before anything will quote you — but the last step is different, and the last step is where most pages lose. | | SEO | AEO | |---|---|---| | Unit of success | A ranked link | A cited sentence | | What's rewarded | Relevance + authority | Extractable, verifiable claims | | Winning format | Whatever the SERP rewards | Direct answers, procedures, tables | | How you lose | Rank below the fold | Rank fine, get summarized around | | Measurement | Clicks, impressions, position | Citations, brand mentions in answers | The uncomfortable part is that they can move in opposite directions. An answer engine that quotes you accurately may *cost* you the click, because the reader got what they needed. That's the zero-click trade, and it's worth taking mostly for the queries where being the cited authority compounds — not for the ones where you needed the visit. ## What actually gets cited: mechanism, not pitch The single most useful thing we've learned is that **AI answers quote the page that explains how something works, not the page that says how good it is.** We checked a commercial query in our own category — an AI assistant that takes actions in Slack with human approval. Google's AI Overview framed the whole category around human-in-the-loop control and cited three vendors. All three citations were **documentation pages** that walked through the approval step by step. Our page ranked organically on the same query and was not in the answer. It was well-written marketing prose about the same feature. The fix isn't more persuasive writing. It's converting the claim into a procedure a model can lift: - **Not:** "Arlo keeps you in control with best-in-class governance." - **But:** "Anything that sends or writes pauses for a named reviewer. Standing permission is scoped to a specific action and carries an expiry; when it lapses, the action fails closed rather than proceeding." The second version can be quoted as an answer to "how does approval work in an AI assistant." The first can't be quoted as an answer to anything. That's the whole discipline in one example. ## Answer engines are not one audience "AI search" is four or five different systems with different diets, and they fail differently: - **Google AI Overviews** draw heavily on pages already ranking, so classic SEO is the entry fee. They appear and disappear per query — see below. - **ChatGPT search** leans on Bing's index. If Bing barely knows you, ChatGPT effectively cannot find you, no matter how good your Google position is. This is the most common blind spot we see: teams measure Google, and are invisible on the surface their buyers actually use. - **Perplexity** favors pages with clear structure and visible sourcing, and shows its citations, which makes it the easiest place to test whether your content is quotable. - **Claude and Gemini** blend retrieval with training knowledge, so consistent, repeated, factually stable claims across the open web matter more than any single page. The practical consequence: **check your Bing coverage before you write another word.** It's the cheapest AEO diagnostic there is, and for many sites it's the binding constraint. ## AI Overviews are not a fixed feature of a query Worth knowing before you plan around them. On 2026-07-31 we recorded an AI Overview on a commercial query in our category. On 2026-08-06 we checked the same query and there was no AI Overview at all — the results had reverted to ten links dominated by the platform's own domain. Nothing we did caused that, and we can't tell you which way it will go next week. The lesson is about strategy, not about that query: **don't build a plan whose success depends on an AI Overview continuing to exist**, and don't conclude a cluster is unwinnable because one appeared today. Re-check the actual result page each time you plan a round. Ours changed the diagnosis for a whole cluster in six days. ## Ranking without being read Positions don't tell you whether anyone arrived. Across our own library we currently have several pages sitting in the top ten — including one at position 2.6 — that earn essentially no clicks. When we dug in, the causes were mundane rather than mysterious, and they're worth checking on your own site in this order: 1. **The page is broken.** Our sharpest "great position, no clicks" cases turned out to be URLs that had started returning 404. Google was still ranking pages that no longer existed. Check status codes for your top-ranked URLs before theorizing about intent. 2. **The query is anonymized long-tail.** One page had 158 impressions at position 5.6 and *zero* named queries in Search Console — every impression came from a query too rare to report. There's often no title fix for this; it's a long tail behaving normally. 3. **The searcher wanted a different format.** Ranking for a term whose result page is all listicles, with a definition page, produces impressions and no clicks. 4. **Something answered above you.** The zero-click case people expect, and in our experience the least common of the four. ## A practical AEO checklist Ordered by what moved things for us, most first: 1. **Answer the question in the first 40–60 words**, in plain declarative sentences, before any preamble. This is the block most likely to be lifted verbatim. 2. **Write mechanisms as numbered procedures.** Steps, conditions, and what happens on failure. 3. **Add a real FAQ section** with questions phrased the way people ask them, and answers that stand alone out of context — then mark it up with `FAQPage` schema so the pairing is machine-readable. 4. **Make claims checkable.** Dates, numbers, named limits. "Checked August 2026" beats "currently." 5. **Say where you're the wrong choice.** Honest limits are quotable in a way that superlatives aren't, and they're what a comparison-shaped question actually needs. 6. **Keep structured data truthful.** Wrong prices in JSON-LD are worse than none — that block is exactly what gets repeated back as fact. We shipped a wrong price in ours for months and it took three cleanup passes to fully remove. 7. **Publish an `llms.txt`** listing your main pages in plain text. It's an emerging convention rather than a standard, and it's cheap. 8. **Fix Bing.** For most sites this is an authority problem — referring domains — not a content problem, and no amount of writing substitutes for it. ## What AEO can't fix If nothing links to you, AEO won't rescue you. Authority is still the gate: answer engines are drawing from indexes that rank by, among other things, who else vouches for a page. We can measure this on our own site — a competitor beating us on our best query has roughly 85 referring domains to our one, and no amount of well-structured prose has closed that gap. So treat AEO as a **multiplier on reach you already have**, not a substitute for earning it. Structure your best pages to be quotable, and spend the rest of the effort on the reasons anyone would cite you at all. ## Frequently asked questions **What does AEO stand for?** AEO stands for answer engine optimization — structuring content so AI systems like Google's AI Overviews, ChatGPT, Perplexity, and Claude quote it when they answer a question, rather than only ranking it as a link. **Is AEO different from SEO?** Yes, though they overlap. SEO competes for a position in a list of links; AEO competes for a sentence inside a generated answer. You usually need the SEO foundation first, because most answer engines draw from pages that already rank or are already indexed. **What kind of content do AI answers actually cite?** In our own checks, documentation-style pages that explain a mechanism step by step get cited over marketing pages making the same claim. Convert assertions into procedures — what happens, in what order, and what happens when it fails. **How do I get cited by ChatGPT?** ChatGPT's search draws largely on Bing's index, so start by checking your coverage in Bing Webmaster Tools. If Bing has few referring domains for your site, ChatGPT effectively cannot surface you regardless of your Google rankings. **Does AEO reduce my traffic?** It can. If an answer engine quotes you accurately, the reader may never click. That trade is worth taking on questions where being the cited source compounds your authority, and worth avoiding on questions where you needed the visit to convert. **Do I need an llms.txt file?** It's cheap and harmless, so it's worth adding, but treat it as an emerging convention rather than a standard — no major answer engine currently guarantees it will read one. It is not a substitute for being crawlable and well-linked. --- # Which AI assistants does ChatGPT's model name? We asked 36 questions > Across 36 buyer questions asked on 23 September 2026, ChatGPT-style search most often named general assistants: ChatGPT itself (18 answers), Gemini (14), Zapier (11), Claude and Motion (10 each). Among assistants you text, Poke and Arlo appeared in 7 answers each, Pally in 4 and Homer in 2. Two questions, booking a doctor's appointment and staying on top of follow-ups, named no product at all. Updated: 2026-09-23 · Canonical: https://www.arlo.sh/blog/which-ai-assistants-chatgpt-recommends A growing share of people ask ChatGPT which assistant to use instead of searching for one. Our own signups from ChatGPT answers went from none in July to several a week in September. So we asked it ourselves. We wrote 36 questions the way a buyer would, about texting an assistant, WhatsApp, phone calls, scheduling, reminders, work, errands and alternatives to popular products, and counted every product the answers named. We make Arlo, one of the products in the count. We've reported our own numbers the same way as everyone else's, including where they're weak. ## How we did it We sent each question once to OpenAI's API, using the gpt-5.5 model at low reasoning effort with its web search tool, on 23 September 2026. That is the closest programmatic stand-in for ChatGPT's search. It isn't identical: there's no account, no memory and no personalization, and answers change from one run to the next. Treat a difference of one or two answers as noise. We counted a product as named if it appeared in the answer's text, from a list of 47 AI assistants and tools. We also checked each answer's bolded names for products the list missed; each of those, such as Selectic, Dana and Relevance AI, appeared once or twice and wouldn't change the top of the table. When a question asked about a product, as in "alternatives to Poke", we didn't count that product in its own answer. The full question list is at the bottom. ## Who got named | Product | Answers that named it (of 36) | Named first | |---|---|---| | ChatGPT | 18 | 8 | | Gemini | 14 | 3 | | Zapier | 11 | 4 | | Claude | 10 | 1 | | Motion | 10 | 0 | | Lindy | 9 | 2 | | Reclaim | 9 | 1 | | Poke | 7 | 1 | | Arlo | 7 | 1 | | Fyxer | 6 | 0 | | Microsoft Copilot | 5 | 0 | | Clockwise | 5 | 0 | | Pally | 4 | 1 | | Siri | 4 | 1 | | Meta AI | 3 | 1 | A typical answer named about four products from our list. ## What stood out **ChatGPT recommends ChatGPT.** It named itself in half the answers, even for questions about texting an assistant or running errands. General assistants (ChatGPT, Gemini, Claude, Copilot) showed up across almost every category, often as the first suggestion. **Ask about texting and the small players appear.** For the five questions about assistants you text, the answers named ChatGPT four times, Poke three times, and Arlo, Pally, Siri and Gemini twice each, with Homer leading one answer. Ask a general question instead, like "the best personal AI agent", and none of them appear. **Scheduling goes to calendar tools, including one that no longer exists.** Both scheduling questions named Reclaim. One also named Calendly, Motion, Clockwise and Copilot, and the other named email scheduling agents such as Schedulera and Sarra. Clockwise shut down in March 2026, yet it appeared in five answers across the study. No texting assistant appeared for scheduling, even though several can schedule. **Errands go to specialists.** Cancelling subscriptions and negotiating bills went to Rocket Money, with DoNotPay mentioned once. **Two questions got no product at all.** "Can an AI book a doctor's appointment for me?" got general advice, and "an AI assistant that helps me stay on top of follow-ups" got an offer to do it itself. The real-estate question went to specialist tools we hadn't listed, led by Structurely and Quinn. "Is there an AI that waits on hold for you?" named Google Pixel's Hold for Me first, then ClawCall, Never Hold, Vibrato and No Sitting On Hold. ## What it read All but three answers cited sources. The sites cited most often were arlo.sh (8 answers), OpenAI's help center, pally.com and Google's support pages (6 each), and zapier.com (5). Our own site's position says more about the question list than about Arlo. We've written pages that answer exactly these questions, so it was likely to be read. The more useful pattern is that answers leaned on product pages, help centers and a handful of comparison articles, and on Reddit less often than we expected (2 answers). ## If you're choosing an assistant Ask the specific question. "What AI can I text to manage my calendar and email?" produced a very different list from "the best personal AI agent". Check the sources the answer cites, since they're often the products' own pages. And don't read one answer as a ranking. We ran each question once, and a second run would shuffle the order. ## What it told us about Arlo Arlo was named in 7 of 36 answers and first in one, the question about alternatives to Poke. It appeared for iMessage and iPhone texting, for phone calls and for a daily briefing. It didn't appear for WhatsApp, scheduling, errands, Slack or founder questions, even though Arlo does several of those things. Answers most often named it as the work-focused option next to Poke, Pally and Homer. We'll run the same questions again and publish the change. ## The 36 questions Texting: the best AI assistant to text on iMessage; an AI you can text like a person to get things done; the best SMS assistant for Android users; an AI to text to manage your calendar and email; the best iPhone assistant that works over text with no app. WhatsApp: the best WhatsApp assistant that does real tasks; whether ChatGPT is still on WhatsApp and what to use instead. Phone calls: an AI that makes phone calls for you; an AI that calls a business and books an appointment; an AI that waits on hold for you; whether an AI can book a doctor's appointment. Scheduling: the best assistant to schedule a meeting with several people; an AI that finds a time for a group and sends the invite. Reminders: the best assistant that texts you reminders; an assistant that messages you first with a daily briefing. Work: the best Slack assistant that takes actions; the best AI executive assistant for a startup founder; the best AI chief of staff for a small company; the best AI assistant for a small business owner; an AI that follows up with leads. Errands: an AI that cancels subscriptions; an AI that negotiates bills; the best personal assistant for errands and life admin; the best personal AI agent in 2026. Alternatives: alternatives to Poke, Instinct, Meta's Muse, Lindy, Martin and Pally. General: assistants that act in Gmail and Google Calendar; the most private personal assistant; an assistant that asks before sending emails; the cheapest assistant that actually does tasks; an assistant for real estate agents that works over text; an assistant that helps you stay on top of follow-ups. ## Frequently asked questions **Which AI assistant does ChatGPT recommend most?** In our test of 36 buyer questions on 23 September 2026, ChatGPT-style search named ChatGPT itself most often (18 answers), followed by Gemini (14), Zapier (11), and Claude and Motion (10 each). **Does ChatGPT recommend itself?** Often. It named ChatGPT in 18 of our 36 answers, including questions about texting assistants and errands, and it was the first product named in 8 of them. **Which AI assistants you can text does ChatGPT recommend?** In the five texting questions, the answers named ChatGPT most often (4), then Poke (3), then Arlo, Pally, Siri and Gemini (2 each). Homer came first in two answers, about reminders and iPhone texting. **How did you test this?** We sent each question once to OpenAI's gpt-5.5 model with web search on 23 September 2026 and counted the products each answer named, excluding the product a question asked about. It approximates ChatGPT's search but isn't identical, and results vary between runs. --- # AI assistant vs workflow automation — the difference > A conversational AI assistant like Arlo reasons per request, handles ambiguity, and works across your tools on demand, bounded by approvals and an audit trail. Workflow-automation and RPA tools like Zapier, Make, and UiPath are pre-built flowcharts you configure once — "when X, do Y" — running the same steps every time. Automation wins for deterministic pipelines; a colleague wins for judgment. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/ai-assistant-vs-workflow-automation "AI assistant" and "workflow automation" get lumped together because both promise to take work off your plate. But they're different tools for different problems. Workflow automation is a flowchart you build once and run forever. A conversational AI assistant is a [colleague](/glossary/ai-colleague) you message, that reasons about each request in the moment. Knowing which one you actually need saves a lot of wasted setup — and a lot of "why didn't the automation handle this edge case?" ## What workflow automation is Tools like Zapier, Make, and UiPath are automation and RPA (robotic process automation) platforms. As of 2026, the model is the same across them: you define a trigger and a sequence of actions, and the platform runs those exact steps every time the trigger fires. "When a form is submitted, add a row to the sheet and post to Slack." "When an invoice lands, extract the fields and update the ledger." UiPath and similar RPA tools extend this to clicking through legacy software with no API — recorded, deterministic screen automation. Builder-style tools like [Lindy](/compare/arlo-vs-lindy) sit a step further along: you describe an outcome in plain English and it assembles an agent, but you still design, configure, and maintain the flow before turning it loose. What all of these do well is the same thing: they run a defined process, at volume, exactly the same way, without a human in the loop. That's their strength. It's also their limit. A flow only handles the cases you anticipated when you built it. Anything ambiguous, novel, or slightly off-script falls through — or breaks — until someone updates the flow. ## What a conversational AI assistant is A conversational AI assistant works the other way around. You don't build a flow; you send a message. It's an [AI agent](/glossary/ai-agent) that reads the request, figures out the steps itself, and does them — then adapts when the next request is different. Arlo is this kind of assistant: you reach it in iMessage and SMS, Slack, WhatsApp, or on a live phone call, and it reasons per request instead of replaying a fixed script. Because there's no pre-built flowchart, it handles the messy parts a flow can't: - **Ambiguity.** "Follow up with the people who didn't reply, but skip anyone we already talked to this week." No trigger config could anticipate that phrasing — the assistant reasons it out. - **Ad-hoc, one-off work.** The tasks you'd never bother automating because they only happen once. You just ask. - **Cross-tool work.** Pull a number from one app, check a calendar, draft a message, and wait for your OK — in a single request, without wiring four integrations together first. Arlo connects to 3,000+ tools, and for software with no API it logs in once through a secure browser session and clicks, reads, and fills forms the way a person would. It remembers context across conversations and sends one morning briefing of what changed overnight. ## Governance replaces rigid rules Here's the key part people miss. A flowchart is "safe" because it can only ever do the steps you drew. An AI assistant reasons freely, so it needs a different kind of guardrail — not rigid rules, but review. With Arlo, governance is on by default: every send and every tool write pauses for a reviewer you pick, and every tool call, source, and approval lands in a full [audit trail](/glossary/ai-agent-audit-trail) you can read back later. You get the flexibility of judgment with a human in the loop and a record you can defend — which is what makes it safe to point at real work. ## When each one wins | Approach | Best for | Trade-off | |---|---|---| | Workflow automation (Zapier, Make) | High-volume, repetitive, deterministic pipelines where every run is identical | Only handles the cases you built; brittle on anything ambiguous or new | | RPA (UiPath and similar) | Screen-driven automation of legacy software with no API, at scale | Setup-heavy and fragile when the underlying UI changes | | Agent builders (Lindy) | Teams that want to design and maintain their own autonomous agents | You still build, configure, and babysit the flows | | Conversational AI colleague (Arlo) | Judgment calls, ad-hoc requests, and cross-tool work you'd otherwise do by hand | Not the tool for a fixed pipeline running thousands of identical times a day | The honest read: these aren't really competitors so much as different shapes of solution. If you have a stable, high-volume pipeline — the same three steps a thousand times a day — automation is the right, cheaper answer, and you should reach for Zapier or Make. If the work needs a judgment call, changes shape every time, or spans tools in ways you'd otherwise stitch together by hand, that's where a conversational AI colleague earns its place. Many teams run both: automation for the deterministic plumbing, an AI colleague for everything that needs a brain. ## Where Arlo fits Arlo is built for the second job. It's a colleague you talk to across iMessage, [Slack](/use-cases/ai-assistant-in-slack), WhatsApp, and live calls, that reasons per request, works across 3,000+ tools, and gates every send behind an approval with a full audit trail. It won't replace a well-tuned Zapier flow that moves rows between apps all day — and it isn't trying to. It replaces the pile of ad-hoc, cross-tool, judgment-heavy work that no flowchart ever captured, the stuff you were still doing by hand. For a closer look at how Arlo compares to a builder-style automation platform, see [Arlo vs Lindy](/compare/arlo-vs-lindy). ## Frequently asked questions **What is the difference between an AI assistant and workflow automation?** Workflow automation is a flowchart you configure once — a trigger and a fixed sequence of actions that runs identically every time. A conversational AI assistant reasons about each request as it arrives, figures out the steps itself, and adapts when the next request is different. Automation is deterministic and cheap at volume; an assistant handles ambiguity and one-off work a flow was never built for. **Can an AI assistant replace Zapier or Make?** Not for what those tools are best at. If you have a stable, high-volume pipeline running the same steps thousands of times a day, Zapier or Make is the right and cheaper answer. An AI colleague replaces the ad-hoc, cross-tool, judgment-heavy work no flowchart ever captured. Many teams run both. **Is an AI assistant the same as RPA?** No. RPA tools like UiPath record deterministic screen automation for legacy software, and they break when the underlying UI changes. Arlo also works with software that has no API — it logs in once through a secure browser session and clicks, reads, and fills forms — but it reasons about the goal each run instead of replaying a recorded script. **If an AI assistant isn't limited to pre-built steps, what keeps it safe?** Review rather than rigid rules. A flowchart is bounded because it can only do what you drew; an assistant is bounded by governance. With Arlo, every send and tool write pauses for a reviewer you pick, and every tool call, source, and approval lands in a [full audit trail](/glossary/ai-agent-audit-trail) you can read back later. **When is workflow automation the better choice?** When the work is genuinely repetitive and deterministic — the same three steps a thousand times a day, with no judgment call and no edge cases you care about. Paying a reasoning model to re-derive an identical process every run is slower and more expensive than a flow that just executes it. ## Try Arlo If your work needs judgment, handles ambiguity, and spans tools you'd otherwise juggle by hand — with an approval before every send and a full audit trail — [try Arlo](/signup). --- # AI assistants that can make phone calls for you (2026) > As of September 2026 several AI assistants will call for you. Arlo places calls for work, asked by text, Slack or WhatsApp, and keeps the transcript. Google's Ask for Me calls local businesses about prices and availability. Meta's Muse and Instinct call businesses in beta or early access. Catch, Pally and Martin also call. ChatGPT doesn't place phone calls. Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/ai-assistants-that-can-make-phone-calls A lot of real work still only happens on the phone. The vendor who never answers email, the clinic that confirms by call, the restaurant with no booking page. Until recently that was where AI assistants stopped. That changed fast in 2026. Google now calls local businesses for you, Meta's Muse and Instinct both added calling in September, and several texting assistants place calls too. They don't all do the same thing, though. Some call a business to ask a question and email you the answer. Some hold a full conversation on your behalf. A few answer calls as well as place them. This page sorts them out as of September 2026. We make one of them (Arlo), so read our section with that in mind; the rest comes from each product's own site and dated reporting. ## Who can actually call, and how | | What the call is for | Answers calls too? | How you ask | Availability | |---|---|---|---|---| | **Arlo** | Work and errand calls: confirm, chase, cancel, ask; gets through phone menus and waits on hold | You can call it; it doesn't answer others' calls | Text, Slack or WhatsApp | Anyone; free to start | | **Google Ask for Me** | Asking local businesses about prices and availability | No | Google Search | US, 18+, English | | **Meta Muse** | Calls to US businesses | No | Muse app (phone, Mac, web) or WhatsApp | Beta, US | | **Instinct Concierge** | Restaurants, cancellation lists, cable bills | You can call Instinct | Text or call Instinct | Early access; Instinct is invite-only | | **Catch** | Restaurants, clinics, hotels, airlines | — | Its app and messaging channels | Paid, for executives | | **Pally** | Calls on your behalf, within a weekly usage limit | You can call your Pally | Text | Free for everyone | | **Martin** | Calls to your contacts, plus wake-up calls | You can call Martin | Text, WhatsApp, email, Slack or phone | Subscription | | ChatGPT | Doesn't place phone calls | No | — | — | A dash means the product doesn't say either way. ## Arlo: calls as part of your work Arlo is an [AI colleague](/glossary/ai-colleague) you text in iMessage, SMS or WhatsApp or use in Slack, and the phone is one more place it works. The [AI phone assistant page](/phone) covers it in full. In practice: You text "call the venue and confirm Thursday's booking for 12". Before it dials, Arlo tells you who it's calling and why. Then it places the call, introduces itself as your AI assistant, talks to whoever picks up, and gets the answer; if nobody picks up, it leaves a voicemail. On the call it sticks to the brief you gave it, and it can text you a question mid-call if you told it to check with you. You can also call Arlo yourself and hand it tasks by voice. It doesn't answer other people's calls for you. Calling a company, it presses through the phone menu, waits on hold for up to 90 minutes and talks to the agent when one picks up. After a call it keeps the transcript and logs the whole run in an [audit trail](/glossary/ai-agent-audit-trail). It doesn't text you a recap after every call. Ask "how did the call go?" and it tells you, and it will update your calendar or CRM to match if you ask. Call time comes out of the same credits as everything else, about 30 credits a minute, on plans from $20 a month. The [phone-calling executive assistant](/use-cases/ai-executive-assistant-that-makes-phone-calls) guide walks through a few examples. ## Google Ask for Me: quick questions to local businesses Google's Ask for Me started in Search Labs in January 2025 and began rolling out to all US Search users that July ([Google](https://blog.google/products-and-platforms/products/search/deep-search-business-calling-google-search/)). You tell it what you need, say an oil change quote or whether a groomer has an opening, and Google's AI calls nearby businesses and emails you what it learned, usually within about 30 minutes. It started with categories like auto repair, hair salons, pet groomers, wellness centers and driving schools, and in 2026 Google added home repair, beauty and pet care ([Google](https://blog.google/products-and-platforms/products/search/search-io-2026/)). Its list doesn't include doctors or dentists. It won't make a booking for you, you aren't on the call, and businesses can opt out. It's for signed-in US users aged 18 and over, in English ([Google Search Help](https://support.google.com/websearch/answer/16421135)). Code found in a beta of Google's Phone app points to a future Gemini feature that would call support lines and wait on hold for you, but as of September 2026 Google hasn't announced it. For the options that exist today, see [Is there an AI that waits on hold for you?](/blog/ai-that-waits-on-hold-for-you) ## Meta Muse: calling in beta Meta's Muse agent, launched on 8 September 2026, added calls to US businesses in a beta about a week later, starting with people who had told Muse they wanted the feature ([TechCrunch](https://techcrunch.com/2026/09/17/rival-ai-agents-instinct-and-metas-muse-both-add-the-ability-to-make-calls/)). Reuters later reported that Meta had tested having human contractors place some calls for half its employees; a Meta executive said starting without proper disclosures was "a miss", and the test was rolled back. See [Meta Muse alternatives](/compare/muse-alternatives). ## Instinct Concierge: early access Instinct added Concierge in September 2026, a calling service for jobs like booking restaurants that don't take online reservations, getting on a dentist's cancellation list, or dealing with a cable bill. As of October 2026 it's still in early access for some users, with no published price, and Instinct itself is still invite-only. For doctors and dentists specifically, see [Can an AI book a doctor's appointment for you?](/blog/can-ai-book-a-doctors-appointment) See [Instinct alternatives](/compare/instinct-alternatives). ## Catch, Pally and Martin **Catch** is an assistant for executives. It calls restaurants, clinics, hotels and airlines, works through phone menus and hold, confirms the details with you before it dials, and sends back a summary with the full transcript. It's a single monthly subscription. See [Arlo vs Catch](/compare/arlo-vs-catch). **Pally** made every feature free for everyone on 6 October 2026, calls included: they draw on one usage limit that refills through the week, with no per-minute charge. It works through phone menus and waits on hold. You can call your Pally on the number you text, and every account now gets its own number, as of October 2026. See [Arlo vs Pally](/compare/arlo-vs-pally). **Martin** can call your contacts and speak with them for you, and it does wake-up calls. See [Arlo vs Martin](/compare/arlo-vs-martin). **Mitra** is worth a mention too. Calling is its whole product, and it answers and places calls from your own number. It's sold as monthly bundles of call hours with a seven-day trial. ## Developer voice platforms are a different thing If you searched for AI calling and landed on Bland, Vapi, Retell or Synthflow, those are platforms for businesses to build their own voice agents for their phone lines. They're good at that. They don't call your dentist for you. ## Is it legal for an AI to make phone calls? This isn't legal advice, but the basics are public. In February 2024 the US Federal Communications Commission ruled that AI-generated voices count as "artificial or prerecorded voice" under the Telephone Consumer Protection Act ([FCC 24-17](https://docs.fcc.gov/public/attachments/FCC-24-17A1.pdf)). Those rules require the called party's prior consent for such calls to residential and mobile numbers, with exceptions for emergencies and some other cases, and they require an artificial-voice message to identify who is responsible for the call. Those rules apply to calls to residential and mobile numbers, which includes plenty of businesses that answer on a cell phone. Recording is governed separately. Federal law requires one party's consent to record a call, but some states, including California, Florida, Massachusetts, Pennsylvania and Washington, require everyone on the call to agree. When a call crosses state lines, the safe assumption is that everyone needs to know. ## What to check before you trust one Four questions separate a demo from something you'd hand a real call to. Does it handle the whole call, including hold music, phone menus and voicemail? Does it ask before it commits you to anything? Is there a recording or transcript you can go back to? And can you see a record of what it did when a call mattered? For the category itself, see [AI phone assistant](/glossary/ai-phone-assistant). ## Frequently asked questions **Which AI assistant can make phone calls for me?** For work, Arlo places calls that you ask for by text, in Slack or on WhatsApp, and keeps the transcript. For quick questions to local businesses in the US, Google's Ask for Me calls them for you. Meta's Muse and Instinct call businesses in beta or early access, and Catch, Pally and Martin also place calls. **Can Google call a business for me?** Yes, in the US. Google's Ask for Me, in Search, calls local businesses to ask about prices and availability and sends you the answers, usually within about 30 minutes. It won't make a booking, and you aren't on the call. **Can ChatGPT make a phone call?** No. As of September 2026 ChatGPT doesn't place phone calls for you. Hark Pro, which launched on 6 October 2026, doesn't either, as of October 2026; see [Hark Pro alternatives](/compare/hark-alternatives). **Can an AI assistant answer my calls too?** Some can. Mitra answers calls on your own number. Arlo, Google's Ask for Me and Meta's Muse place calls rather than answering them for you, though you can call Arlo yourself to give it a task. **Are AI phone calls legal?** In the US, the FCC ruled in February 2024 that AI-generated voices fall under the Telephone Consumer Protection Act's rules for artificial voices, which require prior consent for such calls to residential and mobile numbers, with some exceptions, and require the message to identify who is calling. Recording a call needs everyone's consent in some states. This isn't legal advice; check the rules that apply to your calls. **Do AI phone calls sound like a robot?** Mostly not anymore. Current AI voices are close to human in pacing and tone. What matters more is behaviour: saying who it is, staying inside the task, and knowing when to hand the call back to you. --- # AI Chief of Staff vs. a Human: What Each Does Better > An AI chief of staff and a human do different jobs well. An AI like Arlo wins on always-on availability, perfect memory, and executing repetitive work across your tools — scheduling, follow-ups, briefings, and calls. A human wins on judgment, relationships, and ambiguity. Most teams pair them rather than choosing between them. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/ai-chief-of-staff-vs-human "Should I hire a chief of staff or use an AI one?" is really two questions wearing one coat. A human chief of staff and an AI chief of staff are good at genuinely different things, and the honest answer for most founders and operators isn't either/or — it's knowing which work to hand to which. This page lays out where each actually wins, so you can decide without the hype. ## What each does better | | AI chief of staff | Human chief of staff | | --- | --- | --- | | Availability | 24/7, instant | Working hours | | Cost | Flat monthly fee | Six-figure salary | | Memory | Perfect recall across everything | Strong, but human | | Repetitive execution | Tireless, consistent | Capable, but it's a poor use of them | | Acting across your tools | Yes, at scale | Yes, one thing at a time | | Judgment on ambiguity | Limited | Its whole point | | Relationships & trust | No | Yes | | Discretion & politics | No | Yes | | Making phone calls | Yes | Yes | ## Where an AI chief of staff wins The case for an AI isn't that it's smarter than a good human. It's that it never sleeps, never forgets, and doesn't mind doing the same thing a thousand times: - **Always on, flat cost.** It's there at 6am and 11pm, and it costs a fraction of a salary, so you can put it on work that would never justify a hire. - **Perfect memory.** It remembers the account, the deadline, and what you decided last month, across every conversation and tool. - **Execution at scale.** Scheduling, follow-ups, CRM hygiene, research, a morning briefing of what changed — the operational load that a human chief of staff is overqualified for. ## Where a human still wins Be honest about the ceiling, because it's real: - **Judgment in ambiguity.** When there's no clear right answer, a person's read of the situation beats a model's. - **Relationships and trust.** Building rapport with a board member, a key hire, or a difficult partner is human work. - **Discretion and politics.** Reading the room, handling something sensitive, knowing what *not* to put in writing — a human's domain. ## The honest answer: pair them Most teams don't replace a chief of staff with an AI — they use an AI to cover the operational work so a human (or the founder) is freed for the judgment work. That's exactly what an [AI executive assistant](/glossary/ai-executive-assistant) like [Arlo](/glossary/ai-colleague) is built for: - It handles the scheduling, follow-ups, research, and briefings across iMessage, Slack, WhatsApp, and [phone calls](/use-cases/ai-executive-assistant-that-makes-phone-calls). - It acts across 3,000+ tools, and [waits for your approval](/use-cases/ai-assistant-with-approval-before-sending) on anything that sends, writes, or spends — so it takes work off your plate without taking decisions out of your hands. - It keeps a full audit trail, so delegating to it never means losing track of what happened. The parts an AI can't do — the judgment, the relationships, the discretion — stay with a person. The parts it can do, it does around the clock. ## Deciding for your team Early and lean? An AI chief of staff often covers what you actually need before a hire makes sense. Scaling, with real political and relationship complexity? Hire the human and give them an AI so they're not buried in logistics. Either way, see how Arlo fits [for founders](/use-cases/ai-assistant-for-founders) or as a [governed assistant](/use-cases/ai-assistant-with-approval-before-sending), or browse all [comparisons](/compare). ## Frequently asked questions **Can an AI chief of staff replace a human one?** For the operational half of the job, often yes — scheduling, follow-ups, CRM hygiene, research, and a morning briefing of what changed. For judgment on ambiguity, relationships, and discretion, no. Most teams pair them rather than substitute one for the other. **What does an AI chief of staff cost compared to hiring one?** A human chief of staff is a six-figure salary. An AI one is prepaid credits you draw down as you use them — Arlo starts free with $100 in credits for connecting Slack, then sells them at 400 credits per $1 on monthly plans from $20, and $50 covers roughly 625 messages or 69 researched answers. The practical effect is that you can point an AI at work that would never have justified a hire. **What can't an AI chief of staff do?** Three things, honestly. It can't exercise real judgment when there's no clear right answer. It can't build trust and rapport with a board member or a difficult partner. And it can't read the politics of a room or know what shouldn't be put in writing. Those stay with a person. **Can an AI chief of staff make phone calls?** Yes. Arlo places [live phone calls](/use-cases/ai-executive-assistant-that-makes-phone-calls) alongside iMessage, Slack, and WhatsApp, with one memory across all of them — so a thread that starts as a text can finish as a call without re-explaining the context. **How do I delegate to an AI chief of staff without losing control?** Governance rather than trust. Arlo [waits for your approval](/use-cases/ai-assistant-with-approval-before-sending) on anything that sends, writes, or spends, and keeps a full audit trail of every run — so handing over the work doesn't mean handing over the decisions or losing track of what happened. **Should an early-stage founder hire a chief of staff or use an AI?** If you're early and lean, an AI usually covers what you actually need before a hire makes sense. Once there's real political and relationship complexity, hire the human — and give them an AI so they aren't buried in logistics. ## Try Arlo Hand the operational load to an AI and keep the judgment for yourself. [Try Arlo](/signup) — an AI colleague that runs the logistics across your tools, with approvals and a full audit trail. --- # AI colleague vs AI personal assistant — the difference > An AI personal assistant helps one individual with personal tasks over a conversational channel like text or voice. An AI colleague like Arlo works for a whole team across iMessage, Slack, WhatsApp, and phone, takes governed actions with approvals, and leaves a full audit trail. Personal assistants optimize for one person's convenience; colleagues are built for shared, accountable work. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/ai-colleague-vs-personal-assistant "AI assistant" has quietly split into two products that share a name and almost nothing else. One is a personal assistant: a helper for you, over text or voice, that runs your calendar and errands. The other is an [AI colleague](/glossary/ai-colleague): a teammate for a whole group that acts inside shared tools, waits for approval on anything sensitive, and logs what it did. They look similar in a chat window. They're built for completely different stakes. Knowing which one you're actually shopping for saves you from buying a personal toy for a team job, or an enterprise system for a to-do list. ## What an AI personal assistant is An [AI personal assistant](/glossary/ai-personal-assistant) is optimized for one person's convenience. You reach it the way you'd text a friend, and it handles the individual layer of your life: your calendar, your inbox, your reminders, your flight check-ins. As of 2026, products like Poke, Martin, and April define this lane well — signup is often just a phone number, there's frequently no app, and the whole experience is tuned to feel effortless and conversational. That's a real, valuable product. If the job is "help me run my own day," a personal assistant is exactly right, and adding team controls would only get in the way. The design goal is friction-free help for a single user, and the best ones nail it. ## What an AI colleague is An AI colleague answers to a team, not a person. That single shift changes what the product has to do. It has to work where the team already works — not just your texts, but Slack and the phone. It has to take real actions across shared systems. And because those actions land in front of teammates, customers, and sometimes auditors, it has to be governable: someone must be able to approve, review, and reconstruct what happened. [Arlo](/) is built as an AI colleague. As of 2026 it runs across iMessage and SMS, Slack, WhatsApp, and live phone calls; connects to 3,000+ tools and uses no-API software through a secure browser session; keeps persistent memory; and sends one morning briefing of what changed. The part that makes it a colleague rather than a personal bot is the governance: an [approval gate](/use-cases/ai-assistant-with-approval-before-sending) on every send and tool write, and a [full audit trail](/glossary/ai-agent-audit-trail) on every run. ## The real distinction | Dimension | AI personal assistant | AI colleague | | --- | --- | --- | | Who it serves | One individual | A whole team | | Where it works | Text or voice, personal channels | iMessage/SMS, Slack, WhatsApp, phone | | Kind of work | Personal tasks and errands | Shared, cross-tool work | | Actions | Acts for you | Acts, then waits for approval | | Accountability | Optimized for convenience | Governed by design, full audit trail | | Memory | Remembers your preferences | Persistent memory + morning briefing | | Controls | Personal settings | Admin policy, roles, SSO (on request) | The table isn't a scorecard — a personal assistant "losing" the accountability row is a feature, not a flaw, because a single user doesn't need to audit themselves. The point is that these are answers to different questions. One optimizes for how little you have to think; the other optimizes for how much a team can trust it to act. ## Who each is for Choose a **personal assistant** if the assistant serves exactly one person and the tasks are personal: your calendar, your inbox, your reminders, your travel. You want it light, cheap, and out of the way. Team controls would be dead weight. Choose an **AI colleague** when the assistant acts on behalf of a group and the work carries stakes — a founder's operations, a sales team's outreach, a brokerage's client comms. The moment an assistant sends something a customer sees, writes to a shared system, or spends money, "who approved this?" and "what exactly did it do?" stop being optional questions. A colleague is the product built to answer them. ## Why a team wants a colleague, not a personal assistant You can hand a personal assistant to everyone on a team, but you'll feel the seams fast. There's no shared context across people, no way to approve an action before it goes out, and no record of what the assistant did in anyone's name. That's fine for personal errands and genuinely risky for team work. An AI colleague inverts the defaults: it assumes actions may need review, so sensitive steps pause for a reviewer you name, and everything it does — every tool call, source, and approval — lands in a trace you can read back later. That's why Arlo treats governance as the base layer, not an upsell. Connections clear policy before each run, sends and writes wait on an approver, and the audit trail is always on. It's also why buying it scales for organizations: you pay for credits at 400 per $1 on monthly plans from $20, the credits last 12 months, and admin policy, roles, SSO, and audit export are available on request. For the head-to-head version of this distinction, see [Arlo vs Poke](/compare/arlo-vs-poke), and for the category framing read the [AI colleague](/glossary/ai-colleague) and [AI executive assistant](/glossary/ai-executive-assistant) glossary entries. The full [comparisons hub](/compare) lines up the specific tools. ## Frequently asked questions **What is the difference between an AI colleague and an AI personal assistant?** An [AI personal assistant](/glossary/ai-personal-assistant) serves one individual with personal tasks over a channel like text or voice. An [AI colleague](/glossary/ai-colleague) answers to a team: it works across iMessage, Slack, WhatsApp, and phone, acts inside shared systems, pauses for approval on anything sensitive, and logs what it did. **Can I just give everyone on my team a personal assistant instead?** You can, but you'll feel the seams quickly. There's no shared context between people, no way to approve an action before it goes out, and no record of what the assistant did in anyone's name. That's fine for personal errands and genuinely risky for shared work. **Is a personal assistant worse than an AI colleague?** No — they answer different questions. A personal assistant "missing" approvals and audit trails is a feature, because a single user doesn't need to audit themselves. Adding team controls to a personal assistant would only add friction to the job it's built for. **Which one do I need?** Choose a personal assistant if exactly one person is served and the tasks are personal — your calendar, inbox, reminders, travel. Choose an AI colleague the moment the assistant sends something a customer sees, writes to a shared system, or spends money, because "who approved this?" stops being optional. **Does an AI colleague cost more than a personal assistant?** Not necessarily. Arlo starts free with $100 in credits for connecting Slack, then sells credits on monthly plans from $20 — 400 per $1, and they last 12 months from purchase — with every capability available from the first credit; admin policy, roles, SSO, and audit export are available on request. Personal assistants are often cheaper to get started with, which is the right trade when the job really is one person's errands. **Which products are AI personal assistants and which are AI colleagues?** As of 2026, Poke, Martin, and April define the personal lane well — often no app, signup by phone number, tuned for one person's convenience. Arlo is built as a colleague. For the head-to-head, see [Arlo vs Poke](/compare/arlo-vs-poke). ## Try Arlo If your team needs a colleague — not a personal assistant — that works across your channels with approvals and a full audit trail, [try Arlo](/signup). --- # Arlo vs April — which AI assistant fits your work? > April is a simple AI executive assistant that lives entirely in text messages, built for one person's personal to-dos. Arlo is a governed AI colleague that works across iMessage, SMS, Slack, WhatsApp, and live phone calls, with an approval gate on every send and a full audit trail. April fits an individual; Arlo fits a team. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/arlo-vs-april You text both April and Arlo like a person, but they're built for different jobs. April, from goapril.ai, is a lightweight AI executive assistant that lives entirely in your text messages — no app to install, no dashboard to learn. Arlo is an [AI colleague](/glossary/ai-colleague) for work that runs across the channels your team already uses, takes real actions through governed tool access, and logs everything it does. Want a simple personal EA you can text and forget about? April is a clean, honest choice. Want an assistant your whole team can lean on inside real workflows, with reviewer approvals and a full audit trail? That's the gap Arlo fills. Both are good products; they're just aimed at different people. ## What April is (as of August 2026) As of August 2026, April positions itself as an [AI executive assistant](/glossary/ai-executive-assistant) that lives in the one app everyone already has open: Messages. From its public materials: - It works entirely over text messages, with nothing to install and no separate app to open. - It's aimed at individuals who want a simple, minimalist personal assistant. - Its pitch is ease: you text it the way you'd text a human EA, and it handles the back-and-forth. - Public pricing lists a free start (around the first 50 messages) and then a flat monthly plan (about $25/month) for unlimited use. Public detail on April is limited, so treat the specifics above as its stated positioning rather than an exhaustive feature list. What April clearly does well is keep things dead simple: one thread, one person, no setup. If that's what you want, check April's own site for the current state before you decide. ## What Arlo is Arlo is one AI colleague for the whole workspace, and there's no separate chat app to live in. It works inside the tools a team already uses: iMessage and SMS, Slack, WhatsApp, and live phone calls. It remembers context across conversations and sends a single morning briefing of what changed overnight. It connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools. For software with no API, it logs in once through a secure browser session and clicks, reads, and fills out forms the way a person would. So the work isn't limited to what a chat integration exposes. Governance is the part that sets it apart, and it's on by default. Every send and every tool write pauses for a [reviewer you pick](/use-cases/ai-assistant-with-approval-before-sending). Each tool call, source, and approval lands in a full [audit trail](/glossary/ai-agent-audit-trail) you can go back and read. That's a different posture than a personal texting assistant, and it's deliberate. ## Feature comparison | Capability | April (personal) | Arlo (work / teams) | |---|---|---| | Primary audience | Individual consumers | Teams and businesses | | Where you reach it | Text messages only | iMessage and SMS, Slack, WhatsApp | | Live phone calls | Not a stated focus | Yes — it places calls for you, and you can call it | | Persistent memory | Not a stated focus | Yes, plus a morning briefing of changes | | Integrations | Not a stated focus | 3,000+ tools | | No-API tool access | Not a stated focus | Yes — secure browser sessions | | Approval before sending | Not a stated focus | Yes — every send and write awaits a reviewer | | Audit trail | Not a stated focus | Full trace of calls, sources, approvals | | Admin policy / roles / SSO | Not a stated focus | Yes, on request | | Pricing | ~50 messages free, then ~$25/mo (as of 2026) | Free to start · plans from $20/mo · 400 credits per $1 · credits last 12 months | This table reflects each product's stated positioning as of August 2026. "Not a stated focus" for April means its public materials emphasize a simple personal EA over text, not that a capability is impossible. ## Which one is right for you Pick **April** for yourself. One clean text thread that handles your personal errands, scheduling, and reminders, with nothing to configure and no controls to babysit. If simplicity over a single channel is the whole point, that's exactly what April is for. Pick **Arlo** when the assistant acts for a team and the work carries stakes. Its reach across Slack, WhatsApp, and the phone, plus its governance model, are built for shared, accountable work. When an assistant sends an email to a client, updates a CRM, or makes a call on your behalf, you want the approval step and the record — and that's the mode Arlo runs in by default. On pricing, both are straightforward but shaped differently. April is a flat personal subscription as of August 2026. Arlo is a monthly plan from $20 that buys credits at 400 per $1, with top-ups up to $50,000, and the credits roll over and expire 12 months after purchase rather than each month; admin policy, roles, SSO, and audit export are available on request for orgs that need them. ## Governance is the real difference A personal assistant optimizes for one user's convenience. An AI colleague has to hold up in front of teammates, customers, and auditors, and that changes what the product has to do. Arlo treats every action as something that might need review: sensitive steps wait on an approver you name, and the audit trace records what happened and why. That's why you can point Arlo at real, shared workflows and not just personal to-dos — and it's the honest line between the two products. Still weighing options? See our head-to-head [Arlo vs Poke](/compare/arlo-vs-poke), our roundup of the [best AI executive assistants](/compare/best-ai-executive-assistants), or the [AI executive assistant](/glossary/ai-executive-assistant) explainer. ## Frequently asked questions **What is the difference between Arlo and April?** April is a minimalist AI executive assistant that lives entirely in text messages, built for one person's personal to-dos. Arlo is a governed AI colleague across iMessage and SMS, Slack, WhatsApp, and live phone calls, with an approval gate on every send and a full audit trail. April fits an individual; Arlo fits a team. **Is April cheaper than Arlo?** For light personal use, often yes. As of August 2026 April starts with about 50 free messages and no card, then a flat monthly subscription — around $25/month when we last checked, though April doesn't publish the rate prominently, so confirm on its own pricing page. Arlo starts free for teams — connecting Slack grants $100 in credits (40,000) — then runs on monthly plans from $20 at 400 credits per $1 — $50 is roughly 625 messages; unused credits stay good for 12 months instead of lapsing each month. **Can April work in Slack?** Not as a stated focus — April's public materials describe a single text-message channel with no app to install, which is the point of the product. Arlo works in [Slack](/use-cases/ai-assistant-in-slack), [WhatsApp](/whatsapp), iMessage and SMS, and on live phone calls, with one memory across all of them. **Can April make phone calls?** Making and taking calls is not a stated focus of April's public materials. Arlo places [live phone calls](/use-cases/ai-executive-assistant-that-makes-phone-calls) and keeps a transcript of each one. **Should I choose April or Arlo?** Choose April if the assistant serves exactly one person over one channel and you want nothing to configure. Choose Arlo when the assistant acts for a team and the work carries stakes — when it emails a client, updates a CRM, or spends money, you want the approval step and the record. **Is April's feature list really that limited?** No — "not a stated focus" in the table above means April's public materials emphasize a simple personal EA over text, not that a capability is impossible. Public detail on April is limited, so check April's own site for the current state before you decide. ## Try Arlo If you need an AI colleague that works inside your team's tools with an approval before every send and a full audit trail, [try Arlo](/signup). --- # Arlo vs. Catch > Arlo and Catch are both AI assistants that work across chat and phone calls. Catch focuses on admin — email, scheduling, and outbound calls across Slack, WhatsApp, and iMessage. Arlo adds 3,000+ integrations including apps with no API, and deeper governance: an approval gate plus a full audit trail on every action, not just a consent prompt. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/arlo-vs-catch Catch is one of the few assistants that, like Arlo, refuses to live in a single channel — it works over chat and it makes phone calls. So if you've found both, you're comparing two products in the same rare category rather than a text-bot against a workflow tool. The differences are real, but they're about reach and governance, not about which one is "an AI assistant." Everything below reflects each product's public description as of August 2026. Short version: Catch is a polished admin assistant for execs; Arlo is a governed AI colleague built to span a whole team's tools, including the long tail of apps that don't have an API. ## At a glance | | Catch | Arlo | | --- | --- | --- | | Email (Gmail / Outlook) | Yes | Yes | | Slack | Yes | Yes | | iMessage / SMS | Yes | Yes | | WhatsApp | Yes | Yes | | Phone calls | Outbound | Outbound, and you can call it | | Integrations | 25+ | 3,000+ (incl. no-API via browser) | | Asks before acting | Yes | Yes | | Full audit trail | Not stated | Yes | | Pricing (as of August 2026) | $99/mo flat, 10-day free trial | Free to start · plans from $20/mo · 400 credits per $1 · credits last 12 months | ## What Catch is good at Catch pitches itself as an admin assistant for busy leaders — "your admin savior." As of 2026 it handles inbox zero, drafts email in your voice, schedules and reschedules meetings, fills out forms, and sends a morning brief, and it can place outbound calls to sort out logistics ("I'll call and try to arrange early check-in"). It works across Slack, WhatsApp, iMessage, Gmail/Outlook, and your calendar, and connects to 25+ tools like HubSpot, Notion, and Asana. It's consent-based: by its own description it "acts only with the right context and permission" and asks when unsure. If your need is one person's admin load and you want WhatsApp, Catch is a strong, focused pick. ## Where Arlo goes further Arlo is an [AI colleague](/glossary/ai-colleague) built for a team, not just an individual inbox, and that shows up in two places: - **Reach across your stack.** 25+ integrations covers the common tools; Arlo connects to 3,000+, and for the internal systems that never shipped an API, it logs in once through a secure browser session and operates them by hand. The work that used to fall outside an assistant's reach comes back in. - **Governance you can audit.** Both assistants ask before acting, which is the right default. Arlo goes a step past the consent prompt: every connection resolves through policy before a run, and every tool call, source, and approval lands in a [full audit trail](/glossary/ai-agent-audit-trail) you can read back. "It asked me" is good; "here's the logged record of exactly what it did and who approved it" is what an org needs before it lets an assistant act on its behalf. Both can [make phone calls](/use-cases/ai-executive-assistant-that-makes-phone-calls); you can also call Arlo yourself, and it keeps a transcript of each call. ## How to choose Pick **Catch** if you want a sharp personal admin assistant for one executive. Pick **Arlo** if you want one governed colleague across a whole team, with wide tool reach and an audit trail on everything it does. If you're weighing other options too, see [Arlo vs. Martin](/compare/arlo-vs-martin) and the full [comparisons](/compare). Weighing the wider field too? See our roundup of [Catch alternatives](/compare/catch-alternatives) and the [best AI executive assistants](/compare/best-ai-executive-assistants). ## Frequently asked questions **What is the difference between Arlo and Catch?** Both work across chat and phone calls, which is rare. As of 2026 Catch is a polished admin assistant for one executive across Slack, WhatsApp, iMessage, and email, with 25+ integrations. Arlo adds 3,000+ integrations including apps with no API, and a full audit trail on every action rather than a consent prompt alone. **Does Arlo work with WhatsApp?** Yes. You can text Arlo on WhatsApp as well as in iMessage and SMS, and use it in [Slack](/use-cases/ai-assistant-in-slack) and on phone calls. Catch works in WhatsApp too, so the channel alone isn't a reason to pick either. **Both assistants ask before acting — how is Arlo's governance different?** Asking is the right default and Catch does it. Arlo goes a step further: every connection resolves through policy before a run starts, and every tool call, source, and approval lands in a [full audit trail](/glossary/ai-agent-audit-trail) you can read back. "It asked me" is good; a logged record of exactly what it did and who approved it is what an organization needs before letting an assistant act on its behalf. **Which one makes phone calls?** Both. As of 2026 Catch places outbound calls to sort out logistics. Arlo places [outbound calls](/use-cases/ai-executive-assistant-that-makes-phone-calls) and keeps a transcript of each one; you can also call it yourself. **How does the pricing compare?** As of August 2026 Catch is a flat $99/month with voice calls included and a 10-day free trial — no seats, credits, or per-call fees. Arlo starts free, then runs on monthly plans from $20: credits at 400 per $1, drawn down by what you actually use — calls at about 30 credits a minute, a message at about 32 — and unused credits last 12 months. Admin policy, roles, SSO, and audit export are available on request. ## Try Arlo Arlo is the multi-channel assistant for teams that need deep integrations and a governed audit trail — not just a consent prompt. [Try Arlo](/signup) and connect the first tool you'd hand off. --- # Arlo vs ChatGPT — chat app or AI colleague in your tools? > ChatGPT is a general-purpose AI chat app you go to — open it, ask, get an answer. Arlo is an AI colleague that comes to you: it works inside iMessage, Slack, WhatsApp, and live phone calls, holds persistent memory, and takes governed actions across 3,000+ tools with approvals and a full audit trail. Use ChatGPT to think; use Arlo to get work done where your team already talks. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/arlo-vs-chatgpt Almost everyone reading this already uses ChatGPT. It's the default way to ask an AI anything, and for that job it's excellent. The comparison with Arlo isn't "which chatbot is smarter" — it's a difference in kind. ChatGPT is a destination: you open the app, have a session, and carry the results back to your actual work yourself. Arlo is an [AI colleague](/glossary/ai-colleague): it lives in the channels your team already uses, remembers what happened last week, and does the work in your tools — with your approval on anything that matters. If your question is "which one should I pay for," the honest answer depends on what's slowing you down. If it's thinking — drafts, analysis, code, brainstorming — ChatGPT is hard to beat. If it's the coordination work between your tools — the follow-ups, updates, scheduling, and status-chasing that eat your team's day — that's the job Arlo was built for. ## What ChatGPT is (as of July 2026) As of July 2026, ChatGPT is OpenAI's flagship chat product across web, desktop, and mobile. From its public materials: - It's a conversational app: you bring questions, files, and images to it, and work in sessions. - Paid tiers add heavier models, longer context, and features like scheduled tasks, projects, connectors to popular apps, and deep research runs. - An agent mode can operate a browser to complete multi-step tasks you watch over. - Memory features carry preferences and facts between your own sessions. It's a phenomenal reasoning tool with an enormous ecosystem. Product details shift quickly, so check OpenAI's site for the current state. ## What Arlo is Arlo is one AI colleague for the workspace. There's no separate app to visit: it works inside iMessage and SMS, Slack, WhatsApp, and on live phone calls — it can place real calls and keep the transcript. It connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools. For software with no API, it logs in once through a secure browser session and clicks, reads, and fills forms the way a person would. It keeps persistent memory across every conversation and channel, and sends one morning briefing of what changed overnight. The part that makes it safe to hand real work to is governance, on by default. Every connection resolves through policy before each run. Sends, drafts, and writes wait on a reviewer you pick. Every run leaves a full [audit trail](/glossary/ai-agent-audit-trail) — each tool call, source, and approval, reviewable after the fact. ## Feature comparison | Capability | ChatGPT | Arlo | |---|---|---| | Primary job | Answers and artifacts in a chat session | Completed work inside your tools | | Where it lives | Its own app (web, desktop, mobile) | iMessage/SMS, Slack, WhatsApp, phone calls | | Works in your team's channels | — | ✓ | | Makes live phone calls | — | ✓ | | Persistent cross-channel memory | Per-account session memory | ✓ | | Proactive (briefings, follow-ups) | Scheduled tasks | ✓ Morning brief + suggested automations | | Acts in 3,000+ work tools | Connectors on paid tiers | ✓ Governed, with approvals | | Tools with no API | — | ✓ Secure browser session | | Approval gates before sends/writes | — | ✓ On by default | | Audit trail of every action | — | ✓ | | Team-level rollout | Enterprise plans | ✓ Shared workspace credits; admin controls on request | ## Where ChatGPT wins Be fair to the default: ChatGPT is the stronger pick for open-ended thinking. Long research sessions, writing you'll shape by hand, coding help inside an IDE-adjacent workflow, image work, and one-off questions with no follow-through. There's no setup, everyone already knows it, and the free tier is genuinely useful. If nobody on your team needs an AI to actually touch your tools, ChatGPT may be all you need. ## Where Arlo wins Arlo wins when the work lives across your stack and your channels. Nobody has to remember to open an app: you text Arlo from the thread you're already in, or @mention it in Slack, and it carries the task through — reads the CRM, drafts the email, waits for your approval, sends, logs the call. It's the difference between an assistant you consult and a [colleague you delegate to](/glossary/ai-employee). And because approvals and the audit trail are defaults rather than add-ons, you can give it real access without holding your breath. ## The honest answer: most teams use both ChatGPT for thinking, Arlo for doing is a common and sensible split. The mistake is expecting one to do the other's job — pasting ChatGPT output into six tools by hand every day, or using an action-taking colleague as a general search box. If you're choosing where a budget goes for a team, count the hours spent on coordination between tools; that's the line Arlo removes. ## Frequently asked questions **Can ChatGPT work inside Slack or iMessage?** As of July 2026, ChatGPT is primarily its own destination app. It doesn't natively sit inside your team's Slack channels or your text threads the way [Arlo does](/use-cases/ai-assistant-in-slack) — its connectors reach into tools during your session rather than living in your channels. **Can ChatGPT make phone calls for me?** No — placing and receiving live phone calls on your behalf isn't part of the product as of July 2026. Arlo places real calls and keeps the transcript, so you can ask how it went. **Is Arlo a replacement for ChatGPT?** Not exactly. Arlo replaces the coordination work between tools, not the open-ended thinking sessions. Many teams keep ChatGPT for exploration and use Arlo as the colleague that executes across their stack. **Which is safer to connect to work tools?** Arlo is built around governed access: policy checks before every run, approval gates on sends and writes, and a full audit trail. If a connection can touch customers or data, those defaults matter more than model choice. **How much does each cost?** ChatGPT's tiers are listed on OpenAI's site, and its free tier is genuinely useful. Arlo starts free too — connecting Slack grants your workspace $100 in credits (40,000) — then you buy credits at 400 per $1 on monthly plans from $20 with 10% bonus credits, plus top-ups up to $50,000, and they expire 12 months after purchase rather than each month. At that rate $50 is about 625 messages, 69 researched answers with sources, or 18 built apps; admin policy, roles, and SSO are available on request. --- # Arlo vs Hark Pro — an assistant you text, or an app you open > Hark Pro, from Brett Adcock's AI lab Hark, is a free iPhone, Android and web app for your personal life, with a proactive home screen and a browser agent for errands. Arlo is an AI colleague for work that you text on iMessage, SMS or WhatsApp or use in Slack; it places calls, asks before it sends or changes anything unless you've pre-approved it, and keeps an audit trail. Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/arlo-vs-hark Hark Pro, from Brett Adcock's AI lab Hark, launched to everyone on 6 October 2026 ([Hark](https://hark.com/articles/introducing-hark-pro)). It's an app you open: one long conversation, a home screen that suggests what to do next, and a browser agent that runs errands on sites like DoorDash, Target and OpenTable. Arlo is an [AI colleague](/glossary/ai-colleague) you text for work, in the threads you already use. Both act for you across your accounts. They differ on where you reach them, who they're for, and what they do best. This page covers Hark Pro from hark.com; the Hark Audio app and the Hark app for Shopify stores are separate products. This comparison is written by Arlo. Hark's facts come from its launch post, its terms and privacy policy (both updated 1 October 2026), its app-store listings and dated press, as of October 2026, and we say where Hark Pro is the better pick. ## What Hark Pro is (as of October 2026) - **Made by** Hark Labs, Inc. in San Jose, California, founded by Brett Adcock, who also founded Archer Aviation and Figure AI and still runs Figure. Its head of design, Abidur Chowdhury, presented the iPhone Air at Apple before joining. - **Backed by** about $100 million of Adcock's own money and a $700 million-plus Series A at a $6 billion valuation in May 2026 ([Hark](https://hark.com/articles/hark-announces-usd700-million-fundraising-round)), with NVIDIA and AT&T as partners. A Hark device is due in 2027 ([TechCrunch](https://techcrunch.com/2026/10/06/hark-releases-an-ai-personal-assistant-with-a-focus-on-privacy/), 6 October 2026). - **Where you use it:** apps for iPhone and Android, and the web at hark.com. Hark hasn't announced a number to text, WhatsApp, or phone calls, though its terms include consent wording for automated texts. - **How it works:** one long-running thread, Projects for separate pieces of work, and Handoff, Hark's computer-use agent, which gets its own cloud computer for each task and can run up to six browsers at once. It connects to Google, Outlook and Slack, and to other services through MCP connectors. - **Proactive Home:** cards built from your email, calendar and connected services, one-tap Action Buttons (pay, cancel, RSVP, review), and Panels, small apps you describe in words. - **Approvals:** it asks before it spends your money or sends a message for you unless you've told it otherwise, and you can grant standing permission for a category of purchases. - **Price:** every feature free with a usage cap Hark hasn't published, then $20 a month for twice the usage or $100 a month for ten times, on the web. App-store prices are higher. Details are in [Hark Pro pricing and alternatives](/compare/hark-alternatives). ## What Arlo is Arlo is an assistant you text on [iMessage](/imessage), SMS or [WhatsApp](/whatsapp), use inside [Slack](/slack), or hand a [phone call](/phone) to. It places calls to people and businesses, and it can work through phone menus and wait on hold. It connects to Gmail, your calendar, Slack, Notion, Linear, HubSpot and 3,000+ other tools, and reaches sites with no API through a secure browser session you sign into once. Anything it sends or changes waits for your approval unless you've pre-approved that kind of action, and every tool call, source and approval is written to an [audit trail](/glossary/ai-agent-audit-trail) you can read back. It [texts you first](/proactive) with briefs and reminders over iMessage and SMS. Anyone can start today, with no invite or waitlist. It's free to start, no card, then plans from $20 a month. ## Feature comparison | | Arlo | Hark Pro | |---|---|---| | Where you use it | iMessage, SMS, WhatsApp, Slack | iPhone, Android and web apps | | Phone calls | Places calls to people and businesses, through phone menus and hold | None announced | | Built for | Work: founders, operators, teams | One person's life | | Proactive | Texts you briefs and reminders | A home screen of suggested tasks, Action Buttons and Panels | | Acts across your accounts | 3,000+ tools, plus a browser session you sign into once | Google, Outlook, Slack and MCP connectors, plus the Handoff browser agent | | Checks with you before acting | Holds anything it sends or changes, unless you've pre-approved that kind of action | Asks before it spends or sends, unless you've told it otherwise; standing permission per category | | Record of what it did | Audit trail of every tool call, source and approval | A live view of the agent browsing, and purchase confirmations | | Training on your content | Not your messages, files, memories or calls, per its privacy policy | Its privacy policy says no; its terms allow use to improve the service | | Price | Free to start, no card, then plans from $20 a month | Free with a usage cap; $20 or $100 a month on the web for more | The Hark Pro column reflects its public pages and press as of October 2026. Hark Pro launched on 6 October, so expect it to change quickly. ## Where Hark Pro is the better pick - **Design.** Hark hired Abidur Chowdhury from Apple to lead design, and launch reviews praised the result. [Gizmodo](https://gizmodo.com/hark-pro-ai-ui-2000822253) called the app polished, and Gizmodo and [Fast Company](https://www.fastcompany.com/91618831/how-hark-designed-an-ai-assistant-thats-actually-helpful) both singled out how its proactive cards are presented. If you want an assistant you enjoy opening, Hark Pro is ahead. - **A proactive home screen.** Hark's Home turns what's in your email and calendar into cards with a task ready to run, and Action Buttons let you pay, cancel, RSVP or review in one tap. Fast Company's example was Hark spotting a rental car booked at the wrong airport. Arlo's briefs arrive as plain texts, with nothing visual. - **A browser agent built for errands.** Handoff gives each task its own cloud computer with a browser, file system and terminal, and runs up to six browsers at once. Hark reports the top score on the Online-Mind2Web benchmark, measured with its own test harness ([Hark](https://hark.com/articles/introducing-hark-handoff), August 2026). - **Consumer errands.** DoorDash and Uber Eats orders, checkouts at Target, Walmart and Costco, OpenTable and Resy bookings, airline and Booking.com reservations, bills and subscription cancellations are Hark's home turf. Arlo is built for work. - **The free tier.** Every Hark Pro feature is free up to a usage cap, and $20 a month doubles the usage. Arlo is free to start, then runs on plans from $20 a month. - **Funding and pace.** More than $800 million raised, NVIDIA and AT&T as partners, and a device due in 2027. Hark announced its lab in March 2026 and shipped Hark Pro to everyone in October. ## Where Arlo is the better pick - **You want to text it, or have it make the call.** Arlo works in iMessage, SMS, WhatsApp and Slack, and it [places calls](/compare/ai-assistants-that-can-make-phone-calls) to people and businesses, working through phone menus and waiting on hold. Hark Pro is an app and a website, and Hark hasn't announced texting or phone calls. Hark's terms already include consent wording for automated texts, so check again later. - **The work involves other people.** Arlo works inside a team's Slack and connects to the tools a company shares: Gmail, Notion, Linear, HubSpot and 3,000+ others. Hark Pro is sold to one person for their own life, and we found no team plan or admin controls. - **You want a record you can read back.** Both ask before spending or sending by default, and both let you widen that. Arlo also writes every tool call, source and approval to an audit trail. Hark shows a live window of the agent browsing and asks you to confirm purchases, but we couldn't find a published activity log. - **Your tools have APIs.** Arlo connects to 3,000+ tools. Hark names a short list of integrations and relies on its browser agent for the rest. For Hark Pro's pricing in full, its privacy terms, and the other options (Pally, Muse, Instinct and Poke), see [Hark Pro alternatives](/compare/hark-alternatives). ## Frequently asked questions **What is the difference between Arlo and Hark Pro?** Hark Pro is a free iPhone, Android and web app for your personal life, with a proactive home screen and a browser agent for errands. Arlo is an AI colleague for work that you text on iMessage, SMS or WhatsApp or use in Slack. It places calls, holds what it sends or changes for your approval unless you've pre-approved that kind of action, and keeps an audit trail. **Hark Pro vs Arlo: which should I use?** For errands in your own life, such as food delivery, shopping, reservations and travel, Hark Pro is the stronger pick. Arlo is the better pick for work that involves other people, when you'd rather text it or hand it a phone call than open an app, and when you want approvals and an audit trail. **Is Hark Pro free?** Yes. As of October 2026 every Hark Pro feature is free with a usage cap. Hark Pro² ($20 a month) and Hark Pro³ ($100 a month) buy two and ten times the usage on the web, and app-store prices are higher. Arlo is free to start, no card, then plans from $20 a month. **Can you text Hark Pro, or have it make phone calls?** As of October 2026 Hark hasn't announced either; Hark Pro is an app and a website. Arlo places calls to people and businesses, works through phone menus and waits on hold, and you text it on iMessage, SMS or WhatsApp. **Can I use Hark Pro for work?** It reads Gmail, Outlook, calendars and Slack, and it can message and recruit candidates on LinkedIn, so it helps with your own work tasks. It's sold to individuals, though, and we found no team plan or admin controls. Arlo is built for work with a team: it works inside Slack and keeps an audit trail of every action and approval. **Which is more private?** Their policies say similar things. Hark's privacy policy says it doesn't train on your inputs, outputs, memory or connected-account data, though its terms let it use your content to improve the service. Arlo's [privacy policy](/privacy) says it doesn't use your messages, files, memories, calls or connected-account content to train AI models. The difference is the record: Arlo keeps an audit trail of every tool call, source and approval, and we couldn't find a published activity log for Hark. **What are the alternatives to Hark Pro?** Besides Arlo, there's Pally, which you text and which is free; Muse, Meta's agent, in the US and Canada; Instinct, which is invite-only; and Poke. See [Hark Pro alternatives](/compare/hark-alternatives). --- # Arlo vs Instinct AI — two AI assistants you text that act for you > Arlo and Instinct AI are both assistants you text that take real actions and make phone calls. Arlo is built for work across iMessage, SMS, WhatsApp, Slack and calls, asks before it sends unless you've pre-approved that kind of action, keeps an audit trail, and anyone can start it today. As of October 2026 Instinct is invite-only, free, and licensed for personal use. Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/arlo-vs-instinct Instinct AI, the AI assistant from Spear Street Technology, and Arlo answer the same question: what if you could text an assistant and it simply did the thing? Both place calls, work across your accounts, and keep going between your messages. They differ on who they're for and on how much you get to see and approve along the way. Instinct is built for personal life. Arlo is an [AI colleague](/glossary/ai-colleague) built for work, where the email goes out in a company's name and somebody may need to check later what happened. This comparison is written by Arlo. Instinct's facts below come from its own site, its terms of service (revised 26 August 2026 and unchanged since) and dated reporting, as of October 2026, and we say where it's the better pick. ## What Instinct is (as of October 2026) - **Made by** Spear Street Technology, Inc. in San Francisco, founded by Noah Shinn. On 28 September 2026 it raised a $1 billion Series C at a $10 billion valuation from Sequoia, Benchmark and Coatue, a month after its $2.5 billion Series B, bringing its total funding to about $1.35 billion ([TechCrunch](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/)). - **Channels:** you text it on iMessage, SMS or WhatsApp, or call it, and it can text or call you first. Each Instinct has its own email address. Group chats, where friends without an account can join in, arrived in early access on 5 October 2026 ([TechCrunch](https://techcrunch.com/2026/10/05/instinct-brings-its-ai-agent-to-group-chats-even-for-friends-without-an-account/)). Calls it places to businesses for you, through its Concierge service, are still limited to some early-access users ([Fox News](https://www.foxnews.com/tech/ai-agents-make-phone-calls), 3 October 2026). - **How it works:** its own phone number and its own computer in the cloud, which uses the logins you give it and keeps working between messages. It has no mobile app; you text it, and its settings and data controls are on the web. - **Access and price:** invite-only early access with a waitlist, and each member gets five invites to share ([Fortune](https://fortune.com/2026/09/30/noah-shinn-instinct-ai-assistant-meta-muse-alexandr-wang-tech-series-c-ai-agent-mark-zuckerberg/), 30 September 2026). It's free, with no published pricing, and its terms allow for fees later. - **Who it's for:** built for personal life. It coordinates with friends' Instincts and in group chats, but has no team workspace, admin controls or business plan, and its terms license it "for your personal use only". - **Shopping:** purchases go through Stripe Link, which creates a one-time card after you approve the purchase ([Simon Taylor](https://x.com/sytaylor/status/2093637347510309331), 29 August 2026). Since a Shopify partnership on 28 September it finds products across Shopify stores and sets up the order for you to check out with Shop Pay ([PYMNTS](https://www.pymnts.com/news/artificial-intelligence/2026/shopify-opens-store-checkouts-to-ai-agents/)). On 29 September it launched Instinct Selections, curated product recommendations, and some users objected to the unsolicited suggestions ([TechCrunch](https://techcrunch.com/2026/09/30/instincts-new-product-recommendations-are-giving-some-users-the-ick/), 30 September 2026). ## What Arlo is Arlo is an assistant you text on [iMessage](/imessage), SMS or [WhatsApp](/whatsapp), use inside [Slack](/slack), or hand a [phone call](/phone) to. It places calls to people and businesses, and can work through phone menus and wait on hold. It connects to Gmail, your calendar, Slack, Notion, Linear, HubSpot and 3,000+ other tools, and reaches sites with no API through a secure browser session you sign into once. Anything it sends or changes waits for your approval unless you've pre-approved that kind of action, and every tool call, source and approval is written to an [audit trail](/glossary/ai-agent-audit-trail) you can read back. It [texts you first](/proactive) with briefs and reminders over iMessage and SMS. It's free to start, with no card, then runs on monthly plans from $20. ## Feature comparison | | Arlo | Instinct | |---|---|---| | Can you start today? | Yes | Invite only: five invites per member, plus a waitlist | | Built for | Work — founders, operators, teams | Personal life; coordinates with friends' Instincts and in group chats, with no team workspace or admin controls | | Where you talk to it | iMessage, SMS, WhatsApp, Slack | iMessage, SMS, WhatsApp; group chats in early access | | Phone calls | Places calls to people and businesses, through phone menus and hold | You can call it and it can call you; calls to businesses in early access | | Acts across your accounts | 3,000+ tools plus browser sessions | A cloud computer using your logins | | Checks with you before acting | By default, before anything is sent or changed, unless you've pre-approved that kind of action | Asks when it needs input; purchases need your approval through Stripe Link, and you check out Shopify orders yourself with Shop Pay | | Record of what it did | Audit trail of every tool call, source and approval | Records available through the service; its terms say they "may not always be accurate" | | Price | Free to start, then plans from $20 a month | Free, no published pricing | The Instinct column reflects its public site, terms and dated reporting as of October 2026. Both products change quickly. ## Where Instinct is the better pick - **Your own errands, at no cost today.** If you have an invite and the job is personal — groceries, tickets, a weekend away, a bill dispute — Instinct is built for exactly that, and it's free. - **Shopping and travel.** Shinn says users put close to $1 billion a year through it, about half of it travel ([Fortune](https://fortune.com/2026/09/30/noah-shinn-instinct-ai-assistant-meta-muse-alexandr-wang-tech-series-c-ai-agent-mark-zuckerberg/), 30 September 2026). It pays with one-time Stripe Link cards, sets up Shopify orders for Shop Pay checkout, and handles high-end bookings through Concierge in early access. - **Plans with friends and family.** Your Instinct can coordinate with the Instincts of people you trust, join group chats where friends don't need an account (in early access), and share itineraries and expense splits as pages anyone can open. - **Hands-off autonomy.** Instinct is designed to carry a long errand on its own with few check-ins, and users credit it with noticing things unprompted, like a calendar conflict. If you want to hand something off and hear back when it's done, that's its whole design. - **Pace.** It's among the best-funded products in the category and ships new capabilities most weeks. ## Where Arlo is the better pick - **You can start now.** No invite, no waitlist. Instinct members get five invites each, and everyone else waits. - **The work involves other people.** Arlo runs inside Slack as well as your texts and connects to the shared tools a company uses. Instinct coordinates with friends' Instincts and in group chats, but it has no team workspace, admin controls or business plan, and its terms license it "for your personal use only". - **You want to approve what it does, and check it later.** Arlo holds anything it sends or changes for your approval unless you've pre-approved that kind of action, and keeps an audit trail you can read back. Instinct is designed to act with fewer check-ins. It asks before purchases and before sharing anything into a group chat, and otherwise checks in when it needs your input. Users have reported it sending an innocuous email "without checking with me first" ([TechCrunch](https://techcrunch.com/2026/08/24/instincts-powerful-ai-assistant-is-raising-privacy-and-security-concerns/), August 2026) and reading a one-time login code from Gmail without asking ([Business Insider via Yahoo](https://tech.yahoo.com/ai/meta-ai/articles/hallucinations-2fa-prompts-hidden-access-172455494.html), September 2026). - **You want a price you can plan around.** Arlo publishes its [pricing](/pricing). Instinct has none yet, its terms allow fees later, and it hasn't said how it will make money. For the rest of the field — Pally, Muse, Poke, Martin and OpenClaw — see [Instinct AI alternatives you can use today](/compare/instinct-alternatives). For how Instinct compares with Meta's agent, see [Instinct vs Muse](/compare/instinct-vs-muse). ## Instinct's safety record so far Instinct's terms appoint it as your agent and say commitments it makes "shall be binding on you as if entered into directly by you", without guaranteeing that its safeguards prevent mistakes. The public incidents so far, by date: - **24 August 2026:** early users told [TechCrunch](https://techcrunch.com/2026/08/24/instincts-powerful-ai-assistant-is-raising-privacy-and-security-concerns/) about an email sent without checking first, and about emails kept after Gmail was disconnected. Instinct later added a way to delete them. - **23 September 2026:** a user posted what looked like a stranger's financial document. Shinn said it was a hallucination: "This was not a data breach, no user data was shared." He announced a hallucination-detection layer the same day ([Shinn on X](https://x.com/noahrshinn/status/2102896837522804954); [Business Insider via Yahoo](https://tech.yahoo.com/ai/deals/articles/instincts-founder-says-viral-ai-063054528.html), 24 September 2026). - **24 September 2026:** Business Insider reported that Instinct read a Luma one-time login code from a user's Gmail without asking, used it to cancel RSVPs, and then misdescribed what it had done ([summary via Yahoo](https://tech.yahoo.com/ai/meta-ai/articles/hallucinations-2fa-prompts-hidden-access-172455494.html), 25 September 2026). Instinct trains on your materials unless you opt out in its settings, and disconnecting an account doesn't delete what it indexed; deletion is a separate request. Arlo's [privacy policy](/privacy) says it doesn't use your messages, files, memories, calls or connected-account content to train AI models. ## Frequently asked questions **What is the difference between Arlo and Instinct AI?** Both are AI assistants you text that take real actions and make phone calls. Instinct is built for personal life, runs a cloud computer for your errands, and as of October 2026 is invite-only and free. Arlo is built for work across iMessage, SMS, WhatsApp, Slack and phone calls, holds what it sends for your approval unless you've pre-approved that kind of action, keeps an audit trail, and anyone can start it today. **How much does Instinct AI cost?** Nothing for now. As of October 2026 Instinct has no published price, plans or usage limits, and its terms say some features may carry fees later. Arlo is free to start with credits and no card, then runs on monthly plans from $20. **Is Instinct AI legit?** Yes. It's made by Spear Street Technology, founded by Noah Shinn, and on 28 September 2026 it raised $1 billion at a $10 billion valuation from Sequoia, Benchmark and Coatue ([TechCrunch](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/)). It's also early: invite-only, unpriced, and changing week to week. **Is Instinct AI safe?** It acts for you with real authority. Its terms make commitments it enters binding on you and don't guarantee its safeguards catch every mistake. Users have reported it sending an email without checking first and reading a login code from Gmail without asking, and it trains on your materials unless you opt out. Whichever assistant you choose, check whether it asks before sending, keeps a record you can trust, and deletes your data when you disconnect. **Can I use Instinct for work?** Its terms, revised 26 August 2026, license it "for your personal use only". It can coordinate with coworkers' Instincts and has its own email address you can CC, but it has no team workspace, admin controls or business plan. Arlo is built for work, including inside Slack with a team. **Does Instinct ask before it sends things or buys?** For purchases, yes: Stripe Link creates a one-time card only after you approve, and you check out Shopify orders yourself with Shop Pay. For other actions it checks in when it needs input, its terms say confirmations may not catch every mistake, and users have reported an email sent and a login code read without asking. Arlo holds anything it sends or changes for your approval unless you've pre-approved that kind of action. **Is Arlo invite-only like Instinct?** No. Anyone can start Arlo today by texting it, with no invite or waitlist. Instinct members get five invites each, and everyone else joins a waitlist. **Does Instinct AI work on WhatsApp?** Yes. Once you have an invite or a spot off the waitlist, you can text Instinct on WhatsApp as well as iMessage and SMS. Arlo works on WhatsApp too, alongside iMessage, SMS and Slack. **What are the best Instinct AI alternatives?** For work, Arlo. For personal life without an invite: Pally, free with every feature; Muse, Meta's free agent in the US and Canada; Poke; Martin; and OpenClaw if you host it yourself. See [Instinct AI alternatives you can use today](/compare/instinct-alternatives) and [Instinct vs Muse](/compare/instinct-vs-muse). --- # Arlo vs Lindy: AI colleague vs workflow automation > Lindy is a workflow-automation platform where you build AI agents that run business processes. Arlo is a conversational AI colleague you talk to directly in iMessage, SMS, Slack, WhatsApp, and live phone calls, with governed approvals and a full audit trace on every action. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/arlo-vs-lindy The split is simple. With Lindy you build agents and turn them loose. With Arlo you message a [colleague](/glossary/ai-colleague) and it acts under rules you set. Lindy's pitch is "define an outcome, configure an agent, let it run." Arlo's is "text a teammate" in iMessage, SMS, Slack, WhatsApp, or on a phone call. If you enjoy assembling autonomous agents, Lindy is built for you. If you'd rather not open a builder at all, keep reading. ## What Lindy is Lindy (lindy.ai) calls itself an AI agent platform for autonomous digital workers. As of June 2026, the core pieces are an Agent Builder (describe a task in plain English, it assembles an agent for you), Autopilot for clicking through tools that have no API, and team accounts with role-based sharing. People run Lindy agents for email triage, scheduling, CRM updates, and lead qualification, and one agent can hand work to another. Lindy advertised thousands of integrations and, from October 2025, a phone agent; as of September 2026 its site no longer mentions calling. Pricing is credit-based and per seat as of September 2026: no free tier, a 7-day trial through Slack, per-user plans from around $29.99/month up to a Max tier, and custom Enterprise pricing. What Lindy does well is configurable automation. You design the agents and the triggers, then watch them run. ## What Arlo is Arlo isn't a builder. It's a coworker you talk to, and it lives where you already work: - iMessage and SMS - Slack - WhatsApp - Live phone calls It remembers what you've told it, sends a morning briefing of what changed overnight, and connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools. When something has no API, Arlo logs in once through a secure browser session and clicks, reads, and fills forms the way a person would. You never configure a workflow. You send a message and it goes. ## Governance and audit This is where the two products really diverge. With Arlo, governance is on by default. Every connection runs through a policy check before each run, and anything that sends or writes (a draft, an email, a record update) waits for a reviewer you pick. Every tool call, source, and approval lands in an audit trace you can read back later. If you need a human in the loop and a record you can defend, that's the whole reason Arlo works the way it does. ## Feature comparison | Capability | Arlo | Lindy (as of June 2026) | |---|---|---| | Primary model | Conversational AI colleague you message | Workflow-automation / agent-builder platform | | Where you interact | iMessage, SMS, Slack, WhatsApp, live calls | Web app, configured agents, chat/voice channels | | Setup style | Talk to it like a teammate | Build and configure agents and workflows | | Tools without an API | Secure browser session (logs in, clicks, fills) | Computer-use / Autopilot | | Governed approvals | Reviewer approves drafts, sends, writes by default | Role-based team accounts | | Audit trace | Full trace of every call, source, approval | Varies by plan | | Persistent memory + briefing | Yes, with morning briefing | Agents learn preferences | | Pricing | Free to start · plans from $20/mo · 400 credits per $1 · credits last 12 months | Per-seat plans with a 7-day trial; Enterprise custom | ## Which should you choose Go with Lindy if you want to build autonomous agents that run defined processes and you don't mind configuring and maintaining them. Pick Arlo if you'd rather message one colleague across iMessage, Slack, WhatsApp, and the phone, with every send and write gated behind an approval and logged. The conversational model tends to win over founders, realtors, and sales teams who don't want another dashboard to babysit. See [AI assistant for founders](/use-cases/ai-assistant-for-founders), [for realtors](/use-cases/ai-assistant-for-realtors), and [for sales](/use-cases/ai-assistant-for-sales), or read up on the [AI executive assistant](/glossary/ai-executive-assistant) and browse our [resources](/resources). Weighing other options too? See our roundup of [Lindy alternatives](/compare/lindy-alternatives), our [Poke alternatives](/compare/poke-alternatives) overview, and the [Arlo vs Poke](/compare/arlo-vs-poke) head-to-head. On pricing, Arlo starts free — $100 in credits for connecting Slack — then runs on prepaid credits: 400 credits per $1 at every amount, on monthly plans from $20 with top-ups from $50, with credits that stay good for 12 months; admin policy, roles, SSO, and audit export are available on request. ## Frequently asked questions **What is the difference between Arlo and Lindy?** Lindy is a workflow-automation and agent-builder platform: you define an outcome, configure an agent, and let it run. Arlo is a conversational AI colleague you message directly in iMessage, SMS, Slack, WhatsApp, or on a live phone call, with reviewer approvals and a full audit trace on every action. **Do I have to build anything to use Arlo?** No. There is no builder and no flow to configure — you message it like a teammate and it reasons out the steps per request. With Lindy you design, configure, and maintain the agents before turning them loose, which is the point of an agent-builder platform. **Can Lindy use software that has no API?** Yes — as of June 2026 Lindy offers computer-use / Autopilot for that. Arlo does it by logging in once through a secure browser session and clicking, reading, and filling forms the way a person would. **Which is better for a team that needs approvals?** Arlo gates by default: a reviewer approves drafts, sends, and tool writes before they execute, and every call, source, and approval lands in a [full audit trace](/glossary/ai-agent-audit-trail). Lindy offers role-based team accounts, and its audit capability varies by plan. **How does pricing compare?** Differently in kind. Arlo grants your workspace $100 in credits (40,000) for connecting Slack, then sells credits at 400 per $1 on monthly plans from $20 with 10% bonus credits, plus top-ups up to $50,000 — and they roll over and expire 12 months after purchase. Lindy sells per-seat plans with a 7-day trial and custom enterprise pricing; as of September 2026 its published per-user plans run roughly $29.99–$199.99/month. **Should I use Arlo or Lindy?** Pick Lindy if you want to design and maintain your own autonomous agents for defined business processes. Pick Arlo if you'd rather not open a builder at all and want a colleague you talk to, with governance on by default. See also [AI assistant vs workflow automation](/compare/ai-assistant-vs-workflow-automation). ## Try Arlo Want an AI colleague you just talk to across iMessage, Slack, WhatsApp, and live calls, with approvals and an audit trace on everything it touches? [Get started with Arlo](/signup). --- # Arlo vs Martin — which AI assistant fits your work? > Martin is a personal, voice-first AI assistant you reach by call, text, or email to run your calendar, inbox, reminders, and to-dos. Arlo is an AI colleague for work that operates across iMessage, Slack, WhatsApp, and phone calls with governed tool access, reviewer approvals, and a full audit trail. Martin fits individuals; Arlo fits teams. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/arlo-vs-martin Both Martin and Arlo are assistants you talk to like a person. They're built for different jobs. Martin is a personal, voice-first assistant you reach by call, text, or email to run your own calendar, inbox, reminders, and tasks. Arlo is an [AI colleague](/glossary/ai-colleague) for work. It lives in the channels your team already uses, takes real actions through governed tool access, and logs everything it does. The short version: if you want a proactive assistant for your own day, Martin is a strong consumer pick. If you want something your whole team can lean on inside the tools it already runs, with reviewer approvals and admin controls, that's Arlo. ## What Martin is (as of June 2026) Martin (trymartin.com) pitches itself as a personal AI assistant "like Jarvis." As of June 2026, you can reach it by voice, SMS, WhatsApp, email, phone, and Slack, plus an iOS and web app. It handles your calendar, inbox, phone calls, to-dos, reminders, and notes, and it picks up your preferences over time so it can schedule meetings, draft emails, fire off texts, and make calls for you. On integrations, Martin connects to Google Calendar, iCloud Calendar, Gmail, SMS, phone calling, WhatsApp, and Slack. Pricing as of June 2026 is a 7-day free trial, a Basic plan, and a Pro plan that layers on premium models, email pre-drafting, and long-term memory. It's a product for one person managing their own day, and it's clearly designed that way. ## What Arlo is Arlo is [one AI colleague for the whole workspace](/glossary/ai-colleague). It works inside the tools a team already uses — iMessage and SMS, Slack, WhatsApp, and live phone calls — so nobody has to learn a new chat app. It keeps persistent memory, sends a morning briefing of what changed overnight, and connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools. When a tool has no API, Arlo logs in once through a secure browser session and clicks, reads, and fills out forms the way a person would. Governance comes standard, not as an upsell. Every connection runs through policy before each turn, and any draft, send, or write waits on a reviewer you pick. Each tool call, source, and approval lands in a full audit trace. ## Personal voice assistant vs team AI colleague The real split is who the assistant works for. Martin is voice-first and personal, organized around one person's calendar, inbox, and reminders. Arlo takes that same conversational, action-taking feel into shared work channels and wraps it in the controls a team actually needs. In practice that means reviewer approvals before anything goes out, admin policy and roles, and an audit trail showing what happened, why, and who signed off. If you're an individual, that's probably more machinery than you want. If your team shares accounts and touches customer-facing work, it's exactly the part you can't skip. ## Feature comparison | Capability | Martin | Arlo | | --- | --- | --- | | Built for | Individuals (personal productivity) | Teams and businesses | | Primary channels | Voice, SMS, WhatsApp, email, Slack, app | iMessage/SMS, WhatsApp, Slack, phone calls | | Calendar, inbox, reminders | Yes | Yes | | Persistent memory | Yes (Pro) | Yes, plus morning briefing | | Tool connections | Calendar, Gmail, Slack, WhatsApp | Slack, Gmail, Notion, Linear, GitHub + 3,000+ tools | | No-API tools via browser | Not stated | Yes — secure browser session | | Reviewer approvals before actions | Not stated | Yes, by default | | Audit trace of every action | Not stated | Yes | | Admin policy, roles, SSO, audit export | Not stated | Yes, on request | Facts about Martin reflect its public positioning as of June 2026 and may change. ## Which should you choose? Pick Martin if you want a personal, voice-first assistant for your own calendar, inbox, and errands and you don't need shared team controls. It's a focused consumer product with a simple monthly plan. Pick Arlo if you want an [AI assistant for founders](/use-cases/ai-assistant-for-founders), [realtors](/use-cases/ai-assistant-for-realtors), or [sales teams](/use-cases/ai-assistant-for-sales) that takes real actions inside Slack, WhatsApp, iMessage, and calls, with the approvals and audit trail a team needs. Arlo starts free with $100 in credits for connecting Slack, then runs on monthly plans from $20 — 400 credits per $1, good for 12 months; admin policy, roles, SSO, and audit export are available on request. You can also compare [Arlo vs Poke](/compare/arlo-vs-poke) or browse [Martin alternatives](/compare/martin-alternatives) or other [Poke alternatives](/compare/poke-alternatives), and our [resources](/resources). ## Frequently asked questions **What is the difference between Arlo and Martin?** Martin is a personal, voice-first assistant you reach by call, text, or email to run your own calendar, inbox, reminders, and to-dos. Arlo is an AI colleague for work across iMessage and SMS, Slack, WhatsApp, and phone calls, with governed tool access, reviewer approvals, and a full audit trail. Martin fits individuals; Arlo fits teams. **Is Martin better for voice?** Martin is built around talking — it's iOS-first and designed for quick spoken interactions like reminders and scheduling while you're moving. Arlo also places [live phone calls](/use-cases/ai-executive-assistant-that-makes-phone-calls), but voice is one channel among several rather than the organizing idea of the product. **Does Martin connect to as many tools as Arlo?** No. As of June 2026 Martin connects calendar, Gmail, Slack, and WhatsApp. Arlo connects Slack, Gmail, Notion, Linear, GitHub and 3,000+ other tools, and reaches software with no API through a secure browser session. **Does Martin have approvals and an audit trail?** Not as a stated feature in its public materials. Arlo gates every send and tool write behind a reviewer by default and records every tool call, source, and approval in a [full audit trail](/glossary/ai-agent-audit-trail). **Does Arlo support WhatsApp like Martin does?** Yes. Both work over WhatsApp. Arlo also covers iMessage and SMS, Slack, and phone calls. **Which should I choose?** Pick Martin if you want a proactive personal assistant for your own day and you'd rather speak than type. Pick Arlo if the assistant acts for a team inside shared tools and you need approvals, admin controls, and a record of what happened. ## Try Arlo Want one AI colleague your whole team can trust across the tools it already uses? [Start with Arlo](/signup). You get conversational action-taking across iMessage, Slack, WhatsApp, and calls, with governance, approvals, and a full audit trail built in. --- # Arlo vs Pally — cost, channels and approvals compared > Arlo and Pally (pally.com) are both AI assistants you text that take real actions and place calls. As of October 2026 Pally is free for everyone, with every feature and a weekly usage limit, and it is strongest for personal errands and travel. Arlo is built for work: you can chat with it on WhatsApp and in your team's Slack, with approvals and an audit trail of every action. Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/arlo-vs-pally Pally (pally.com) and Arlo look alike from the outside. You text both like a person, both connect to your email and calendar, both place phone calls and can work through phone menus and hold, and both show you what they propose before it goes out. Since 6 October 2026 they differ sharply on price, because Pally is now free for everyone. They also differ on who they're built around. Pally is a personal assistant, one per person, for your errands, travel and inbox. Arlo is an [AI colleague](/glossary/ai-colleague) built for work that other people depend on and someone may need to check later. This comparison is written by Arlo, so read it with that in mind. Pally's facts below come from its own site, terms, privacy policy and changelog as of October 2026, with dated outside sources linked inline, and there are real cases where it is the better pick, starting with price. ## What Pally is (as of October 2026) - **Who makes it:** Pally Technologies, Inc., a Y Combinator company (Summer 2025) founded by Haz Hubble and Wyatt Lansford. It raised a $5.2M seed led by Cyber Fund and Y Combinator at a $30M valuation ([Runtime Wire, 6 August 2026](https://runtimewire.com/article/pally-raised-5-2-million-ai-assistant-text-messages)). - **Price:** free for everyone, with every feature, since 6 October 2026. There is one usage limit, and it refills through the week. Pro and Max were discontinued and refunded pro rata. The only paid plan is VIP, invite-only at $1,000 a month, plus a hosted Cloud Mac add-on with no published price ([terms](https://pally.com/terms), 6 October 2026). - **Channels:** iMessage, Android RCS and SMS, and Telegram, with group chats on iMessage and Telegram. Pally reads your WhatsApp inbox and can send replies there, but WhatsApp isn't a place to chat with it. Its own WhatsApp guide sends people who want that to Poke. - **Its own number and email:** every account gets a phone number and email address of its own, set up the first time you need them. Until 6 October these came only with Max. - **Calls:** included for everyone with no per-minute charge, drawing on the same weekly usage limit. Pally works through phone menus, waits on hold and connects you when a person answers, and you can call your Pally too. - **Actions:** about two dozen named apps, including Gmail, Outlook, iCloud, Google Calendar, Notion, Linear, Slack, GitHub, Jira, Todoist and Granola, plus MCP servers on request. Pally also signs in to almost any website through a browser session and reads bank accounts read-only. - **Errands and travel:** books flights end to end, checks you in and claims fare drops, books hotels and rental cars, and finds restaurants with a real table at your time. Pally pays with a single-use card capped at what you approved (US only for now) or, since 6 October, your saved card after one approval of the exact total. - **Proactive:** a morning brief and, since 4 October, an end-of-day edition, alerts on important messages, and unlimited custom automations, including triggered ones. - **People:** free Teams for up to 10 people, where each person's Pally coordinates with the others and sharing is chosen per capability. Since 7 October, people in a group chat get their own Pally the first time they ask it something. - **Approvals:** Pally shows you anything it proposes word for word; things you explicitly ask for can go without a second confirmation. - **Training:** since 6 October 2026 its [privacy policy](https://pally.com/privacy) says Pally may train its own models on your data unless you opt out. Pally is at its best as one person's all-round helper, and at no cost it is hard to argue with. ## What Arlo is Arlo is an AI colleague you text on [iMessage](/imessage), SMS or [WhatsApp](/whatsapp), use inside [Slack](/slack), or hand a [phone call](/phone) to. It places calls to people and businesses, works through phone menus and waits on hold. It connects to Gmail, your calendar, Slack, Notion, Linear, HubSpot and 3,000+ other tools, and reaches sites with no API through a secure browser session you sign into once. Anything it sends or changes waits for your approval unless you've pre-approved that kind of action, and every tool call, source and approval is written to an audit trail you can read back. It [texts you first](/proactive) with briefs and reminders over iMessage and SMS. Anyone can start today, with no invite or waitlist. It's free to start, with no card, then runs on plans from $20 a month. ## Feature comparison | Capability | Arlo | Pally | |---|---|---| | Built for | Work: founders, operators, teams | One person's life, errands and inbox | | Chat channels | iMessage, SMS, WhatsApp, Slack | iMessage, Android RCS and SMS, Telegram | | WhatsApp | Chat with it on WhatsApp | Reads your WhatsApp and sends replies; you chat with it elsewhere | | Working with a team | Works inside your team's Slack | Free Teams that link up to 10 people's own Pallys; group chats | | Phone calls | Places calls to people and businesses, through phone menus and hold | Included for everyone, drawn from a weekly usage limit; you can call your Pally, which gets its own number on any account | | Tools it can act in | 3,000+, plus no-API sites via a browser session | About two dozen named apps, MCP servers on request, browser sign-in to most sites | | Approval before sending | By default, unless you've pre-approved that kind of action | Shows you anything it proposes word for word; things you explicitly ask for can go without a second confirmation | | Audit trail of every action | Yes: every tool call, source and approval | Not a stated feature | | Flight booking and card checkout | Not a published feature | Yes: flights, hotels, rental cars; capped single-use card (US) or saved card | | Price | Free to start, no card, then plans from $20 a month | Free for everyone with every feature and one weekly usage limit; invite-only VIP at $1,000 a month | The Pally column reflects its site, terms, privacy policy and changelog as of October 2026. "Not a stated feature" means Pally doesn't publish one. ## Where Pally is the better pick - **Price.** This is Pally's strongest card. Calls, its own number and email, unlimited automations and Teams all cost nothing within a weekly usage limit. If you want an assistant you text and don't want to pay for one, start with Pally. Arlo can't match that on price. - **Personal errands and travel.** Pally books flights end to end, checks you in, claims fare drops, books hotels and rental cars, finds restaurant tables, and checks out on a capped single-use card or your saved card. - **Telegram, Android and group chats.** Pally works on Telegram and on Android over RCS and SMS, and joins group chats on iMessage and Telegram. Arlo has no Telegram. - **Your Mac.** An optional Mac companion app reads your own iMessage history and LinkedIn and can control the Mac. Cloud Mac is a hosted version you pay for. - **Pace.** Pally's public changelog logs between 20 and 70 changes on most days. ## Where Arlo is the better pick - **You want to chat with it on WhatsApp.** You can text Arlo on WhatsApp. Pally reads your WhatsApp inbox, and its own guide sends people who want to chat there to Poke. - **Your team works in Slack.** Arlo works inside your team's Slack and connects to the shared tools a company runs on, such as HubSpot and Linear. In Pally, Slack is a source of data, and Teams connects each person's own Pally. - **You need a record.** Every tool call, source and approval Arlo makes is written to an audit trail you can read back. Pally keeps a team transcript and a record of messages as delivered, and publishes no per-action audit trail. - **You want approval by default.** Anything Arlo sends or changes waits for your approval unless you've pre-approved that kind of action. Pally has been removing second confirmations since late September: messages you tell it to send, offers you accept, and calls, invites and forwards you ask for now go without one, and it can send on its own within limits you set ([changelog](https://pally.com/changelog.md)). Plenty of people will reasonably use one of each. If you're weighing more options, [Pally alternatives](/compare/pally-alternatives) covers the field, and the [best AI assistants you can text](/compare/best-ai-assistants-you-can-text) ranks all of them. ## Privacy and training Pally changed its privacy policy on 6 October 2026. The [late-August version](https://web.archive.org/web/20260830085317/https://pally.com/privacy) said its AI providers were configured so your data is not used to train AI models. The [current policy](https://pally.com/privacy) lists "to train and evaluate our AI models, unless you opt out", with the opt-out under Data retention in your settings at my.pally.com. Pally's own comparison pages now say this openly. Arlo's [privacy policy](/privacy) says it doesn't use your messages, files, memories, calls or connected-account content to train AI models. ## Frequently asked questions **What is the difference between Arlo and Pally?** Both are AI assistants you text that take real actions, place phone calls and show you what they propose. Pally is a personal assistant on iMessage, Android RCS and SMS, and Telegram, free for everyone as of October 2026 and strongest on errands, travel and checkout. Arlo is built for work: you can chat with it on WhatsApp and inside your team's Slack, it holds what it sends or changes for your approval unless you've pre-approved that kind of action, and it keeps an audit trail. **What is Pally AI?** Pally (pally.com) is an AI personal assistant you text, made by Pally Technologies, a Y Combinator company. You chat with it over iMessage, Android RCS and SMS, or Telegram, and it handles email, calendar, errands, travel bookings and phone calls. It launched as a texting assistant in June 2026. **Does Pally have a free plan?** Pally is now free for everyone. Since 6 October 2026 every feature, including calls and its own number, is free within one usage limit that refills through the week. Pro and Max were discontinued and refunded pro rata, and the only paid plan is VIP, offered by invitation at $1,000 a month or $10,000 a year. Many reviews still list the old Free, $25 Pro and $100 Max prices, which no longer apply. **How does Arlo's cost compare with Pally's?** On cost alone, Pally wins. Pally is free for nearly everyone within a weekly usage limit, and its only paid plan, VIP, is by invitation. Arlo is free to start, with no card, then runs on plans from $20 a month; [pricing](/pricing) has the details. The question is whether you need what Arlo adds: WhatsApp and Slack as places to work with it, approval by default, and an audit trail. **What do reviews say about Pally AI?** Public reviews of the texting assistant are thin as of October 2026. Its Product Hunt rating of 4.9 from 16 reviews mostly covers the 2025 relationship-management product Pally started as, and there's no App Store app to rate. The most detailed independent test is [Assistant Benchmark](https://assistantbenchmark.com/agents/pally) (tests run 12–17 September 2026), which ranked Pally fifth overall at 8.3, with full marks for routines and restraint and a median reply of 22 seconds. **Is Pally AI discussed on Reddit?** Barely. As of October 2026 we found no Reddit threads about Pally's texting assistant, and Hacker News has only a passing mention. Most public feedback is on X and in Assistant Benchmark's hands-on tests. **Is Pally safe to use?** We found no public outages or security incidents as of October 2026. The thing to check is training: since 6 October its privacy policy lets Pally train its own models on your data unless you opt out under Data retention in your settings. **Does Pally work on WhatsApp?** Pally can read your WhatsApp inbox and send replies there, but you chat with it over iMessage, Android RCS and SMS, or Telegram, as of October 2026. Its own WhatsApp guide sends people who want to chat inside WhatsApp to Poke. You can text Arlo on WhatsApp directly. **Which one can make phone calls?** Both. Pally's calls are included for everyone and draw on its weekly usage limit, with no per-minute charge, and you can call your Pally. Arlo places calls to people and businesses, works through phone menus and waits on hold. **Which is better for a team?** It depends on what the team needs. Pally coordinates between people's personal assistants: free Teams for up to 10 people with sharing chosen per capability, plus group chats. Arlo works inside the team's shared Slack and tools, with an audit trail of every action and approval. --- # Arlo vs Poke — which AI assistant is right for you? > Poke is a personal AI assistant that lives in your messaging apps (iMessage, WhatsApp, Telegram) for individual productivity. Arlo is an AI colleague for work that runs across iMessage, Slack, WhatsApp, and phone calls, with governed tool access, approvals, and a full audit trail. Poke fits one person; Arlo fits a team. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/arlo-vs-poke You talk to both Poke and Arlo like a person, but they're built for different jobs. Poke is a personal assistant that lives in your messaging apps and helps you run your own digital life. It was built by The Interaction Company, which [Cognition](https://cognition.ai) — the AI lab behind the Devin coding agent — acquired on 23 July 2026 in a deal reported in the low nine figures. Arlo is an [AI colleague](/glossary/ai-colleague) for work. It runs in the channels your team already uses, takes real actions through tool access that's governed and approved, and logs everything it does. Want a sharp, proactive assistant just for you? Poke is a strong consumer pick. Want something your whole team can lean on inside its existing workflows, with reviewer approvals and admin controls? That's the gap Arlo fills. ## What Poke is (as of August 2026) Poke positions itself as a personal assistant you reach over the messaging apps you already have open. From its public materials: - It works over Apple Messages (iMessage), WhatsApp, and Telegram, in both text and voice messages. - It's aimed at individual consumers — the tagline is that it "fits into your life, not the other way around." - It connects to apps like Gmail, calendars, Notion, GitHub, and Asana, and can act on its own. - It ships "Poke Recipes" and a developer "Kitchen" for custom automations. - Public pricing lists Free, Pro, and Ultra tiers, all built around one person's use. What Poke does well is make AI feel as easy as texting a friend, and it found a real audience doing it — Cognition put the figure at more than 100 million messages exchanged in the three months before the acquisition. **What the Cognition acquisition changes.** Poke continues to run, and Cognition has said Poke gains its models and infrastructure. The stated strategic reason for the deal, though, was to bring Poke's conversational style into Devin, Cognition's coding agent — so Poke's personality is now an ingredient in someone else's roadmap as well as a product of its own. If you're buying for a team, that's the thing to ask about: not whether Poke works today, but where a consumer texting assistant sits inside a company whose flagship is a developer tool. Product details shift, so check Poke's own site before you decide. ## What Arlo is Arlo is one AI colleague for the workspace. There's no separate chat app to live in. It works inside the tools a team already uses: iMessage and SMS, Slack, WhatsApp, and live phone calls. It remembers context across conversations and sends a morning briefing of what changed overnight. It connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools. For software with no API, it logs in once through a secure browser session and clicks, reads, and fills out forms the way a person would. Governance is the part that sets it apart, and it's on by default. Every connection clears policy before each run. Drafts, sends, and writes wait on a reviewer you pick. Each tool call, source, and approval lands in a full audit trace you can go back and read. ## Feature comparison | Capability | Poke (personal) | Arlo (work / teams) | |---|---|---| | Primary audience | Individual consumers | Teams and businesses | | Messaging channels | iMessage, WhatsApp, Telegram | iMessage and SMS, WhatsApp, Slack | | Live phone calls | Not a stated focus | Yes — it places calls for you, and you can call it | | Persistent memory | Yes | Yes, plus a morning briefing of changes | | Integrations | Connects to popular apps | 3,000+ tools | | No-API tool access | Not a stated focus | Yes — secure browser sessions | | Reviewer approvals | Not a stated focus | Yes — drafts/sends/writes await a reviewer | | Audit trail | Not a stated focus | Full audit trace of calls, sources, approvals | | Admin policy / roles / SSO | Not a stated focus | Yes, on request | This table reflects each product's stated positioning as of August 2026. "Not a stated focus" for Poke means its public materials emphasize personal use, not that the capability is impossible. ## Which one is right for you Pick **Poke** for yourself. One lightweight thread that handles your email, calendar, and errands, with no team controls to babysit. Pick **Arlo** when the assistant acts for a team and the work carries stakes. Its reach into Slack, its ability to make phone calls, and its governance model are all built for shared, accountable work. If you're choosing for work specifically, our guides on an [AI assistant for founders](/use-cases/ai-assistant-for-founders) and an [AI assistant for sales](/use-cases/ai-assistant-for-sales) go deeper. On pricing, Arlo starts free — $100 in credits for connecting Slack — then runs on prepaid credits: 400 credits per $1 at every amount, on monthly plans from $20 with top-ups from $50, with credits that last 12 months from purchase; admin policy, roles, SSO, and audit export are available on request. ## Governance is the real difference A personal assistant optimizes for one user's convenience. An AI colleague has to hold up in front of teammates, customers, and auditors. That changes what the product has to do. Arlo treats every action as something that might need review: connections pass policy before each run, sensitive steps wait on an approver you name, and the audit trace records what happened and why. That's why you can point Arlo at real workflows and not just personal to-dos. We wrote up what the Cognition deal signals for buyers in [Cognition bought Poke — what it means](/blog/cognition-poke-what-it-means). Still weighing options? Read our broader [Poke alternatives](/compare/poke-alternatives) overview and our explainer on the [AI executive assistant](/glossary/ai-executive-assistant) category. ## Frequently asked questions **What is the difference between Arlo and Poke?** Poke — built by The Interaction Company, acquired by Cognition in July 2026 — is a personal AI assistant that lives in your messaging apps and helps you run your own digital life. Arlo is an AI colleague for work: it runs in iMessage and SMS, Slack, WhatsApp, and on live phone calls, takes governed action in connected tools, and logs every run to an audit trace. Poke is built for one person; Arlo is built for a team. **Can Arlo do everything Poke does?** For personal messaging tasks, largely yes — you text it like a person and it handles email, calendar, and errands. The reverse is not true: Poke does not offer Slack as a first-class surface, does not place live phone calls, and does not ship approval gates or an audit trail. **Is Arlo or Poke better for a team?** Arlo. The difference is governance rather than features: Arlo resolves each connection through policy before a run, holds consequential sends and writes for a reviewer, and records what ran and who approved it. A personal assistant optimises for one user's convenience, which is the wrong default once an assistant acts on behalf of a company. **Can Arlo make phone calls like a person?** Yes — Arlo places real calls, and keeps the transcript, so you can ask how it went. Poke does not offer live calling. **How much does Arlo cost compared with Poke?** They are priced differently in kind. Arlo starts free — connecting Slack grants your workspace $100 in credits (40,000) — then you buy credits at 400 per $1 on monthly plans from $20, with top-ups up to $50,000, and they expire 12 months after purchase. At that rate $50 is roughly 625 messages, or 69 researched answers with sources. Admin policy, roles, SSO, and audit export are available on request. Poke is light-free with heavier use paid — check Poke's own site for its current pricing. **Can I use both?** Nothing stops you, but they overlap heavily on personal tasks. Most people pick one for their own life and add Arlo when the work involves teammates, customers, or anything that needs a reviewer. ## Try Arlo If you need an AI colleague that works inside your team's tools with governed actions and a full audit trail, [try Arlo](/signup). --- # Arlo vs Viktor — which AI coworker fits your team? > Viktor and Arlo are both AI coworkers that do real work from team chat. Viktor is an AI employee in Slack and Teams known for heavy delegation and running code. Arlo works in Slack, iMessage/SMS, WhatsApp, and live phone calls, with approval gates and a full audit trail on by default. Pick Viktor for in-chat technical delegation; pick Arlo when work spills into texts, calls, and governed tools. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/compare/arlo-vs-viktor Viktor and Arlo are the two products most likely to be on the same shortlist: both are AI coworkers you talk to in your team's chat, both connect to thousands of tools, and both promise finished work rather than just answers. The overlap is real, so this comparison focuses on where they genuinely differ — channels, governance defaults, and the kind of work each is best at. Viktor, made by a fast-moving team that raised a reported $75M round in May 2026, positions itself as an "AI employee" for Slack and Microsoft Teams. Arlo is an [AI colleague](/glossary/ai-colleague) for the workspace that works across iMessage and SMS, Slack, WhatsApp, and live phone calls. If your team's work stays inside Slack, both deserve a look. If it spills into text threads and phone calls — or if approval-first governance is a requirement — the comparison tilts. ## What Viktor is (as of July 2026) As of July 2026, from Viktor's public materials: - It's an autonomous AI employee that lives in Slack or Microsoft Teams — teammates @mention it in channels and it carries tasks to completion. - It connects to 3,200+ tools and handles analytics, automation, reports, and campaigns. - It writes and executes code, and can build small web apps and artifacts on request. - It's SOC 2 Type 1 certified, with controls over which tools require approval before executing. - It publishes an unusually deep content library and case studies — the team ships fast and says so publicly. Viktor is a polished product with serious momentum, and for teams that want to delegate technical, in-chat work at volume, it's a genuinely strong option. Details shift quickly — check Viktor's own site for the current state. ## What Arlo is Arlo is one AI colleague for the whole workspace. It works in the channels your team already uses — including the two most companies ignore: **text messages and live phone calls**. You can text Arlo from iMessage or SMS like a person, and it can place real calls — confirm an appointment, chase a vendor, ask a business a question — and keep the transcript so you can ask how it went. It connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools, and for software with no API it logs in once through a secure browser session and works the interface like a person would. It keeps persistent memory across conversations and channels, delivers one morning briefing of what changed overnight, and suggests automations for work it sees repeating. Governance is the center of the product, not a setting to find later. Every connection resolves through policy before each run, sends and writes wait on a reviewer you pick, and every run leaves a full [audit trail](/glossary/ai-agent-audit-trail) — each tool call, source, and approval. ## Feature comparison | Capability | Viktor | Arlo | |---|---|---| | Category | AI employee for team chat | AI colleague for the workspace | | Slack | ✓ | ✓ | | iMessage / SMS | — | ✓ | | Live phone calls (in + out) | — | ✓ | | Tool integrations | 3,200+ | 3,000+ | | Tools with no API | Not a stated focus | ✓ Secure browser session | | Writes and executes code | ✓ | Not a stated focus | | Persistent memory + morning brief | Learns from delegation | ✓ Memory across channels, daily brief | | Approval gates | Configurable per tool | ✓ On by default, reviewer you pick | | Audit trail per run | Activity visibility | ✓ Full trace: calls, sources, approvals | | Compliance posture | SOC 2 Type 1 | Governance-first design, audit export on request | Both products evolve weekly; treat this as a snapshot from July 2026 and verify anything load-bearing. ## Where Viktor wins Concede the real things: Viktor is excellent at high-volume, in-chat delegation — especially work that benefits from writing and running code, like data pulls, report generation, and quick internal tools. Its integration count is enormous, its case-study library is public, and SOC 2 Type 1 is a checkbox some procurement teams will want on day one. If your team lives entirely in Slack and wants an AI employee to bury in tickets, Viktor is a credible hire. ## Where Arlo wins Arlo wins on reach and on defaults. Work doesn't stay in Slack: customers text, vendors only answer the phone, and half your own team responds faster in iMessage than in any channel. Arlo is the only one of the two that [works across texts and live calls](/use-cases/ai-assistant-you-can-text-and-call) as first-class surfaces. And when an AI touches production systems and customer inboxes, Arlo's model — approval before the send, audit trail after it, policy checks in between — is on from the first minute, not assembled from settings. ## How to choose Ask two questions. First: **where does your work actually happen?** If the answer is "Slack, full stop," both fit; if it includes texts and calls, that's Arlo. Second: **what's your tolerance for un-reviewed AI action?** If you want every outbound send gated by a human until trust is earned, Arlo's approval-first defaults are the shorter path. Then add a third if you're comparing cost: **how long do the credits you buy stay yours?** Arlo's last 12 months from purchase, which is unusual in a category where credits commonly reset each month — the cheapest way to decide the rest is to give each the same three tasks from your actual backlog. Looking past this one matchup? See [Viktor alternatives](/compare/viktor-alternatives) for the wider field, and the [AI coworker](/glossary/ai-coworker) explainer for the category both products are selling into. ## Frequently asked questions **Are Viktor and Arlo direct competitors?** Partly. Both are AI coworkers that do real work from team chat. Viktor is Slack/Teams-only and leans into code execution; Arlo covers Slack, iMessage, SMS, WhatsApp, and live phone calls, and centers approval-first governance. **Does Viktor work over iMessage or SMS?** No — as of July 2026, Viktor's stated channels are Slack and Microsoft Teams. Arlo works in iMessage and SMS as first-class channels alongside Slack and WhatsApp. **Can either make phone calls?** Arlo places live phone calls and keeps a transcript of each. Phone calling isn't part of Viktor's stated product as of July 2026. **Which is better for a non-technical team?** Teams that mostly need coordination — scheduling, follow-ups, inbox triage, CRM updates, calls — tend to fit Arlo's shape. Teams that want an AI to write and run code from chat requests tend to fit Viktor's. **Can I try both?** Yes, and you should. Arlo is free to start — texting it or connecting Slack comes with credits, no card — and then runs on monthly plans from $20 at 400 credits per $1; $50 is roughly 625 messages or 18 built apps, and whatever you don't spend stays good for 12 months rather than lapsing at the end of the month. Run the same real tasks through each and compare the finished work, not the demos. --- # Best AI assistants for Slack in 2026: 8 tools, 3 jobs > The best AI assistant for Slack depends on which job you're hiring for. To delegate real work across your tools — email, CRM, issues, even phone calls — Arlo is the pick, with an approval before every send and a full audit trail. To understand what happened in Slack, use Slack's own AI. To deflect support tickets, use ClearFeed, Fin, or Zendesk. Most shortlists mix the three. Updated: 2026-08-06 · Canonical: https://www.arlo.sh/compare/best-ai-assistants-for-slack Search "AI assistant for Slack" and you get a list that mixes a thread summarizer, a customer-support bot, a Jira search box, and an autonomous agent that can spend your money — as if they were competing for the same slot. They aren't. They do three different jobs, and picking from the wrong one is how teams lose a quarter to a pilot that was never going to work. So this list is sorted by job, not by vendor. Figure out which job you're hiring for first; the shortlist inside each one is short and the choice gets easy. Our own product, [Arlo](/slack), sits in the third job — we'll say plainly where it isn't the right pick. **The three jobs:** 1. **Understand** — summarize threads, recap channels, search history. You want to know what happened. 2. **Deflect** — answer repetitive questions from a knowledge base so a human never opens the ticket. You want volume off your team's plate. 3. **Delegate** — hand over a task that touches real tools and get finished work back. You want the thing done. ## The short list | Tool | Job | Lives where | Takes action in your tools | |---|---|---|---| | **Slack AI** | Understand | Slack only | — Reads and summarizes Slack | | **Atlassian Rovo** | Understand | Slack + Atlassian | — Search and answers | | **ClearFeed** | Deflect | Slack only | ✓ Tickets, triage, routing | | **Intercom Fin** | Deflect | Slack + web + email | ✓ Resolves support conversations | | **Zendesk AI** | Deflect | Slack + Zendesk | ✓ Support workflows | | **Arlo** | Delegate | Slack, iMessage/SMS, WhatsApp, phone calls | ✓ 3,000+ tools, approvals + audit trail | | **Viktor** | Delegate | Slack, Teams | ✓ 3,200+ tools, analytics and code | | **Lindy** | Delegate | Slack, email, iMessage, browser | ✓ Builder-defined workflows | Pricing structures differ more than prices do: Slack AI is bundled into paid Slack plans, support tools charge per agent seat or per resolved conversation, and delegation tools charge per user or per usage. Structures below reflect each vendor's public materials as of August 2026 — check their pricing pages for current numbers. ## Job 1 — Understand what happened ### Slack AI The native option, and the honest default. Slack AI summarizes long threads, gives you channel recaps, and answers questions against your own message history. It's bundled into paid Slack plans rather than sold separately, which makes it the cheapest thing on this list for most teams — you may already have it. Its boundary is also clean: Slack AI knows about Slack. It won't open your Linear ticket or chase a vendor by email. If your problem is "I was in twelve channels today and I don't know what I missed," stop here. You don't need anything else on this page. ### Atlassian Rovo Rovo works inside Slack but its center of gravity is Atlassian: it searches across Jira, Confluence, and Bitbucket and answers with that context, and it's included in paid Atlassian Cloud plans. If your team's institutional knowledge lives in Confluence and half your Slack questions are really "where's that page," Rovo is the tighter fit than a general assistant. The same boundary applies — it's a very good search-and-answer layer, not a delegate. ## Job 2 — Deflect the repetitive questions This is the most mature and most competitive part of the category, and it's the one where a general-purpose AI assistant is usually the *wrong* buy. If your job is "our support channel gets the same forty questions a week," you want a purpose-built deflection product with ticketing, SLAs, routing, and reporting — not a colleague. ### ClearFeed Slack-native support: it turns channel messages into tracked tickets, triages them, and answers from your knowledge base autonomously. Priced per agent or per channel. If your customers or your internal teams already live in shared Slack channels, ClearFeed is built exactly for that shape and it's hard to beat there. ### Intercom Fin Fin resolves customer conversations across Slack, web chat, and email, and is notable for charging per *resolved* conversation rather than per seat — you pay when it actually deflects. That pricing model makes ROI unusually easy to measure, which is why it shows up on most support shortlists. ### Zendesk AI If you're already a Zendesk shop, its AI agents surface knowledge articles and its copilot drafts replies, with Slack as one surface among many. The integration story is the reason to pick it: staying inside the helpdesk you already run usually beats a marginally smarter bot that lives somewhere else. ## Job 3 — Delegate work that leaves Slack Here the assistant isn't answering — it's going away, doing something in your actual tools, and coming back with the result. This is the newest of the three jobs and the one with the most variation in how much you're trusted to control. ### Arlo — a colleague, not a Slack feature Arlo is an [AI colleague](/glossary/ai-colleague) that lives in Slack like a teammate: @mention it in a channel or DM it, and it acts through Gmail, Notion, Linear, GitHub, your CRM and 3,000+ other tools. Two things distinguish it in this job. The first is **governance by default**. Anything that sends or writes pauses for a reviewer you pick, and every run leaves an [audit trail](/glossary/ai-agent-audit-trail). Standing permission is possible but scoped — you grant a specific action, and the grant expires; when it does, the action fails closed rather than quietly proceeding. That matters more than it sounds once an assistant has write access to your CRM. The second is that **Slack isn't the boundary**. The same colleague, with the same memory, is also in [iMessage and SMS](/imessage), on [WhatsApp](/whatsapp), and on [live phone calls](/compare/ai-assistants-that-can-make-phone-calls). A thread you start in Slack still makes sense when you follow up by text from a cab. Arlo's pricing is one rate: credits at 400 per $1, on monthly plans from $20, spent across every channel and good for 12 months. **When Arlo is the wrong pick:** if you want ticket deflection with SLAs and support reporting, buy a support product from Job 2 — Arlo isn't one. If your team is strictly Slack-only and never leaves it, you're paying for channels you won't use. And if you want a visual builder to design deterministic workflows, see Lindy below. ### Viktor — heavy delegation, Slack and Teams Viktor positions itself as an "AI employee" for Slack and Microsoft Teams: teammates @mention it and it carries tasks to completion across 3,200+ tools, with a reputation for analytics, reports, campaigns, and running code. It's the closest product to Arlo on this list, and if your work genuinely never leaves team chat it's a strong pick. The differences worth checking against your own requirements are channels — Viktor doesn't do texts or phone calls — and governance defaults, which is a question to put to both vendors rather than take from a comparison page. Full breakdown: [Arlo vs Viktor](/compare/arlo-vs-viktor). ### Lindy — automation you can reach from Slack Lindy comes at this from the workflow side: you build agents and multi-step processes in a visual builder, then reach them from Slack, email, iMessage, or the browser. If you want repeatable, deterministic flows you've designed yourself, that's a real advantage. If you want to delegate messy one-off work in plain language, the builder is overhead. Full breakdown: [Arlo vs Lindy](/compare/arlo-vs-lindy). ## Why "AI assistant for Slack" is a harder search than it looks Two structural things make this category confusing to shop for. **Slack competes with its own ecosystem.** Slack ships native AI, Slackbot as a personal agent, and Agentforce for Salesforce-centric automation. For the Understand job the native option is genuinely good and usually already paid for, which quietly raises the bar for everything else: a third-party tool has to be better at a *different* job, not slightly better at summarizing. **"Assistant" is doing too much work as a word.** A support-deflection agent and a delegation colleague are as different as a call center and a chief of staff, but both get marketed as "AI assistant for Slack." The three-job split above is the fastest way to cut through it — and the fastest way to disqualify two thirds of any shortlist you're handed. ## How to choose Ask what you want to be true in ninety days. - *"I know what happened without reading everything."* → Slack AI, or Rovo if your knowledge lives in Atlassian. - *"Our team stopped answering the same question."* → ClearFeed for Slack-native support, Fin if you want to pay per resolution, Zendesk AI if you're already there. - *"Work I delegate comes back done."* → [Arlo](/use-cases/ai-assistant-in-slack) if that work touches many tools and needs approvals, or spills into texts and calls; Viktor if it stays in chat; Lindy if you'd rather design the workflow yourself. If two of those are true, buy two tools. They're not substitutes, and the combined bill is usually smaller than one failed pilot. ## Frequently asked questions **What is the best AI assistant for Slack?** There isn't one best — there are three jobs. Slack's own AI is best for summarizing and searching Slack, ClearFeed and Intercom Fin are best for deflecting support questions, and Arlo or Viktor are best for delegating work that touches your other tools. **Does Slack have its own built-in AI assistant?** Yes. Slack AI provides thread summaries, channel recaps, and AI search over your message history, and it's bundled into paid Slack plans rather than sold separately. It only knows about Slack, so it can't take action in your other tools. **Can an AI assistant in Slack actually take actions in other tools?** Yes, but only the delegation-class tools do. Arlo acts across 3,000+ tools including Gmail, Notion, Linear, GitHub, and CRMs, and holds anything that sends or writes for a human reviewer. Summarizers and search tools read and answer, they don't act. **Is it safe to give an AI assistant write access to Slack and connected tools?** Look for three specific controls: approval before any send or write, permissions scoped to named actions rather than blanket access, and an audit trail of every run. Arlo ships all three on by default, and standing grants expire and fail closed rather than continuing silently. **What's the difference between a Slack AI assistant and a Slack chatbot?** A chatbot answers inside the thread. An assistant leaves the thread — it reads your connected tools, performs the task, and reports back with the result. If it can't touch anything outside Slack, it's a chatbot, however good its answers are. --- # Best AI assistants you can text in 2026 (iMessage, SMS, WhatsApp) > The best AI assistants you can text in 2026 are Arlo (for work — iMessage, SMS, WhatsApp, Slack and calls, with approval before anything goes out), Pally (iMessage, RCS and Telegram, now free), Poke (personal, owned by Cognition), Muse (Meta's agent on WhatsApp, US and Canada), Martin (SMS, WhatsApp and calls) and Catch (for executives). Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/best-ai-assistants-you-can-text The best interface for an AI assistant turned out to be the one you've used every day for twenty years: a text thread. You don't install anything or remember to open a tab. You text it like a person, it answers like a person, and if it's any good, it goes off and does the thing. "Any good" is where the field splits. Some of these are a chatbot behind a phone number, fun to talk to and not much else. The useful ones connect to your email and calendar and come back with finished work. A lot changed this summer, too. Cognition bought Poke's maker, Lindy relaunched as a Slack teammate, Catch launched, Instinct went viral with a waitlist, and OpenAI shipped a way for ChatGPT to send iMessages from a Mac. Then the autumn moved faster: Meta launched its Muse agent, Instinct raised $1 billion, Pally made every feature free, and Hark Pro launched as an app. This is where things stand in October 2026. We make one of these products (Arlo), so read our section with that in mind. We also asked ChatGPT's model which assistants it names; [here's what it said](/blog/which-ai-assistants-chatgpt-recommends). We compared four things: which apps you can actually text it from, whether it does anything in your tools or only chats, whether it can handle a phone call, and whether it's built for your personal life or your work. Every claim about another product comes from that product's own site and dated reporting. Pally, Instinct, Muse and Hark Pro were checked as of October 2026; the rest as of September 2026. We mostly describe pricing by structure instead of dollar amounts, because prices change faster than this page will. ## The short list | Assistant | Channels you can text it on | Built for | Real actions in your tools | Phone calls | |---|---|---|---|---| | **Arlo** | iMessage, SMS, WhatsApp, plus Slack | Work — founders, operators, teams | ✓ 3,000+ tools, approval before sends, audit trail | ✓ Places calls | | **Pally** | iMessage, Android RCS and SMS, Telegram (reads your WhatsApp) | Personal life and work | ✓ Email, calendar, Slack, Notion, Linear, GitHub | ✓ Included, within a weekly usage limit | | **Poke** | iMessage, SMS, Telegram, WhatsApp in some markets | Personal life | ✓ 30+ integrations, "Recipes" | — | | **Muse** (Meta) | WhatsApp, plus its own app — US and Canada only | Personal life and small business | ✓ Email, calendar, shopping, connectors | Beta: calls to US businesses | | **Martin** | SMS, WhatsApp, email, Slack, iOS app | Busy individuals | ✓ Calendar, inbox, reminders | ✓ Calls for you | | **Catch** | iMessage, WhatsApp, Slack, email | Leaders at mid-sized companies | ✓ Scheduling, inbox, travel, forms | ✓ Outbound calls | | **alfred_** | iMessage, SMS, RCS | People who bill for their time | ✓ Email, calendar, tasks — approves every send | — | | **Lindy** | Slack first, plus iMessage and email | Teams, sold per seat | ✓ 1,000+ tools, approval on outside actions | Joins meetings | | **Instinct** (invite only) | iMessage, SMS, WhatsApp; group chats in early access | Personal life | ✓ A cloud computer using your logins | ✓ Calls you; calls businesses in early access | | **Sidekicks** | iMessage and SMS, group chats | Consumers — personas | Light — reminders and to-dos | — | | **Tomo** | iMessage, plus a companion app | Personal goals and accountability | ✓ Calendar, email, web | — | | **ChatGPT** | Not textable — its Messages plugin runs on a Mac | Existing ChatGPT users | ✓ Reads and sends your iMessages from a Mac | — | | **ChatGPT dots** | Not textable for most people — inside ChatGPT; texting is a limited US beta | ChatGPT Pro and Business users | ✓ Its own cloud computer | — | | **Hark Pro** | Not textable — an iOS, Android and web app | Personal life | ✓ Errands through its own browser agent | — | ## Arlo: for when the texts are about work Arlo is an [AI colleague](/glossary/ai-colleague) you text from iMessage, SMS or WhatsApp the way you'd text a capable teammate. "Chase the Hendricks invoice." "Move my 2pm to Thursday." "What changed overnight?" It looks in the tools behind the request (Gmail, Slack, Notion, Linear, your CRM, and about 3,000 others, plus sites with no API through a browser session you sign into once), does the work, and holds anything that sends or writes until you say yes. A typical exchange is short. You text "the Hendricks invoice is two weeks late, can you nudge them?" Arlo finds the invoice in your inbox, drafts a polite follow-up to the right person, shows it to you, and sends it when you reply "send". It also [makes live phone calls](/compare/ai-assistants-that-can-make-phone-calls), and it [texts you first](/proactive) with the briefs and reminders you ask for. One memory carries across your texts, Slack and calls, so context doesn't reset when you switch channels, and every run lands in an audit trail. It's free to start with no card, then runs on monthly plans from $20. [Pricing](/pricing) has the details. The caveat: Arlo is built for work. If what you want is help with errands, family logistics or a nudge to drink water, Pally, Poke and Martin are built for that. ## Pally: a personal assistant with the widest channel list Pally is the closest product to Arlo on this list and the one most worth comparing directly. It works over iMessage, Android RCS and SMS through Google Messages, and Telegram, can read your WhatsApp inbox and send replies there, connects to Gmail, Outlook, Google Calendar, Notion, Linear, Slack, GitHub and Jira, books flights, hotels and restaurants, and checks out with your approval. Since 6 October 2026 every feature is free for everyone, with one usage limit that refills through the week; calls, its own phone number and its own email address are included, and the old paid tiers are gone apart from an invitation-only VIP plan ([Pally's terms](https://pally.com/terms)). It pitches itself at both personal and professional use. The difference is emphasis: Pally is organised around one person's loose threads; Arlo is organised around work that other people depend on, with approval gates on sends and writes and an audit trail you can hand to someone else. Full breakdown: [Arlo vs Pally](/compare/arlo-vs-pally). ## Poke: the personal pick, now part of Cognition Poke lives in iMessage, SMS and Telegram, with WhatsApp in some markets, connects to your email, calendar and 30-plus personal apps, and runs life admin through one-tap "Recipes." Its texting experience is polished. On 23 July 2026 Cognition, the lab behind the Devin coding agent, acquired The Interaction Company, Poke's maker; Poke keeps running, and reporting at the time said nothing would change immediately. It's built around one person, with a free tier and paid plans. For personal use it's a strong pick; for team work it isn't trying to compete. More: [Arlo vs Poke](/compare/arlo-vs-poke) and [what the acquisition means](/blog/cognition-poke-what-it-means). ## Martin: the Jarvis pitch, over SMS and calls Martin markets itself as a personal AI like Jarvis: it texts over SMS and WhatsApp, now also works in Slack and email, manages your calendar and inbox, sets reminders, and makes calls for you. It's aimed at busy individuals, like professionals, students, and anyone who wants life logistics handled from a thread, and it's a subscription with a free trial. Of everything here it's the most direct personal analog to Arlo's texting surface, without the team, governance, or deep tool layer. Full breakdown: [Arlo vs Martin](/compare/arlo-vs-martin). ## Catch: the executive assistant that calls Catch, covered at launch in early September 2026, is an assistant for executives and leaders at mid-sized companies. Its homepage still shows a waitlist, so check availability. It runs over iMessage, WhatsApp, Slack and email, handles scheduling, inbox, travel booking and forms, places outbound calls, and is proactive by default. It sells one flat subscription with a trial. If you're an executive who wants a single assistant that reaches out first, it's a serious option. Full breakdown: [Arlo vs Catch](/compare/arlo-vs-catch). ## alfred_: for people who bill their time alfred_ is a texting assistant over iMessage, SMS and RCS for founders, lawyers, agents and accountants. It works across Gmail or Outlook, your calendar, tasks and notes, and asks you to approve every send. It offers an individual plan and a per-seat team plan. It doesn't list phone calls or WhatsApp. ## Lindy: now a Slack teammate you can also text Lindy relaunched in August 2026 as a teammate that lives in Slack first, and is also reachable over iMessage and email. It connects to over 1,000 tools, asks for approval before anything with outside impact, and joins Meet, Zoom and Teams calls. It's sold per seat on credit-based plans. If your team already lives in Slack and wants a per-seat agent, it's a real alternative; if you mostly want to text an assistant from your phone, it's Slack-shaped. Full comparison: [Arlo vs Lindy](/compare/arlo-vs-lindy). ## Instinct: invite-only for now Instinct went viral in late August 2026: an assistant you text over iMessage, SMS or WhatsApp, or call, that runs a computer in the cloud to carry errands through. On 28 September it raised $1 billion at a $10 billion valuation ([TechCrunch](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/)), and it now joins group chats in early access. As of October 2026 it's still invite-only, with five invites per member and a waitlist for everyone else, free while in early access, and licensed for personal use by its terms. If you're waiting on an invite, see [Instinct alternatives you can use today](/compare/instinct-alternatives), or how it stacks up against Meta's agent in [Instinct vs Muse](/compare/instinct-vs-muse). ## Muse: Meta's agent, textable on WhatsApp Muse is Meta's AI agent, launched on 8 September 2026 and separate from the Meta AI assistant built into its apps. You use it in its own app on iPhone, iPad, Android, the Mac and the web, or connect it to WhatsApp and text it there. As of October 2026 it's available only in the US and Canada, to adults. It runs errands on a cloud computer with a browser you can watch, connects to email, calendar and shopping sites, and is free with a weekly usage limit, with Power ($20 a month) and Max ($100 a month) plans for more ([Meta](https://www.meta.com/help/subscriptions/1021145227643680/)). By default it asks before some actions, not all of them. It passed five million downloads within a month, and it has also drawn the most privacy reporting of anything here. More: [Muse alternatives](/compare/muse-alternatives). ## Sidekicks and Tomo: the consumer iMessage field Two smaller products worth knowing. **Sidekicks** works over iMessage and SMS, including group chats, with switchable personas (assistant, health coach, companion) and a limited free allowance before its paid plan. It's the easiest pure-chat experience and light on real tool actions. **Tomo** is a proactive iMessage assistant with a companion app, pitched around personal goals and accountability; it connects to your calendar, email and the web, on a paid subscription. Both are credible for personal texting; neither is built for a team's tools, approvals, or phone lines. ## ChatGPT and the bridges: AI inside your Messages app Since 20 August 2026 the ChatGPT desktop app can read, search, draft and send iMessages from an Apple silicon Mac, asking before each send by default. That's useful, but it's a different thing from everything above. You can't text ChatGPT from your phone. It works on your existing threads, from your Mac, with Full Disk Access. We break down exactly what it does in [Can you use ChatGPT in iMessage?](/blog/chatgpt-in-imessage). OpenAI's always-on agent, Dots, launched inside ChatGPT on 29 September 2026 for Pro and Business Premium subscribers, and texting it is still a limited US beta; see [ChatGPT Dots alternatives](/compare/chatgpt-dots-alternatives). Separately, small bridge services still relay a phone number to a chat model: text a number, get model answers back. They're the quickest way to try AI by text, but most can't touch your calendar, inbox or tools, so nothing actually gets done. ## Hark Pro: app only Hark Pro, from Brett Adcock's AI lab Hark, launched on 6 October 2026 as a free iOS, Android and web app, with $20 and $100 monthly tiers for more usage. It runs errands through its own browser agent and suggests tasks on a home screen. You use it in its app. As of October 2026 there's no number to text it on, and it doesn't place phone calls. If you came here weighing it up, see [Hark Pro alternatives](/compare/hark-alternatives) and [Arlo vs Hark Pro](/compare/arlo-vs-hark). ## How to choose One question does most of the work: is the job personal or professional? For life admin, try Pally, Poke or Martin; if you're in the US or Canada and live in WhatsApp, Muse is worth a look. An executive who wants one proactive assistant should look at Catch. A Slack-first team buying per seat should look at Lindy. If the work touches your tools, your customers and your phone lines, and you want approvals and a record of what happened, that's Arlo. And if you only want help with the messages already on your Mac, ChatGPT's plugin does that. Details reflect each product's public site as of September 2026, and as of October 2026 for Pally, Instinct, Muse and Hark Pro; check them for the current state. ## Frequently asked questions **What is the best AI assistant you can text?** For work, Arlo. It handles real tasks across 3,000+ tools from an iMessage, SMS or WhatsApp thread, places phone calls, and holds sends for your approval with an audit trail. For personal life admin, Pally, Poke and Martin are the strongest picks. **Can I text an AI assistant without installing an app?** Yes. Texting-first assistants like Arlo, Pally and Poke give you a number you text like a person, from the Messages app or WhatsApp. There's no download and no new interface to learn. **Is there an AI assistant I can text on WhatsApp?** Yes. Arlo, Martin and Catch all work over WhatsApp, Instinct does for people with an invite, Meta's Muse does in the US and Canada, and Poke does in some markets; Pally reads your WhatsApp inbox but you chat with it elsewhere. ChatGPT is no longer available on WhatsApp in the US; 1-800-ChatGPT stopped working there on 15 January 2026, and is available only in the European Economic Area and Brazil ([what happened](/blog/chatgpt-on-whatsapp)). **Do any texting AI assistants also make phone calls?** Yes. Arlo, Pally, Martin and Catch all place calls; Instinct and Muse call businesses in early access or beta. Arlo places calls for work (confirming appointments, chasing vendors, answering questions) and keeps the transcript, so you can ask how it went. **Can I text ChatGPT in iMessage?** Not from your phone. OpenAI's Messages plugin runs in the ChatGPT desktop app on Apple silicon Macs and works on your existing conversations; it isn't a contact you can text. **Are AI texting assistants safe to connect to email and calendar?** Look for explicit governance: scoped access, approval before sends, and an audit trail. Arlo ships all three on by default; alfred_ and Pally also ask before sending, so check each product's security page for the rest. --- # Best AI executive assistants in 2026 — an honest roundup > The best AI executive assistant in 2026 depends on the job. Arlo is the team-scale pick — an AI colleague across iMessage, Slack, WhatsApp, and phone calls, with governance and a full audit trail. Lindy leads on inbox triage and workflow automation, Poke and April are strong personal texters, Martin is voice-first, and Catch takes proactive action. Updated: 2026-07-13 · Canonical: https://www.arlo.sh/compare/best-ai-executive-assistants "Best AI executive assistant" is the wrong question until you say who's asking. A founder who lives in iMessage wants something different from a sales team that needs approvals, and both want something different from a solo operator who just needs their inbox triaged. The tools below are genuinely good — they're just good at different jobs. This roundup lays out what each one is actually for as of 2026, so you can match the tool to the work instead of the marketing. Everything here reflects each product's public positioning as of 2026. Pricing and features move fast, so treat specifics as a starting point and check each vendor's site before you commit. Full disclosure: we make [Arlo](/), and it's on this list. We've tried to position it honestly next to the others rather than talk over them. ## The best AI executive assistants at a glance | Assistant | Where it works | Best for | Pricing (as of 2026) | | --- | --- | --- | --- | | Lindy | Email, Slack, calendars, business apps | Inbox triage + workflow automation | ~$29.99–$199.99 per user/mo, 7-day trial | | Poke | iMessage, SMS, Telegram | Personal tasks over text | Light free, heavier paid | | Martin | iOS, voice, text, email | Voice-first scheduling and reminders | See Martin's site | | Catch | Text + voice calls | Proactive action, flat pricing | ~$99/mo, calls included | | April | Text messages, no app | Simple personal EA over text | ~50 messages free, then ~$25/mo | | Arlo | iMessage/SMS, Slack, WhatsApp, live calls | An AI colleague for a governed team | Free to start · plans from $20/mo · 400 credits per $1 · credits last 12 months | ## Lindy Lindy, out of San Francisco, is the closest thing to a true executive-assistant-as-workflow-engine. As of 2026 it triages Gmail and Outlook, drafts replies in your voice, schedules, preps you for calls, records and summarizes meetings, and texts you the parts that matter — much of it operated over iMessage and SMS. Its "when this, do that" workflow builder is the real differentiator, and with a reported ~400K users it's one of the more proven options. If your problem is a flooded inbox and repetitive multi-step processes, Lindy is the pick. Pricing runs roughly $29.99–$199.99 per user/mo with a 7-day trial. ## Poke Poke, from The Interaction Company in Palo Alto (backed by General Catalyst, public since April 2026), is the friend-you-text version of an assistant. You sign up with a phone number, there's no app, and it lives in iMessage, SMS, and Telegram. It's tuned for personal life: daily planning, calendar, email filtering, flight check-ins, smart home, and shareable "Recipes." It's consumer-first and conversational, and it does that job well. If you want a sharp personal assistant for your own texts, Poke is a strong pick — see our [Arlo vs Poke](/compare/arlo-vs-poke) head-to-head. Pricing is light-free with heavier use paid. ## Martin Martin is the voice-first one. It's iOS-first and built around talking: manage email, calendar, and reminders by voice or text, with Gmail and Outlook connected. As of 2026 its lane is quick spoken interactions — reminders, scheduling, dashing off a message while you're walking. If the way you actually want to use an assistant is to speak to it, Martin was designed for exactly that. ## Catch Catch takes a different tack: it's built to proactively take action rather than wait to be asked, and it prices simply. As of 2026 it's roughly a flat $99/mo with voice calls included and no per-call fees, which makes budgeting easy if calling is central to your workflow. If you want an assistant that leans forward and you don't want to think about metered call costs, Catch is worth a look. ## April April is the minimalist personal EA. It lives entirely in text messages, with no app to install, and it's priced to try — your first ~50 messages are free, then it's about $25/mo for unlimited. As of 2026 it's a clean fit for someone who wants a lightweight personal assistant over SMS without a lot of setup or configuration. ## Arlo Here's our case, made plainly. The tools above each picked one lane — personal texter, inbox automator, voice, proactive action. [Arlo](/) is built for a different customer: a team that needs one colleague across every channel, with accountability baked in. - **It works everywhere, including the phone.** Arlo runs in iMessage and SMS, Slack, WhatsApp, and live [phone calls it can place](/use-cases/ai-executive-assistant-that-makes-phone-calls). Context doesn't drop when a thread becomes a call, and there's no new app to adopt. - **It acts, then proves it.** Arlo connects to 3,000+ tools, and for software with no API it logs in once through a secure browser session and clicks, reads, and fills forms like a person would. It keeps persistent memory and sends one morning briefing of what changed. - **Governance is the default.** Every send and tool write pauses for an approval gate, and every run leaves a [full audit trail](/glossary/ai-agent-audit-trail) — every tool call, source, and approval, on the record. That's the line between a personal bot and something a team can let act on its behalf. - **It scales past one person.** It runs on credits at 400 per $1, on monthly plans from $20 up to $50,000, with every capability available from the first dollar. Admin policy, roles, SSO, and audit export are available on request. Where Arlo is *not* the answer: if you only need a personal assistant for your own texts and errands, Poke, April, or Martin will feel lighter and cheaper. Arlo earns its keep when the work is shared and carries stakes. ## How to choose Match the tool to the job. Pick **Lindy** for inbox triage and workflow automation. Pick **Poke** or **April** for a personal text assistant. Pick **Martin** if you want to talk to it. Pick **Catch** if you want proactive action with flat, call-inclusive pricing. Pick **Arlo** if you need an [AI colleague](/glossary/ai-colleague) for a team — one assistant across iMessage, Slack, WhatsApp, and phone, with approvals and an audit trail. Still deciding between the personal options? Our [Poke alternatives](/compare/poke-alternatives) overview and the [AI executive assistant](/glossary/ai-executive-assistant) explainer go deeper on the category, and the full [comparisons hub](/compare) lines up the head-to-heads. Two head-to-heads go deeper than this roundup can: [Arlo vs April](/compare/arlo-vs-april) for the text-first personal EA lane, and [Arlo vs Catch](/compare/arlo-vs-catch) for the flat-rate exec assistant with calls included. ## Frequently asked questions **What is the best AI executive assistant in 2026?** It depends on the job. Lindy leads on inbox triage and workflow automation, Poke and April are strong personal texters, Martin is voice-first, and Catch takes proactive action with flat, call-inclusive pricing. Arlo is the team-scale pick — an AI colleague across iMessage, Slack, WhatsApp, and phone calls with approvals and a full audit trail. **What is the best AI executive assistant for a founder?** If you're running operations largely yourself and want one assistant that reaches you wherever you are, Arlo fits: it works in iMessage and SMS, Slack, WhatsApp, and on live calls, with one memory across all of them and an approval before anything sends. If you only need your own inbox and calendar handled, Lindy, Poke, or Martin will feel lighter and cheaper — see [AI assistant for founders](/use-cases/ai-assistant-for-founders). **What is the best AI executive assistant for a small team?** Look for three things a personal assistant won't give you: shared context across people, an approval step before an action goes out in someone's name, and a record of what happened. Arlo is built around those defaults. Most of the alternatives in this roundup are designed for a single user, which is a genuine strength for personal work and a real gap for shared work. **How much do AI executive assistants cost in 2026?** Roughly: Lindy ~$29.99–$199.99 per user/mo with a 7-day trial; Catch ~$99/mo with calls included; April ~50 messages free then ~$25/mo; Poke light-free with heavier use paid; Martin per its own site. Arlo is the outlier on this list because you pay for work done rather than seats: credits at 400 per $1, on monthly plans from $20 with top-ups up to $50,000, drawn down by what you use and good for 12 months from purchase. Pricing moves fast — check each vendor before you commit. **Which AI executive assistants can make phone calls?** As of 2026, Catch places outbound calls, Martin is voice-first for spoken interaction, and Arlo [places calls](/use-cases/ai-executive-assistant-that-makes-phone-calls) for you and keeps a transcript of each one. Most text-first assistants do not make calls at all. **When is Arlo the wrong choice?** When you only need a personal assistant for your own texts and errands. Poke, April, or Martin will feel lighter and cheaper, and the governance layer that makes Arlo useful for a team is dead weight for a single user. Arlo earns its keep when the work is shared and carries stakes. ## Try Arlo If the assistant you need has to act for a team and stand up to review, [try Arlo](/signup) — an AI colleague inside the tools you already use, governed by design and auditable end to end. --- # Catch alternatives — AI assistants for teams and personal life > The main Catch alternatives are Poke, Martin, Lindy, and Arlo. Catch, Poke, and Martin are personal AI assistants over text and voice; Lindy automates work across business apps. Arlo is an AI colleague across iMessage, Slack, WhatsApp, and phone calls with approval gates and a full audit trail. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/catch-alternatives If you're shopping for Catch alternatives, the first thing to sort out is what job you're hiring for. Catch is a personal AI assistant you talk to. Poke and Martin sit in the same lane. Lindy is a work-automation tool. Arlo is an AI colleague built for a team — it works across every channel your team uses, makes phone calls, and holds every action behind an approval gate. Those are genuinely different products, so the right pick depends on whether you're outfitting yourself or a team. Everything below is real to the best of our knowledge as of July 2026. Because we can't verify each competitor's live feature set, treat the specifics as "as of July 2026." After the rundown, I'll make the case for [Arlo](/) if you want one governed assistant spanning iMessage, Slack, WhatsApp, and live calls. ## The main Catch alternatives at a glance | Assistant | Where it works | Best for | | --- | --- | --- | | Catch | Text and chat (personal assistant) | Personal tasks and daily help | | Poke | iMessage, SMS, Telegram (WhatsApp in some markets) | Personal tasks over text | | Martin | Voice, text, email | Voice-first scheduling and reminders | | Lindy | Email, Slack, calendars, business tools | Workflow automation at work | | Arlo | iMessage/SMS, Slack, WhatsApp, live phone calls | An AI colleague across a team's tools, with governance | ## Catch Catch is positioned as a personal AI assistant — the kind you message to manage your day and get quick help. As of July 2026 the pitch is a friendly, conversational helper for one person's tasks and reminders. That's a real and useful lane. Want the direct head-to-head? See [Arlo vs Catch](/compare/arlo-vs-catch). ## Poke Poke comes from The Interaction Company and the whole pitch is that using it feels like texting a friend. As of July 2026 it runs over iMessage, SMS, and Telegram, with WhatsApp in some markets. The sweet spot is everyday personal stuff — planning your day, juggling your calendar, the back-and-forth you'd do over text anyway. For the direct comparison, see [Arlo vs Poke](/compare/arlo-vs-poke). ## Martin Martin is the voice-first one. As of July 2026 it hooks into your calendar, email, and messaging, and you can talk to it, text it, or email it. It's mostly described around scheduling, reminders, drafting quick messages, and daily updates. If what you want is something you can speak to like a personal assistant, that's Martin's lane. See [Arlo vs Martin](/compare/arlo-vs-martin) for the head-to-head. ## Lindy Lindy is built to get work done, not chat. As of July 2026 it's a workflow automator that plugs into email, Slack, your calendar, and a long list of business apps — triaging inboxes, sitting in on meetings, and chaining multi-step workflows on its own. If your problem is repetitive business processes rather than a chatty assistant in your texts, Lindy is the obvious pick here. ## Why Arlo is the best Catch alternative for teams Catch, Poke, and Martin are personal assistants — one person, their tasks, their texts. That's a fine fit for an individual. But if you're bringing an assistant into a team, the questions change: Can everyone reach it where they already work? Can you trust it to act? Can you prove what it did? Arlo is built for exactly those questions. - **It works everywhere, including the phone.** Arlo is in iMessage and SMS, Slack, WhatsApp, and live phone calls. No new app to learn, and no dropped context when a thread becomes a call. - **It remembers, and it briefs you.** Arlo holds onto what matters across conversations and sends a morning briefing on what changed overnight, so nobody re-explains themselves every time. - **It acts, and connects to real tools.** Arlo works across 3,000+ tools. When an app has no API, it signs in once through a secure browser session and operates it by hand — clicking, reading, and filling forms the way a person would. - **Governance is the default, not an add-on.** Every send and every tool write pauses for a reviewer you choose, and every action lands in a full [audit trail](/glossary/ai-agent-audit-trail). That's the real line between an AI colleague you can put on real work and a personal chat-bot. See how [approvals before sending](/use-cases/ai-assistant-with-approval-before-sending) work. - **It scales past one person.** Arlo runs on credits at 400 per $1, on monthly plans from $20 up to $50,000 as you scale. Admin policy and audit export are available on request for orgs that need them. ## How to choose Pick **Catch, Poke, or Martin** if you want a personal assistant for your own tasks, texts, and reminders. Pick **Lindy** if the goal is automating business workflows across apps. Pick **Arlo** if you want one [AI colleague](/glossary/ai-colleague) across iMessage, Slack, WhatsApp, and calls, with memory and governed actions your whole team can audit. Comparing options for a specific team? See how Arlo works [in Slack](/use-cases/ai-assistant-in-slack) or as an assistant that [makes phone calls](/use-cases/ai-executive-assistant-that-makes-phone-calls), and browse the full [use cases](/use-cases) and [comparisons](/compare). ## Frequently asked questions **What are the best Catch alternatives?** Poke, Martin, Lindy, and Arlo. Catch, Poke, and Martin are personal AI assistants over text and voice; Lindy automates work across business apps. Arlo is an [AI colleague](/glossary/ai-colleague) across iMessage and SMS, Slack, WhatsApp, and phone calls with approval gates and a full audit trail. **What is the best Catch alternative for a team?** Arlo. Catch is a polished admin assistant for one executive; Arlo is built for shared work, adding 3,000+ integrations, including apps with no API, and a logged record of every action. See the head-to-head at [Arlo vs Catch](/compare/arlo-vs-catch). **Is there a cheaper alternative to Catch?** It depends on your volume. As of August 2026 Catch is a flat $99/month with voice calls included and a 10-day free trial. Arlo starts free, then runs on monthly plans from $20: credits at 400 per $1 that you spend on what you use — about 30 credits a minute on calls, about 32 for a message. Whether Catch's flat monthly rate is better value depends on how much calling you do. **Which Catch alternative supports WhatsApp?** Arlo does — you can text it on WhatsApp as well as in iMessage and SMS — and Martin lists it too. Catch itself works in WhatsApp, so if that is the only channel you need, it isn't a reason to switch. **Do the alternatives ask before taking action?** Catch is consent-based and asks when unsure, which is the right default. Arlo goes further: connections resolve through policy before a run, sends and writes wait on a named reviewer, and every tool call, source, and approval lands in a [full audit trail](/glossary/ai-agent-audit-trail). ## Try Arlo Arlo is the Catch alternative for teams that want one assistant everywhere they work, with an approval gate on every send and a full audit trail. [Try Arlo](/signup) to put a colleague inside the tools you already use. --- # ChatGPT Dots alternatives: price, review and Dots vs Muse > ChatGPT Dots is OpenAI's always-on agent, launched on 29 September 2026 inside ChatGPT. It needs ChatGPT Pro, from $100 a month, or a Business Premium seat, and isn't on Free, Go or Plus. The main alternatives are Arlo for work (text it on iMessage, SMS or WhatsApp, use it in Slack, and it places calls), Muse and Pally (free with weekly limits), Instinct (invite-only) and Hark Pro (an app). Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/chatgpt-dots-alternatives OpenAI announced Dots on 29 September 2026 at its DevDay keynote in San Francisco: always-on agents that live inside ChatGPT, each with its own cloud computer, that work toward goals you set and come back to you for decisions ([OpenAI](https://openai.com/index/introducing-dots/)). Early reviewers found it strongest at work tasks. The catch for most people is getting in. Dots isn't on ChatGPT's Free, Go or Plus plans, so the cheapest way in costs $100 a month, and texting your dot is a limited US beta. This page covers what Dots is and costs, how it asks before acting, what has been reported about its safety, and five alternatives: Arlo, Muse, Pally, Instinct and Hark Pro. It's written by Arlo, which makes one of them, so weigh our section accordingly. Facts about Dots come from OpenAI's own pages and dated reporting as of October 2026, and we say where Dots is the better pick. ## What ChatGPT Dots is (as of October 2026) - **Who makes it:** OpenAI. Sam Altman announced it at DevDay on 29 September 2026, and it has been rolling out in stages since; OpenAI's help centre says access may take several days to reach an account. OpenAI writes the name in lowercase, "dots", and calls each agent "your dot". Dots run on OpenAI's GPT-6 Astra model. - **Where you use it:** inside ChatGPT, in the desktop app for Mac and Windows, on the web, and in the iPhone and Android apps after setup, where mobile access is available. You have to create your dot on a desktop: you can't set one up on a phone, and mobile web isn't supported ([OpenAI Help](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot)). You can also bring your dot into a Slack channel. - **Texting:** a limited beta for Pro users in the US only, through a third-party provider, and not in Business or Enterprise workspaces. OpenAI says access "may not be available to every Pro user", and its marketing pages still say texting is coming soon. No dots page mentions WhatsApp. - **How it works:** each dot has its own cloud computer and browser, running Linux and Chrome, which you can open to watch or take over. With your permission it can also use your own computer through the ChatGPT desktop app, and it can start Codex tasks. It reaches more than 4,000 apps through plugins it shares with ChatGPT and Codex. At launch you get one primary dot; OpenAI says you'll be able to add more. - **Memory:** it starts with your ChatGPT memories and recent context, and forms its own, including from connected apps. - **Working on its own:** when you aren't using it, it researches in the background with read-only tools that can't send messages, change app content or control a browser, a limit OpenAI says it enforces in code. It runs reminders and recurring checks you schedule, and messages you with progress or questions. - **Calls:** you can call your dot by voice inside the ChatGPT app. "Your dot cannot initiate calls to you at launch", and OpenAI doesn't list phone calls to businesses or people on your behalf. - **For teams:** ChatGPT Space, a shared workspace where teammates, ChatGPT and dots build on shared knowledge, comes with Pro, Business and Enterprise plans ([OpenAI](https://openai.com/index/devday-2026-recap/)). Specialist dots, with their own identity and credentials for a team role, are in enterprise pilots only. - **Who can use it:** adults 18 and over on the plans below. Pro isn't available in the EEA, Switzerland or the UK at launch. Business Premium is available in every supported ChatGPT region. ## What ChatGPT Dots costs (as of October 2026) Dots comes only with ChatGPT's higher-priced plans. OpenAI's [pricing page](https://chatgpt.com/pricing/) lists it under Pro and not under Free, Go or Plus. | ChatGPT plan | Dots included? | Price | |---|---|---| | Free, Go, Plus | No | — | | Pro | Yes, your first dot at no extra cost | $100, $200 or $500 a month ([OpenAI Help](https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers)) | | Business, Standard seat | No | $20 a user a month on annual billing, $25 monthly | | Business, Premium seat | Yes | $100 a user a month on annual billing, $125 monthly ([OpenAI](https://openai.com/business/pricing/)) | | Enterprise, Edu, Healthcare | In beta, off until an admin turns it on | Custom | What that buys is still being set. Conversations with your dot don't count toward ChatGPT's usage limits, but tasks it starts in Codex or ChatGPT Work do. For the first month after launch, dots usage doesn't count against eligible plans' allowances, and OpenAI says it will share usage terms for each plan after that ([release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes), 29 September 2026). Users have also hit an unannounced cooldown with the message "Your dot is on a break", and on 6 October an OpenAI Support reply said it couldn't confirm how long the cooldown lasts ([OpenAI Community](https://community.openai.com/t/dot-is-on-a-break-check-back-in-a-bit/1403373)). OpenAI says Dots won't stay premium-only. Sam Altman and CFO Sarah Friar said a mass-market version will come, without giving a date or price ([Business Insider](https://www.businessinsider.com/openai-dots-new-ai-agents-wont-stay-premium-product-forever-2026-9), 29 September 2026). ## How Dots asks before it acts Both products ask first: Dots before it sends or buys, and Arlo before anything it sends or changes, unless you've pre-approved that kind of action. Dots gives you detailed control over when. Per OpenAI's [privacy and safety FAQ](https://help.openai.com/en/articles/20001529-dots-privacy-security-and-safety-faqs) and its [safety post](https://openai.com/index/how-we-build-safety-security-and-privacy-into-dots/): - **Always handed to you:** the most sensitive actions, such as changing a password or transferring money, need you to take over. No setting changes that. - **Asked every time:** permanently deleting data, installing or running software from an unrecognised source, and granting new security-sensitive access. - **Approvable in advance:** some actions, such as recurring messages. Purchases with a card you've saved on a merchant's site need approval, which can be given ahead of time if it specifically covers that purchase. Approving one message doesn't give your dot ongoing permission to contact people. - **Sharing information:** the approval has to cover both what is shared and the kind of recipient. Health information always needs a named recipient. - **Custom Rules:** for each kind of action you choose "Take action without asking", "Take action if pre-approved", "Ask before taking action" or "Hand off to you". Your dot can draft rules, but changing them needs your approval. - **Auto-review:** a separate model, carried over from Codex, checks any planned action that could affect your accounts or share information against your instructions, your rules and OpenAI's safety rules. For an email it checks the recipient and the content. Your dot can't turn it off. - **Activity View:** the desktop app shows ongoing and delegated tasks, their status and the steps your dot has taken, and its profile lists what's in progress, scheduled and completed. You can open its computer to inspect the work. Arlo writes every tool call, source and approval to an [audit trail](/glossary/ai-agent-audit-trail) you can read back. On approvals and a record of what was done, the two are close. ## Is ChatGPT Dots safe? OpenAI has published a lot about Dots' safeguards, and the product is barely a week old. Here is what's on record, as of October 2026. - **The launch context.** Dots launched a day after OpenAI apologised for one of its agents accessing Australian government systems in June, and a day after it shelved its GPT-6.1 Astra model. Its head of safety systems, Saachi Jain, said that model "didn't quite meet the bar in terms of staying within scope and authorization" ([NBC News](https://www.nbcnews.com/tech/tech-news/openai-launches-dots-ai-agents-safety-questions-rcna600338), 29 September 2026). Dots run on the earlier GPT-6 Astra. - **Published test figures.** In OpenAI's system card, as reported, flagged boundary problems rose from 8.6% to 19.7% when chained tasks doubled from five to ten, with no high-severity breaches ([The New Stack](https://thenewstack.io/openai-dots-agent-permissions/), 30 September 2026). - **Memory you can't edit.** You can't view, edit or delete individual dot memories; the only way to clear the dot's own context is to reset it. Disconnecting an app doesn't delete what the dot already learned from it. Its context doesn't keep credentials, images or screenshots ([OpenAI Help](https://help.openai.com/en/articles/20001529-dots-privacy-security-and-safety-faqs)). - **Training.** On personal plans, ChatGPT's "Improve the model for everyone" setting covers your dot's conversations, actions and the connected-app data used in them. Business, Enterprise and Edu don't train on it by default. OpenAI says human review may occur in limited circumstances even with model improvement turned off. - **What isn't documented.** We couldn't find an exportable or admin audit log for a primary dot. The New Stack noted that OpenAI hasn't said whether a dot's actions in services like GitHub or Slack are logged under your identity or marked as an agent's. - **First-week bug reports.** Users on OpenAI's community forum have reported voice calls that fail to connect ([2 October](https://community.openai.com/t/dot-voice-calls-fail-to-connect-on-iphone-mac-desktop-and-web-oct-1-2/1402858)), a dot that couldn't resume after a safety pause ([2 October](https://community.openai.com/t/dot-cannot-resume-after-safety-pause-http-409-your-dot-s-safety-failure-has-changed/1402937)), a cloud browser that stayed blocked after a manual sign-in ([5 October](https://community.openai.com/t/dot-cloud-browser-stays-blocked-after-completed-manual-sign-in/1403535)), and the unannounced cooldowns above. These are user reports from the first week, and some may already be fixed. The same three questions apply to Dots and to every alternative here: does it ask before it sends or buys, is there a record you can trust of what it did, and what happens to your data when you disconnect an account or leave? ## The alternatives at a glance | Assistant | To start | Where you use it | Built for | Phone calls | |---|---|---|---|---| | **Arlo** | Free to start, no card | iMessage, SMS, WhatsApp, Slack | Work: founders, operators, teams | Places calls, through phone menus and hold | | **Muse** | Free up to a weekly limit; US and Canada | Its own apps (iPhone, iPad, Android, Mac), web, WhatsApp | Personal life and small business | Calls US businesses, in beta | | **Pally** | Free for everyone, with a weekly limit | iMessage, Android RCS and SMS, Telegram | Personal life | Included, within the weekly limit | | **Instinct** | Invite only, free | iMessage, SMS, WhatsApp, or a call | Personal use | Calls you; calls businesses in early access | | **Hark Pro** | Free with a cap | iPhone, Android and web apps | Personal life | None announced | | ChatGPT Dots | ChatGPT Pro from $100 a month, or a Business Premium seat | Inside ChatGPT, plus Slack; texting in a limited US beta | Work and personal | Voice calls with your dot in the app; none on your behalf listed | ## Arlo: if you want to text it, or have it make the call Arlo is an [AI colleague](/glossary/ai-colleague) you text on [iMessage](/imessage), SMS or [WhatsApp](/whatsapp), use inside [Slack](/slack), or hand a [phone call](/phone) to. It places calls to people and businesses, and it can work through phone menus and wait on hold. It connects to Gmail, your calendar, Slack, Notion, Linear, HubSpot and 3,000+ other tools, and reaches sites with no API through a secure browser session you sign into once. Anything it sends or changes waits for your approval unless you've pre-approved that kind of action, and every tool call, source and approval is written to an audit trail you can read back. It [texts you first](/proactive) with briefs and reminders over iMessage and SMS. It's built for work that other people depend on: a company's inbox, a team's Slack, customers and vendors on the phone. Anyone can start today, with no invite or waitlist. Arlo is free to start, no card, then plans from $20 a month. How it compares with ChatGPT itself: [Arlo vs ChatGPT](/compare/arlo-vs-chatgpt). ## Muse: free, from Meta, in the US and Canada Muse, Meta's AI agent, is the product Dots is most often measured against; Wired, The Verge and Axios all framed Dots as OpenAI's answer to it. It launched on 8 September 2026 and, like Dots, runs your errands on a cloud computer with a browser you can watch and step into. You use it in its own app on iPhone, iPad, Android, the Mac and the web, or connect it to WhatsApp and text it there. It's free up to a weekly usage limit, with Power at $20 a month and Max at $100 a month for more ([Meta](https://www.meta.com/help/subscriptions/1021145227643680/)), and it's open to adults in the US and Canada. It pays with one-time Link by Stripe card numbers, asks before completing a purchase, and is testing calls to US businesses. Two things to weigh: by default Muse asks before only some actions, and Meta uses Muse data for training unless you opt out. More: [Muse alternatives](/compare/muse-alternatives) and [Instinct vs Muse](/compare/instinct-vs-muse). ## Pally: free, and you text it Pally (pally.com), a Y Combinator company, is an assistant you chat with over iMessage, Android RCS and SMS, or Telegram. Since 6 October 2026 every feature is free for everyone, under one usage limit that refills through the week, and its Pro and Max subscriptions are gone ([Pally's terms](https://pally.com/terms)). Calls are included under that limit, and it works through phone menus and waits on hold. It books flights, hotels and rental cars, and pays with a single-use card capped at the amount you approve (US only for now) or with your saved card. It reads your WhatsApp inbox, but WhatsApp isn't where you chat with it. Pally shows you what it proposes word for word, though things you explicitly ask for can go out without a second confirmation, and since 6 October its [privacy policy](https://pally.com/privacy) lets it train on your data unless you opt out. More: [Pally alternatives](/compare/pally-alternatives) and [Arlo vs Pally](/compare/arlo-vs-pally). ## Instinct: text it or call it, if you can get an invite Instinct, the AI assistant from Spear Street Technology, is an assistant you text over iMessage, SMS or WhatsApp, or call, and it can call you. It runs a computer in the cloud with the logins you give it and keeps working between your messages, and calls it places to businesses are in early access. It raised $1 billion at a $10 billion valuation on 28 September 2026 ([TechCrunch](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/)). Getting in is the hard part. As of October 2026 it's invite-only: each member gets five invites, and everyone else joins a waitlist ([Fortune](https://fortune.com/2026/09/30/noah-shinn-instinct-ai-assistant-meta-muse-alexandr-wang-tech-series-c-ai-agent-mark-zuckerberg/), 30 September 2026). It's free, with no published pricing. Its terms license it for personal use only, and it trains on your materials unless you opt out. More: [Instinct alternatives](/compare/instinct-alternatives) and [Arlo vs Instinct](/compare/arlo-vs-instinct). ## Hark Pro: an app, free with a cap Hark Pro, from Brett Adcock's AI lab Hark, launched on 6 October 2026 as a free iPhone, Android and web app ([Hark](https://hark.com/articles/introducing-hark-pro)). Every feature is free up to a usage cap Hark hasn't published, and $20 or $100 a month buys more. Like a dot, its browser agent gets its own cloud computer for each task, and a proactive home screen turns your email and calendar into one-tap tasks. It asks before it spends your money or sends a message, unless you've told it otherwise. You use it in its app: as of October 2026 Hark hasn't announced a number to text or phone calls. More: [Hark Pro alternatives](/compare/hark-alternatives) and [Arlo vs Hark Pro](/compare/arlo-vs-hark). ## Where Dots is the better pick - **Long desktop and coding work.** Each dot has its own cloud computer that you can open to watch or take over. It can start Codex tasks, and with your permission it can use your own computer through the ChatGPT desktop app. If your work is long tasks at a desk, this is the strongest reason to pick Dots. - **You already live in ChatGPT.** A dot starts with your ChatGPT memories and recent context, runs on OpenAI's GPT-6 Astra model, and shares plugin permissions with ChatGPT and Codex across more than 4,000 apps. If you already pay for ChatGPT Pro, your first dot costs nothing extra. - **Background research that can't act.** While you're away, a dot works with read-only tools that OpenAI says are limited in code, so its proactive research can't send anything or change your apps. - **Approvals you can tune.** Custom Rules per kind of action, Auto-review of planned actions that touch your accounts or share information, and an Activity View of each step. - **OpenAI's scale.** ChatGPT has 1.2 billion weekly users, according to OpenAI's DevDay recap. Dots work with ChatGPT Space and Pages, where teammates and dots build on shared documents, and Business Premium seats include Dots in every supported ChatGPT region. ## Where Arlo is the better pick - **You want to text it today.** Arlo works in the iMessage, SMS and WhatsApp threads you already use. Dots texting is a limited beta for US Pro users, and no dots page mentions WhatsApp. - **You want it to make the call.** Arlo places calls to people and businesses, works through phone menus and waits on hold. OpenAI doesn't list phone calls on your behalf for Dots. - **Price of entry.** Arlo is free to start, no card, then plans from $20 a month. Dots starts with a ChatGPT subscription at $100 a month. Dots details on this page reflect OpenAI's pages and dated reporting as of October 2026. Dots is rolling out in stages, so check OpenAI's help centre for current plans, regions and texting. ## Frequently asked questions **What are the best ChatGPT Dots alternatives?** Arlo for work: you text it on iMessage, SMS or WhatsApp, use it in Slack, and it places phone calls for you. Muse, Meta's AI agent, is free up to a weekly limit for adults in the US and Canada. Pally is free for everyone, and you text it. Instinct is free but invite-only. Hark Pro is a free app with a usage cap. Each of them costs less to start than Dots. **Is ChatGPT Dots free?** No. As of October 2026 Dots isn't included with ChatGPT's Free, Go or Plus plans. It comes with ChatGPT Pro, where your first dot costs nothing on top of the $100, $200 or $500 monthly price, and with a Business Premium seat. OpenAI says a mass-market version will come, with no date or price yet. Muse and Pally are free with weekly limits, Hark Pro is free with a cap, and Instinct is free with an invite. Arlo is free to start, no card, then plans from $20 a month. **How much does ChatGPT Dots cost?** At least $100 a month. ChatGPT Pro comes in $100, $200 and $500 tiers, and all three include Dots. A Business Premium seat costs $100 a user a month on annual billing or $125 month to month ([OpenAI](https://openai.com/business/pricing/)). Some launch coverage got this wrong: BGR reported Dots only on the $200 tier, but The Verge tested it on the $100 one ([The Verge](https://www.theverge.com/ai-artificial-intelligence/1004096/openai-chatgpt-dots-hands-on-agent), 2 October 2026), and CBS News gave $20 a month for Business Premium, which is the Standard seat's price. OpenAI says it will publish usage terms for each plan after the first month. **What's the difference between ChatGPT Dots and Muse?** Both are always-on agents with their own cloud computer and a browser you can watch and take over, and both ask before they buy. Muse is free up to a weekly limit for adults in the US and Canada, works in its own apps and on WhatsApp, and is built for personal errands and small business. Dots costs at least $100 a month, lives inside ChatGPT, connects to Codex, and leans toward work: The Verge's hands-on found it stronger at work tasks than at errands. The Verge also asked whether Dots can compete with free ([The Verge](https://www.theverge.com/ai-artificial-intelligence/1003399/meta-openai-ai-agents-muse-dots-battle), 1 October 2026). For personal errands in North America, Muse is the cheaper answer. **Is ChatGPT Dots safe?** OpenAI has published detailed guardrails. Password changes and money transfers always go back to you, a separate Auto-review model checks emails and other sharing before they go out, and background research is read-only. The open questions are memory you can't edit one item at a time, a model-improvement setting on personal plans that covers your dot's conversations and actions, and the bugs users reported in its first week. It also launched a day after OpenAI apologised for one of its agents accessing Australian government systems. **Is ChatGPT Dots any good?** Early reviews are positive about work. Casey Newton estimated his dot did about two hours of work for about 15 minutes of his time ([Platformer](https://www.platformer.news/openai-dots-agents-devday-2026/), 30 September 2026). The Verge's hands-on found it stronger at work tasks than at errands. The criticism is about access and cost: [ZDNET](https://www.zdnet.com/innovation/openai-dots-openclaw-chatgpt-pro/) called the cost structure open-ended and underdefined, Benedict Evans called the launch confused, and users in Europe and the UK have complained that they can't get it on Pro. **Can you text ChatGPT Dots?** Only in a limited beta. As of October 2026 texting is open to some Pro users in the US, through a third-party provider, and not to Business or Enterprise workspaces, while OpenAI's marketing pages still say texting is coming. No dots page mentions WhatsApp. Separately, the ChatGPT desktop app can read and send your iMessages from an Apple silicon Mac ([how that works](/blog/chatgpt-in-imessage)); OpenAI hasn't said whether a dot can use it. To text an assistant today, Arlo works in iMessage, SMS and WhatsApp, Pally in iMessage, RCS, SMS and Telegram, and Instinct in iMessage, SMS and WhatsApp with an invite. More: [the best AI assistants you can text](/compare/best-ai-assistants-you-can-text). **Can ChatGPT Dots make phone calls?** You can call your dot by voice inside the ChatGPT app, though users reported those calls failing to connect in its first week. Your dot can't call you at launch, and OpenAI doesn't list phone calls to businesses or people on your behalf. Arlo places calls to people and businesses, works through phone menus and waits on hold. Pally includes calls, and Muse and Instinct call businesses in beta or early access. More: [AI assistants that can make phone calls](/compare/ai-assistants-that-can-make-phone-calls). **Is ChatGPT Dots available in the UK and Europe?** On ChatGPT Pro, not yet. At launch Pro excludes the EEA, Switzerland and the UK, and OpenAI hasn't given a date. Business Premium seats include Dots in every supported ChatGPT region. Sam Altman told NBC News the EU process "takes longer, and sometimes unpredictably longer". --- # Fyxer Alternatives — Beyond the Email Inbox > Most Fyxer alternatives are other AI email tools — Superhuman, Gmelius, Missive — that, like Fyxer, live inside Gmail and Outlook. Arlo is the different kind of alternative: an AI colleague that works across email, iMessage, Slack, WhatsApp, and phone calls, connects to 3,000+ tools, and keeps an approval gate and audit trail on every action. Updated: 2026-07-04 · Canonical: https://www.arlo.sh/compare/fyxer-alternatives Fyxer is a well-liked AI email assistant: as of July 2026 it organizes your inbox, drafts replies in your tone, and takes notes in your meetings, all inside Gmail and Outlook. If email is the whole problem, it's a solid tool. But if you're looking for a *Fyxer alternative*, it's usually because you want help that doesn't stop at the inbox — and most of the obvious alternatives have the same ceiling. ## The two kinds of Fyxer alternative **Other email assistants.** Superhuman, Gmelius, and Missive live in the same place Fyxer does — your email client. They're faster inboxes and better drafting, and if that's your bottleneck, any of them is a reasonable swap. What none of them change is the boundary: the help ends where email ends. No Slack, no phone. **An AI colleague across your whole stack.** [Arlo](/glossary/ai-colleague) is the other kind of alternative — not a better inbox, but an assistant that treats email as one channel among several. It does the inbox work Fyxer does, then keeps going into the tools and conversations that live outside it. ## Fyxer vs. Arlo | | Fyxer | Arlo | | --- | --- | --- | | Email triage and drafting | Yes | Yes | | Meeting notes | Yes | Yes | | Slack | No | Yes | | iMessage / SMS | No | Yes | | Phone calls | No | Yes, places them | | Integrations | Gmail & Outlook | 3,000+ (incl. no-API via browser) | | Takes action in other tools | Limited to email | Yes, across your stack | | Approval gate + audit trail | Not stated | Yes | ## Why teams switch to Arlo - **It works past the inbox.** The same assistant that triages your email also answers in [Slack](/use-cases/ai-assistant-in-slack), texts in iMessage and WhatsApp, and [picks up the phone](/use-cases/ai-executive-assistant-that-makes-phone-calls). Context follows the work instead of resetting at the edge of Gmail. - **It acts across your tools.** Beyond drafting a reply, Arlo opens the ticket, updates the CRM, and runs the follow-up — across 3,000+ integrations, including internal systems with no API. - **It's governed.** Anything that sends or writes waits for a reviewer you choose, and every action lands in a [full audit trail](/glossary/ai-agent-audit-trail). That's the difference between a smart inbox and something a team can hand real work to. ## How to choose If your bottleneck is strictly email and meeting notes, Fyxer — or another inbox tool like Superhuman — will serve you well. If you want one assistant that does the email work *and* everything around it, across your team's channels and with governance you can audit, that's Arlo. See how it fits as an [AI executive assistant](/glossary/ai-executive-assistant) or [for founders](/use-cases/ai-assistant-for-founders), or browse all [comparisons](/compare). ## Frequently asked questions **What are the best Fyxer alternatives?** They come in two kinds. Other email assistants — Superhuman, Gmelius, Missive — live inside your inbox like Fyxer does and make it faster. Arlo is the other kind: an [AI colleague](/glossary/ai-colleague) that does the inbox work and then keeps going into Slack, iMessage, WhatsApp, and phone calls. **Is there a Fyxer alternative that works outside email?** That's the main reason to look past the inbox tools. Fyxer, Superhuman, Gmelius, and Missive all stop where email stops. Arlo treats email as one channel among several and connects to 3,000+ tools, including internal systems with no API. **Can a Fyxer alternative make phone calls?** Arlo [places calls](/use-cases/ai-executive-assistant-that-makes-phone-calls) and keeps a transcript of each one. Fyxer and the other email-native tools do not. **Does Fyxer work with Slack?** No — Fyxer is an email tool, living in Gmail and Outlook. Arlo works in [Slack](/use-cases/ai-assistant-in-slack), [WhatsApp](/whatsapp), iMessage and SMS, and on calls, with one memory across all of them. **Do Fyxer alternatives ask before sending an email?** Arlo does: anything that sends or writes waits for a reviewer you choose, and every action lands in a [full audit trail](/glossary/ai-agent-audit-trail). An approval gate is not a stated feature of Fyxer's public materials. **When is Fyxer still the better choice?** When your bottleneck genuinely is email and meeting notes. Fyxer, or an inbox tool like Superhuman, will serve you well and feel lighter than a full AI colleague. Switch when the work you want handled starts outside the inbox. ## Try Arlo Arlo is the Fyxer alternative for people who want more than a better inbox — an AI colleague across email, chat, and calls, with a full audit trail. [Try Arlo](/signup) and hand off the first thing that isn't email. --- # Hark Pro alternatives — pricing, privacy and what else to try > Hark Pro, from Brett Adcock's AI lab Hark, launched on 6 October 2026 as a free iPhone, Android and web app, with $20 and $100 monthly tiers for more usage. The main alternatives are Arlo for work (text it on iMessage, SMS or WhatsApp, use it in Slack, and it places calls), Pally, Muse, Instinct and Poke. Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/hark-alternatives Hark Pro, from Brett Adcock's AI lab Hark, went live to everyone on 6 October 2026 ([Hark](https://hark.com/articles/introducing-hark-pro)). Adcock also founded Archer Aviation and Figure AI, and the launch drew same-day coverage from Bloomberg, TechCrunch, Gizmodo and Fast Company. This page is about that Hark Pro, from hark.com. The Hark Audio app and the Hark app for Shopify stores are separate products. The questions people are searching for are what it costs, whether it's safe to hand it your accounts, and what else does the same job. This comparison is written by Arlo, one of the alternatives, so weigh our section accordingly. Hark's facts come from its launch post, terms, privacy policy, app-store listings and dated press, as of October 2026, and we say where Hark Pro is the better pick. ## What Hark Pro is (as of October 2026) - **Who makes it:** Hark Labs, Inc. of San Jose, California. Brett Adcock founded it and is still CEO of Figure. Its head of design, Abidur Chowdhury, presented the iPhone Air at Apple before joining. - **Funding:** about $100 million of Adcock's own money, then a $700 million-plus Series A at a $6 billion valuation in May 2026 ([Hark](https://hark.com/articles/hark-announces-usd700-million-fundraising-round)). NVIDIA and AT&T are partners, and AT&T invested in August 2026 ([Hark](https://hark.com/articles/at-and-t-and-hark-formalize-strategic-partnership)). - **Where it runs:** an iPhone app, an Android app, and the web at hark.com. Hark hasn't announced a number you can text, WhatsApp, or phone calls. Its terms do include consent wording for automated texts, so a texting channel may come. - **How it works:** one long conversation, plus Projects, which are separate threads with their own files. Its browser agent, Handoff, gets its own cloud computer for each task and can run up to six browsers at once. It connects to Gmail and Google Workspace, Outlook, Google Calendar and Slack, and to other services through MCP connectors. - **Proactive Home:** cards that turn what's in your email and calendar into tasks, one-tap Action Buttons such as pay, cancel, RSVP or review, and Panels, small apps you describe in words. - **Errands it names:** DoorDash and Uber Eats orders, checkouts at Target, Walmart or Costco, OpenTable and Resy bookings, flights on United, Delta, American or JetBlue, hotels on Booking.com, paying bills and cancelling subscriptions. - **Approvals:** it asks before it spends your money or sends a message on your behalf, unless you've told it otherwise. You can give it standing permission for a category of purchases. - **Who can use it:** adults, 18 and over, per its terms. A Hark device with its own AT&T connection is due in 2027 ([TechCrunch](https://techcrunch.com/2026/10/06/hark-releases-an-ai-personal-assistant-with-a-focus-on-privacy/), 6 October 2026). ## Hark Pro pricing (as of October 2026) Hark Pro is free, with a cap. Hark's pricing page at hark.com/pricing returned a 404 when we checked on 7 October 2026, so these prices come from its [launch post](https://hark.com/articles/introducing-hark-pro) and the app stores. | Plan | Price on the web | What you get | |---|---|---| | Hark Pro | $0 | Every feature, with a usage cap Hark hasn't published | | Hark Pro² | $20 a month | Twice the usage | | Hark Pro³ | $100 a month | Ten times the usage | Buying inside the apps costs more. On 7 October 2026 the US App Store listed Pro² at $26.00, Pro³ at $129.99 and a fourth tier, Pro⁴, at $200.00, without stating a billing period ([App Store](https://apps.apple.com/us/app/hark-pro/id6772362941)). Google Play listed items from $23 to $115. Hark hasn't announced Pro⁴ anywhere else. Usage top-ups are opt-in, charged when your allowance runs low, and not refundable, per its [terms](https://hark.com/terms). On 4 October, Adcock [posted on X](https://x.com/adcock_brett/status/2106889503919292529) that the first 100,000 sign-ups would get a paid plan for free. Hark's launch post doesn't mention the offer, and reports differ on which plan it covers and for how long, so treat it as unconfirmed until Hark says more. The free, $20 and $100 ladder is the same one Meta set for Muse about four weeks earlier ([The Next Web](https://thenextweb.com/news/hark-pro-agent-hardware-europe), 6 October 2026). ## The alternatives at a glance | Assistant | Can you start today? | Where you use it | Built for | Phone calls | |---|---|---|---|---| | **Arlo** | Yes, free to start | iMessage, SMS, WhatsApp, Slack | Work: founders, operators, teams | Places calls | | **Pally** | Yes, free | iMessage, Android RCS and SMS, Telegram | Personal life and work | Included, within a weekly usage limit | | **Muse** | Yes, in the US and Canada | Its own app, web, Mac, WhatsApp | Personal life and small business | Calls US businesses, in beta | | **Instinct** | Invite only | iMessage, SMS, WhatsApp, or a call | Personal use | Calls you; calls businesses in early access | | **Poke** | Yes | iMessage, SMS, Telegram; WhatsApp in some markets | Personal life | — | | Hark Pro | Yes, free with a cap | iPhone, Android and web apps | Personal life | None announced | ## Arlo: if you want to text it, or the work involves other people Arlo is an [AI colleague](/glossary/ai-colleague) you text on [iMessage](/imessage), SMS or [WhatsApp](/whatsapp), use inside [Slack](/slack), or hand a [phone call](/phone) to. It places calls to people and businesses, and it can work through phone menus and wait on hold. It connects to Gmail, your calendar, Slack, Notion, Linear, HubSpot and 3,000+ other tools, and reaches sites with no API through a secure browser session you sign into once. Anything it sends or changes waits for your approval unless you've pre-approved that kind of action, and every tool call, source and approval is written to an [audit trail](/glossary/ai-agent-audit-trail) you can read back. It [texts you first](/proactive) with briefs and reminders over iMessage and SMS. Two differences from Hark Pro matter most. You reach Arlo in the threads you already use, and it makes the phone call for you; Hark hasn't announced texting or calls. And Arlo is built for work that other people depend on, like a company's inbox, a team's Slack and customers on the phone, while Hark Pro is sold to one person for their own life, with no team plan or admin controls that we could find. Anyone can start Arlo today, with no invite or waitlist. It's free to start, no card, then plans from $20 a month. Head-to-head: [Arlo vs Hark Pro](/compare/arlo-vs-hark). ## Pally: free, and you text it If what drew you to Hark Pro was a free assistant for errands and travel, Pally (pally.com) is the closest match you reach by text. It comes from Pally Technologies, a Y Combinator company, and you chat with it over iMessage, Android RCS and SMS, or Telegram. On 6 October 2026, the same day Hark Pro launched, Pally made every feature free for everyone, with one usage limit that refills through the week. Its Pro and Max plans ended with pro-rata refunds, and the only paid plan left is an invitation-only VIP plan at $1,000 a month ([Pally's terms](https://pally.com/terms)). Calls are included and draw on the same usage limit: it works through phone menus, waits on hold, and you can call it. It books flights, hotels and rental cars, and pays with a single-use card capped at the amount you approve, US only for now. It reads your WhatsApp inbox, but WhatsApp isn't a place to chat with it. Two things to check before you connect it: its privacy policy, updated 6 October, now lists training its models on your data unless you opt out, and messages you explicitly ask it to send can go without a second confirmation. More: [Arlo vs Pally](/compare/arlo-vs-pally) and [Pally alternatives](/compare/pally-alternatives). ## Muse: the same price ladder, from Meta Muse, Meta's AI agent, is the closest like-for-like. It launched on 8 September 2026 and is free with a weekly usage limit, with Power at $20 a month and Max at $100 a month for more ([Meta](https://www.meta.com/help/subscriptions/1021145227643680/)). Like Hark Pro, it runs your errands on a cloud computer with a browser you can watch, comes back with goals and ideas on its own, and keeps an audit trail. You use it in its app on iPhone, iPad, Android, the Mac and the web, or connect it to WhatsApp and text it there. It's available only in the US and Canada, to adults. By default it asks before only some actions, and Meta uses Muse data for training unless you opt out. It has also drawn more privacy reporting than anything else here, including a [TIME](https://time.com/article/2026/10/06/meta-muse-ai-agent-privacy/) report on 6 October 2026 that Muse keeps profiles of the people in a user's life, including people who don't use Muse; Meta didn't dispute it. More: [Muse alternatives](/compare/muse-alternatives). ## Instinct: text it or call it, if you can get in Instinct, the AI assistant from Spear Street Technology, is an assistant you text over iMessage, SMS or WhatsApp, or call, and it can call you. It runs a computer in the cloud with the logins you give it, keeps working between your messages, and has its own email address. Calls it places to businesses are in early access. It raised $1 billion at a $10 billion valuation on 28 September 2026 ([TechCrunch](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/)). Getting in is the hard part. As of October 2026 it's invite-only: each member gets five invites, and everyone else joins a waitlist ([Fortune](https://fortune.com/2026/09/30/noah-shinn-instinct-ai-assistant-meta-muse-alexandr-wang-tech-series-c-ai-agent-mark-zuckerberg/), 30 September 2026). It's free, with no published pricing. Its terms license it for personal use only, and it trains on your materials unless you opt out. More: [Instinct alternatives](/compare/instinct-alternatives) and [Arlo vs Instinct](/compare/arlo-vs-instinct). ## Poke: personal texting, now owned by Cognition Poke, from The Interaction Company in Palo Alto, is built to feel like texting a friend. It runs over SMS and Telegram, with WhatsApp in some markets, and it was one of the first third-party AI agents you could reach through iMessage. It's at its best on everyday personal things, like planning your day and juggling your calendar. Its pricing is negotiated: light requests are free and heavier ones cost. Cognition, the lab behind the Devin coding agent, acquired its maker on 23 July 2026 in a reported low-nine-figure deal, and Poke keeps running. Poke's details here come from our August 2026 check. More: [Arlo vs Poke](/compare/arlo-vs-poke) and [what the Cognition deal means](/blog/cognition-poke-what-it-means). ## Is Hark Pro safe and private? Hark leads with privacy, and its documents back much of that up. Here is what they say, as of October 2026. - **It asks first by default, and you can change that.** Hark's [terms](https://hark.com/terms), updated 1 October 2026, say it will ask you to confirm before it spends your money or sends a communication on your behalf "unless you tell us otherwise". They also provide for "standing permission" to complete a category of transactions without asking. The launch post gives an example: if you allow it, Hark can top up a child's lunch card on its own. - **If a site bans your account, that's on you.** Handoff signs in to websites as you. The terms put the risk on you if a site detects the automation and suspends your account. - **Training: the privacy policy and the terms read differently.** The [privacy policy](https://hark.com/privacy-policy), updated 1 October 2026, says Hark doesn't train on your inputs, outputs, memory, connected-account data or sandbox data, and that no opt-out is needed. The terms separately let Hark use your content to improve the service and for product development, and review it for safety. One outlet, [ProgressiveRobot](https://www.progressiverobot.com/2026/10/06/hark-pro-ai-agent-web-ios-android-launch/), read that wording as a training risk. The privacy policy's no-training statement is the more specific of the two. The terms also bar the outside AI model providers Hark uses from training on your data. - **Passwords and cards** go in what Hark calls "Secured by Hark", an encrypted vault it says even Hark can't see inside. Hark says it will never sell your data or share it with advertisers, and that you can always delete it. - **What you hand over:** to be useful it wants your email, calendar and logins. [TechCrunch](https://techcrunch.com/2026/10/06/hark-releases-an-ai-personal-assistant-with-a-focus-on-privacy/) and [Gizmodo](https://gizmodo.com/hark-pro-ai-ui-2000822253) both pointed out how much of your digital life that is. - **A record of what it did:** you can watch the agent browse in a small live window, and you confirm purchases. We couldn't find a published description of an activity log you can review later. On training, Pally, Muse and Instinct all use your data unless you opt out, per their own policies as of October 2026. Arlo's [privacy policy](/privacy) says it doesn't use your messages, files, memories, calls or connected-account content to train AI models. ## Where Hark Pro is the better pick - **Errands in your own life.** Food delivery, store checkouts, restaurant bookings, airline sites and bills are what Hark built Handoff for, with its own cloud computer for each task. - **A home screen that does the triage.** Proactive cards, Action Buttons and Panels put the next thing to do one tap away. Launch reviews from Gizmodo and Fast Company praised how it's presented. - **Every feature, free, up to a cap.** You can use all of Hark Pro without paying, and $20 a month doubles the usage. - **Staying power.** More than $800 million raised, NVIDIA and AT&T as partners, and a device due in 2027. The full head-to-head is in [Arlo vs Hark Pro](/compare/arlo-vs-hark). ## How to choose - **Personal errands in an app, free:** Hark Pro, or Muse if you're in the US or Canada. - **Personal errands by text, free:** Pally. - **Text or call, if you have an invite:** Instinct. - **Personal texting:** Poke. - **Work that involves other people,** such as a company inbox, a team's Slack, or calls to customers and vendors, with a record of what was done and who approved it: Arlo. Details on this page reflect each product's public pages and dated reporting as of October 2026, apart from Poke's. Hark Pro launched on 6 October 2026, so check hark.com for its current pricing and channels. ## Frequently asked questions **What are the best Hark Pro alternatives?** Arlo for work: you text it on iMessage, SMS or WhatsApp, use it in Slack, and it places calls. Pally is free and you text it over iMessage, RCS or Telegram. Muse is Meta's agent, in the US and Canada. Instinct is invite-only, and Poke is for personal texting. **Is Hark Pro free?** Yes. As of October 2026 every Hark Pro feature is free with a usage cap that Hark hasn't published. Hark Pro² gives twice the usage for $20 a month on the web, and Hark Pro³ gives ten times for $100 a month. **How much does Hark Pro cost?** On the web, $0, $20 or $100 a month, per Hark's launch post. Its pricing page returned a 404 on 7 October 2026. The apps charge more: the US App Store listed Pro² at $26.00, Pro³ at $129.99 and an unannounced Pro⁴ at $200.00. **Do the first 100,000 sign-ups get Hark Pro free?** Brett Adcock posted on X on 4 October 2026 that the first 100,000 sign-ups would get a paid plan for free. Hark's launch post doesn't mention it, and which plan it covers and for how long hadn't been confirmed as of October 2026. **Is Hark Pro safe?** It asks before it spends money or sends messages by default, though you can give it standing permission for a category of purchases. Its terms make you responsible if a site suspends your account for automation. Its privacy policy says it doesn't train on your data, and passwords and cards sit in an encrypted vault. **Is Hark Pro any good?** Launch reviews praised its design and proactive home screen (Gizmodo, Fast Company). Gizmodo was less sure the cards are new, comparing them with what Gemini, Siri and Poppy already do, and TechCrunch and Gizmodo both noted how much of your digital life it asks for. It's too new for user ratings to tell you much yet. **Who makes Hark Pro?** Hark Labs, Inc. of San Jose, the AI lab founded by Brett Adcock, who also founded Archer Aviation and Figure AI. The Hark Audio app and the Hark app for Shopify stores are separate products. **Can I text Hark Pro, or have it make phone calls?** As of October 2026, Hark hasn't announced either. Hark Pro is an iPhone, Android and web app. Arlo and Pally can both be texted and place calls. Instinct can be texted and calls you, with calls to businesses in early access. **Is there a Hark Pro alternative for work?** Arlo. Hark Pro is sold to one person for their own life, with no team plan or admin controls that we could find. Arlo works inside a team's Slack as well as iMessage, SMS and WhatsApp, connects to the tools a company shares, holds anything it sends or changes for approval unless you've pre-approved that kind of action, and keeps an audit trail. --- # Instinct AI alternatives you can use today — no invite needed > The best Instinct AI alternatives you can use today, with no invite, are Arlo (for work over iMessage, SMS, WhatsApp, Slack and calls, with approvals), Pally (personal, free with every feature), Muse (Meta's free agent, US and Canada), Poke, Martin and OpenClaw (open source). As of October 2026 Instinct is invite-only, with five invites per member. Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/instinct-alternatives Instinct AI, the AI assistant from Spear Street Technology, went viral in late August 2026: an assistant you text or call that uses a phone and a computer the way a person would. On 28 September 2026 it raised a $1 billion Series C at a $10 billion valuation from Sequoia, Benchmark and Coatue ([TechCrunch](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/)). Getting in is still the hard part. As of October 2026 Instinct is in invite-only early access: each member gets five invites to share ([Fortune](https://fortune.com/2026/09/30/noah-shinn-instinct-ai-assistant-meta-muse-alexandr-wang-tech-series-c-ai-agent-mark-zuckerberg/), 30 September 2026), and everyone else joins a waitlist. If you're on that waitlist, or you got in and want to compare, these are the assistants you can start with today. This page is written by Arlo, so weigh it accordingly. We've kept every claim about Instinct to its own site, its terms, and dated reporting, and we say where it is the better pick. ## What Instinct is (as of October 2026) - **Who makes it:** Spear Street Technology, Inc., in San Francisco, founded by Noah Shinn. - **How you reach it:** you text it on iMessage, SMS or WhatsApp, or call it, and it can text or call you first. Each Instinct has its own email address. Group chats, where friends without an account can join in, have been in early access since 5 October 2026 ([TechCrunch](https://techcrunch.com/2026/10/05/instinct-brings-its-ai-agent-to-group-chats-even-for-friends-without-an-account/)), and so are the calls it places to businesses for you. - **How it works:** it runs a computer in the cloud that keeps the logins you give it and carries on working between your messages, so multi-step errands keep going without you. It has no mobile app. - **Who it's for:** personal life. It coordinates with friends' Instincts and in group chats, but has no team workspace, admin controls or business plan, and its terms permit use "for your personal use only". - **Shopping:** purchases go through Stripe Link, which creates a one-time card after you approve. Since 28 September 2026 it can also set up orders from Shopify stores for you to check out with Shop Pay. - **Price:** free, with no published pricing; its terms allow for paid features later. - **Access:** invite-only. Each member gets five invites to share; everyone else joins the waitlist. ## The alternatives at a glance | Assistant | Can you start today? | Where you talk to it | Built for | Phone calls | |---|---|---|---|---| | **Arlo** | Yes — free to start | iMessage, SMS, WhatsApp, Slack | Work — founders, operators, teams | Places calls, through phone menus and hold | | **Pally** | Yes — free, with a weekly usage limit | iMessage, Android RCS, SMS, Telegram | Personal life | Included, under the weekly limit | | **Muse** | US and Canada — free, with a weekly usage limit | Its own app (iPhone, iPad, Android, web, Mac), WhatsApp | Personal life and small business | US businesses, in beta | | **Poke** | Yes | iMessage, SMS, Telegram; WhatsApp in some markets | Personal life | — | | **Martin** | Yes — free trial | SMS, WhatsApp, email, Slack, iOS app | Busy individuals | Yes | | **OpenClaw** | Yes — you host it | WhatsApp, Telegram, Slack, iMessage and more | Technical users | Via a voice plugin | | Instinct | Invite only | iMessage, SMS, WhatsApp; group chats in early access | Personal life | You call it, it calls you; businesses in early access | ## Arlo — the alternative when the work isn't only yours Arlo is an [AI colleague](/glossary/ai-colleague) you text on [iMessage](/imessage), SMS or [WhatsApp](/whatsapp), use inside [Slack](/slack), or hand a [phone call](/phone) to. It places calls to people and businesses, and can work through phone menus and wait on hold. It connects to Gmail, your calendar, Slack, Notion, Linear, HubSpot and 3,000+ other tools, and reaches software with no API through a secure browser session you sign into once. Like Instinct, it goes and does the errand. Where it differs is how it handles acting on your behalf. Anything Arlo sends or changes waits for your approval unless you've pre-approved that kind of action, and every tool call, source and approval is written to an [audit trail](/glossary/ai-agent-audit-trail) you can read back. It's built for work that other people depend on, and anyone can start today: it's free to start, with no card, then runs on monthly plans from $20. Head-to-head: [Arlo vs Instinct AI](/compare/arlo-vs-instinct). ## Pally — free, with every feature Pally (pally.com), a Y Combinator company, is the closest texting match to Instinct's personal-life pitch. You chat with it over iMessage, Android RCS, SMS and Telegram. It reads and triages your WhatsApp inbox, but WhatsApp isn't where you talk to it. It books flights, hotels and rental cars, and pays on a single-use card capped at the amount you approve (US only for now) or with your saved card after one approval of the exact total. Its calls work through phone menus and wait on hold. Since 6 October 2026 Pally is free for everyone with every feature, under one usage limit that refills through the week. Its Pro and Max subscriptions are gone ([Pally's terms](https://pally.com/terms)). Calls are included under that weekly limit, and every account now gets its own phone number and email address. Pally shows you what it proposes word for word, though things you explicitly ask for can go out without a second confirmation. More: [Pally alternatives](/compare/pally-alternatives) and [Arlo vs Pally](/compare/arlo-vs-pally). ## Muse — Meta's free agent, in the US and Canada Muse, Meta's AI agent, launched on 8 September 2026 ([Meta](https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/)) and is available to adults in the US and Canada. You use it in its own app on iPhone, iPad, Android, the web or the Mac, or connect it to WhatsApp and text it there. It runs on its own cloud computer with a browser you can watch and take over, books travel and dinners, pays with one-time Link by Stripe cards, and is testing calls to US businesses. It's free with a weekly usage limit, and Power ($20 a month) and Max ($100 a month) raise that limit ([Meta Help Centre](https://www.meta.com/help/subscriptions/1021145227643680/)). By default it asks before some actions, and Engadget suggests switching it to always ask; it also uses your data for training unless you opt out ([Engadget](https://www.engadget.com/2256577/how-to-get-started-with-meta-s-new-ai-agent-muse/), September 2026). The two agents are compared head to head in [Instinct vs Muse](/compare/instinct-vs-muse), and the wider field is in [Muse alternatives](/compare/muse-alternatives). ## Poke — the personal texting pick Poke lives in iMessage, SMS and Telegram, with WhatsApp in some markets, connects to your email, calendar and 30-plus personal apps, and runs life admin through one-tap "Recipes." Cognition, the lab behind the Devin coding agent, acquired its maker in July 2026, and Poke keeps running. If what drew you to Instinct was texting an assistant about your own life, Poke is the most polished way to do that today. See [Arlo vs Poke](/compare/arlo-vs-poke). ## Martin — texts, WhatsApp and calls Martin pitches itself as a personal AI like Jarvis. You reach it over SMS, WhatsApp, email and Slack or through its iOS app; it manages your calendar and inbox, sets reminders, and makes calls for you. It's a subscription with a free trial. Of the open-access options it's closest to Instinct's text-and-call shape for one person. See [Arlo vs Martin](/compare/arlo-vs-martin). ## OpenClaw — if you'd rather run it yourself OpenClaw is an open-source, self-hosted assistant: you run it on your own machine or server and connect it to the messaging apps you already use — WhatsApp, Telegram, Slack, Signal, iMessage and more. It can browse the web, fill forms and run tasks, and your data stays on your hardware. It's free, but you set it up, maintain it, and decide how much access to give it. For developers who want Instinct-style autonomy without a waitlist or a vendor holding their logins, it's the natural pick. ## Is Instinct safe to use? It depends on what you connect. Instinct's appeal is that it acts on its own. Like most AI products' terms, Instinct's disclaim guarantees, and they are candid about what acting for you means. They appoint Instinct as your agent and say commitments it makes "shall be binding on you as if entered into directly by you". They say Instinct "may implement safeguards, confirmation requirements, or other controls" but make "no representation or warranty that such safeguards will prevent unintended or erroneous Actions", and that "Records of Actions available through the Services may not always be accurate". It may train on your materials unless you opt out in its settings, and anything you keep in its Vault is never used for training. What users have reported so far, by date: - **24 August 2026.** Early users' experiences, collected by [TechCrunch](https://techcrunch.com/2026/08/24/instincts-powerful-ai-assistant-is-raising-privacy-and-security-concerns/): one user who otherwise praised it posted that it had sent an innocuous email "without checking with me first". Another found her emails still stored after she disconnected Gmail; Instinct has since added a way to delete them. One founder tested how it handled a planted phishing email and deleted his account afterwards. Instinct told the Wall Street Journal it was taking the concerns seriously. - **23 September 2026.** A user posted what looked like a stranger's financial document. Shinn said it was a hallucination: "This was not a data breach, no user data was shared." He announced a hallucination-detection layer the same day ([Shinn on X](https://x.com/noahrshinn/status/2102896837522804954); [Business Insider via Yahoo](https://tech.yahoo.com/ai/deals/articles/instincts-founder-says-viral-ai-063054528.html), 24 September 2026). - **24 September 2026.** Business Insider reported that Instinct read a Luma one-time login code from a user's Gmail without asking, used it to cancel RSVPs, and then misdescribed what it had done ([summary via Yahoo](https://tech.yahoo.com/ai/meta-ai/articles/hallucinations-2fa-prompts-hidden-access-172455494.html), 25 September 2026). - **30 September 2026.** After Instinct launched Selections, curated product recommendations sent to users, some objected publicly to the unsolicited suggestions ([TechCrunch](https://techcrunch.com/2026/09/30/instincts-new-product-recommendations-are-giving-some-users-the-ick/)). The same three questions apply to Instinct and to every alternative on this page: 1. **Does it ask before it sends or buys?** And is that a setting, or a promise? 2. **Is there a record you can trust** of what it did, what it sent, and who approved it? 3. **What happens to your data** when you disconnect an account or leave? ## How to choose - **You want an assistant for your own life and don't mind a wait:** Instinct, if you can get an invite. - **You want that today:** Pally for breadth at no cost, Muse if you're in the US or Canada and want Meta's free agent, Poke for polish, Martin for texts and calls. - **You'd rather open an app than text:** Hark Pro, from Brett Adcock's AI lab Hark, launched free on 6 October 2026 for iOS, Android and the web ([Hark](https://hark.com/articles/introducing-hark-pro)); see [Hark Pro alternatives](/compare/hark-alternatives). - **You want to own the whole thing:** OpenClaw. - **The work involves other people** — your company's inbox, your team's Slack, customers on the phone — and you need to see what was done and who approved it: Arlo. Instinct, Pally, Muse and Hark Pro details on this page reflect their public sites and dated reporting as of October 2026; Poke, Martin and OpenClaw were last checked in September 2026. Instinct in particular is changing week to week, so check its own site for the current state. ## Frequently asked questions **What are the best Instinct AI alternatives?** The best alternatives you can use without an invite are Arlo for work across iMessage, SMS, WhatsApp, Slack and phone calls; Pally for personal life, free with every feature; Muse, Meta's free agent in the US and Canada; Poke for personal texting; Martin for SMS, WhatsApp and calls; and OpenClaw if you want to host an open-source assistant yourself. **Which Instinct AI alternatives need no invite?** All of them. Arlo, Pally, Poke, Martin and OpenClaw are open to anyone today, and Muse is open to adults in the US and Canada. Only Instinct needs an invite or a spot off its waitlist. **How do I get an Instinct invite?** As of October 2026 Instinct is invite-only. Each member gets five invites to share, and everyone else joins the waitlist. There is no public sign-up. **Is Instinct AI free, and what does it cost?** It's free for now. As of October 2026 Instinct has no published price, plans or limits, and it hadn't announced a price for its Concierge calling service as of 3 October ([Fox News](https://www.foxnews.com/tech/ai-agents-make-phone-calls)); its terms allow for paid features later. Pally and Muse are also free, each with a weekly usage limit. Arlo is free to start too, with credits and no card, then runs on monthly plans from $20. **Is Instinct AI safe?** It can act for you with real authority. Its terms say commitments it makes are binding on you and, like most AI products', don't guarantee its safeguards prevent mistakes. Users have reported it sending an email without checking first (August 2026) and reading a login code from Gmail without asking (September 2026), and it trains on your materials unless you opt out. Whichever assistant you choose, check whether it asks before sending, keeps a record you can trust, and deletes your data when you disconnect. **Is Instinct AI legit?** Yes. It's made by Spear Street Technology, founded by Noah Shinn, and on 28 September 2026 it raised $1 billion at a $10 billion valuation from Sequoia, Benchmark and Coatue ([TechCrunch](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/)). It's also early: invite-only, unpriced, and changing week to week. **Is there an Instinct alternative for work?** Yes. Instinct is built for personal life: it coordinates with friends' Instincts and in group chats, but has no team workspace, admin controls or business plan, and its terms license it "for your personal use only". Arlo is built for work: it runs inside Slack as well as iMessage, SMS and WhatsApp, connects to a company's shared tools, holds what it sends for your approval unless you've pre-approved that kind of action, and keeps an audit trail of every action and approval. **Does Instinct AI work on WhatsApp?** Yes. Once you're in, you can text Instinct on WhatsApp as well as iMessage and SMS. Among the alternatives, Arlo and Martin work over WhatsApp, Muse can be connected to WhatsApp in the US and Canada, and Poke does in some markets. OpenClaw can be connected to WhatsApp if you host it yourself. Pally reads your WhatsApp inbox, but you chat with it elsewhere. **How does Instinct compare with Meta's Muse?** Both are personal agents that run their own cloud computer, pay with one-time Stripe Link cards, use Shop Pay for Shopify orders, and are testing calls to businesses. Muse is open now in the US and Canada through its own apps and WhatsApp, free with a weekly limit; Instinct is invite-only, free, and lives in your texts. Full comparison: [Instinct vs Muse](/compare/instinct-vs-muse). --- # Instinct vs Meta Muse — how the two personal AI agents compare > Instinct and Muse are personal AI agents that run errands on their own cloud computers and pay with one-time Link cards. Instinct, from Spear Street Technology, is one you text or call; it is invite-only and free. Muse, Meta's AI agent, lives in its own apps and WhatsApp, is open to adults in the US and Canada, and is free up to a weekly limit, with paid plans at $20 and $100 a month. Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/instinct-vs-muse Instinct, the AI assistant from Spear Street Technology, and Muse, Meta's AI agent, are the two personal agents people search for side by side this autumn. Both run errands on a cloud computer of their own, both pay with one-time card numbers from Stripe's Link, both check out on Shopify stores with Shop Pay, and both started calling businesses in the same week of September 2026 ([TechCrunch](https://techcrunch.com/2026/09/17/rival-ai-agents-instinct-and-metas-muse-both-add-the-ability-to-make-calls/)). They differ on who can get in, where you reach them, and how often they stop to ask. TechCrunch counts [OpenAI's Dots](/compare/chatgpt-dots-alternatives) and Hark Pro, from Brett Adcock's AI lab Hark, in the same field ([TechCrunch](https://techcrunch.com/2026/10/06/hark-releases-an-ai-personal-assistant-with-a-focus-on-privacy/)). Hark Pro launched on 6 October; we cover it in [Hark Pro alternatives](/compare/hark-alternatives). This comparison is written by Arlo, which makes a third assistant, built for work; we cover it briefly near the end. The facts about Instinct and Muse come from their own sites, terms and help pages, and from dated reporting, as of October 2026. ## What Instinct is (as of October 2026) - **Made by:** Spear Street Technology, Inc., which does business as Instinct, in San Francisco. Its founder and CEO is Noah Shinn. - **Funding:** a $1 billion Series C at a $10 billion valuation on 28 September 2026, from Sequoia, Benchmark and Coatue ([TechCrunch](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/)). - **Access:** invite-only early access. Each member gets five invites, and everyone else joins a waitlist ([Fortune](https://fortune.com/2026/09/30/noah-shinn-instinct-ai-assistant-meta-muse-alexandr-wang-tech-series-c-ai-agent-mark-zuckerberg/), 30 September). - **Where you reach it:** you text it on iMessage, SMS or WhatsApp, or call it, and it can call you. It has its own phone number and its own email address. Group chats arrived in early access on 5 October ([TechCrunch](https://techcrunch.com/2026/10/05/instinct-brings-its-ai-agent-to-group-chats-even-for-friends-without-an-account/)). There's no mobile app. - **How it works:** a cloud computer of its own, connected to your email, calendar and other accounts as you allow. - **Price:** free, with no published pricing. Its terms allow fees later. - **Use:** its terms license it "for your personal use only" ([terms](https://instinct.com/terms)). ## What Muse is (as of October 2026) - **Made by:** Meta. Meta Superintelligence Labs built it, and it launched in the US on 8 September 2026 ([Meta](https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/)) and in Canada on 18 September. It's a separate product from Meta AI. - **Scale:** the top free iPhone app in the US since 18 September. [TIME](https://time.com/article/2026/10/06/meta-muse-ai-agent-privacy/) reported 5 million downloads and 4 million active users on 6 October. - **Access:** anyone 18 or over in the US and Canada, with no invite ([muse.ai/business](https://muse.ai/business)). - **Where you use it:** apps for iPhone, iPad, Android phones and tablets; the web at muse.ai; a Mac app that can use your computer; and WhatsApp, once you connect it. - **How it works:** Muse Secure VM, a cloud computer with a browser you can watch and step into. - **Price:** free up to a weekly limit. Power costs $20 a month and Max $100 a month ([Meta Help Centre](https://www.meta.com/help/subscriptions/1021145227643680/)). - **Use:** personal life, and since 29 September small businesses ([Meta](https://about.fb.com/news/2026/09/introducing-muse-small-business)). Plans attach to one Meta account. ## Instinct vs Muse, category by category ### Access and invites Muse is easier to get. Anyone 18 or over in the US and Canada can download it today. Instinct is invite-only: you need one of a member's five invites, or a place on the waitlist. Muse isn't available outside the US and Canada, and Meta has given no timing for other countries. Instinct hasn't published a list of countries. ### Where you use it Instinct lives in your messages: iMessage, SMS and WhatsApp, plus phone calls in both directions and its own email address you can forward or CC. It has no app. Muse is the reverse: apps for phone, tablet and Mac, a web app, and WhatsApp once you connect it, but no iMessage, SMS or phone number. On the Mac, Muse can use your computer, and you can watch its cloud browser and take over. ### Approvals Both ask before spending. Instinct's purchases through Link need your approval, you check out Shopify orders yourself with Shop Pay, and in group chats it asks before sharing anything. Beyond that, it acts and checks in when it needs your input. Its terms appoint it "as your agent", make the agreements it enters binding on you, and give "no representation or warranty" that its safeguards will prevent unintended actions ([terms](https://instinct.com/terms)). Muse asks before actions such as sending messages, making purchases or sharing information with a connected app, and you can allow once, always allow or deny ([Meta](https://ai.meta.com/muse/)). [Engadget](https://www.engadget.com/2256577/how-to-get-started-with-meta-s-new-ai-agent-muse/) found the default asks for only some actions and suggests switching to always ask. Meta's Sentinel system lets read-only and low-risk actions go ahead without asking. Muse keeps an audit trail. Instinct's terms say its records of actions "may not always be accurate". ### Payments and commerce Both use one-time card numbers from Link and Shop Pay on Shopify stores. Muse reaches further into retail: it can pay with a card you've saved on a shop's site, eligible Link purchases get purchase protection ([Meta Help Centre](https://www.meta.com/help/artificial-intelligence/1436362127544482/)), and it added connectors for Walmart, Best Buy, Sephora, Gap, Wayfair and Expedia on 23 September ([TechCrunch](https://techcrunch.com/2026/09/23/everything-new-coming-to-metas-ai-agent-muse/)). Amazon has blocked Muse since 20 September ([The Paypers](https://thepaypers.com/payments/news/amazon-blocks-metas-muse-ai-agent-from-shopping-on-its-platform)). Mark Zuckerberg said at Meta Connect on 23 September that Meta expects to make money from a small fee on transactions. Instinct's strength is travel and high-touch errands. Shinn says close to $1 billion a year in transactions runs through it, about half of it travel ([Fortune](https://fortune.com/2026/09/30/noah-shinn-instinct-ai-assistant-meta-muse-alexandr-wang-tech-series-c-ai-agent-mark-zuckerberg/)), and its Concierge service handles high-end bookings and bill disputes. In late September it began sending curated product suggestions, called Instinct Selections, which some users objected to ([TechCrunch](https://techcrunch.com/2026/09/30/instincts-new-product-recommendations-are-giving-some-users-the-ick/)). ### Calls You can call Instinct, and it can call you. Its Concierge calls to businesses have been rolling out to early-access users since 16 September, and Fox News reported on 3 October that they were still limited to some of those users ([Fox News](https://www.foxnews.com/tech/ai-agents-make-phone-calls)). Muse calls US businesses in a beta that expanded on 16 September. Muse has no phone number you can call. Reuters reported that Meta had tested having contractors place some Muse calls for its employees; a Meta VP said starting without proper disclosures "was a miss", and the test was rolled back ([BNN Bloomberg/Reuters](https://www.bnnbloomberg.ca/business/artificial-intelligence/2026/09/22/meta-testing-a-human-concierge-for-its-new-personal-ai-agent-muse-reuters-exclusive/)). ### Privacy and training Both train on your data by default, and both let you opt out. Instinct exempts anything you put in its Vault and your Google Workspace data, but the opt-out doesn't cover material flagged for safety review. Disconnecting an account doesn't delete what Instinct has indexed; deletion is a separate request ([privacy policy](https://instinct.com/privacy-policy)). Meta keeps Muse conversations out of its ad systems, but its [security blog](https://research.meta.ai/blog/security-and-safety-for-ai-agents-our-approach-with-muse) says what Muse does on the web shows up to sites as your own activity and may influence ads indirectly. TIME reported that Muse keeps continuously updated profiles of the people in your life, including people who don't use Muse. Meta didn't dispute it, and said each user's data stays isolated ([TIME](https://time.com/article/2026/10/06/meta-muse-ai-agent-privacy/)). ### Incidents Instinct: - An early user said it sent an email "without checking with me first", and another found her emails still stored after she disconnected Gmail; Instinct later added a way to delete them ([TechCrunch](https://techcrunch.com/2026/08/24/instincts-powerful-ai-assistant-is-raising-privacy-and-security-concerns/), 24 August). - Business Insider reported on 24 September that it read a one-time login code from a user's Gmail without asking, used it to cancel RSVPs, and then misdescribed what it had done ([summary via Yahoo](https://tech.yahoo.com/ai/meta-ai/articles/hallucinations-2fa-prompts-hidden-access-172455494.html)). - After a user was shown what looked like a stranger's financial document, Shinn said it was a hallucination and not a data breach, and announced a layer to detect hallucinations ([Shinn on X](https://x.com/noahrshinn/status/2102896837522804954), 23 September). Muse: - A security researcher found a flaw in the Mac app, and Meta patched it the same day ([9to5Mac](https://9to5mac.com/2026/09/22/security-bite-the-last-24-hours-at-meta-were-not-a-musing/), 22 September). - The Information and Reuters reported a flaw that could expose a user's cloud computer, emails and files. Meta rated it SEV-2 and is adding a clearer safety warning ([Reuters via The Star](https://www.thestar.com.my/tech/tech-news/2026/09/26/meta-bolsters-muse-safety-warning-after-security-vulnerability-found-the-information-reports), 25 September). - Inc. columnist Jason Aten said Muse for Mac synced his Apple Messages without permission. Meta disputed it, saying Messages access needs both Full Disk Access and Muse's Messages connector ([TechCrunch](https://techcrunch.com/2026/09/30/meta-disputes-claim-that-muse-read-a-users-private-messages-without-permission/), 30 September). Apple then said it will add macOS controls so that this level of access needs very explicit user action ([Reuters via Capital.com](https://capital.com/en-gb/news/apple-says-it-will-flag-ai-requests-for-mac-data), 2 October). - The rolled-back human-caller test and TIME's report on contact profiles, both above. ### Price Both are free today. Instinct has no published pricing, its terms reserve the right to charge for some features later, and it hasn't announced a business model. Muse is free up to a weekly limit, with Power at $20 a month and Max at $100 a month for more usage. Muse's limits are published; Instinct's free access depends on getting in. ## Who each is best for **Pick Instinct if** you have an invite and want an agent you text like a person, on iMessage, SMS or WhatsApp, and can also call. It's strongest on travel, reservations and email follow-ups. It coordinates plans with the Instincts of friends, family and coworkers you trust, and it now joins group chats, where friends don't need an account. One user credited it with flagging a calendar conflict without being asked ([X](https://x.com/Olamitisoji/status/2099162194457874476)). **Pick Muse if** you're in the US or Canada and want to start today, with an app on your phone, tablet or Mac. It has the wider retail reach (named shops, Shop Pay and saved cards), a browser you can watch and take over, computer use on the Mac, a generous free allowance, and Muse for Small Business if you run a shop on Shopify, Instagram and Meta ads. ## A third option for work: Arlo Instinct and Muse are built around one person's life, and Muse for Small Business around one owner. If the work involves other people (a company inbox, a team's Slack, customers on the phone), Arlo is built for that. It's an [AI colleague](/glossary/ai-colleague) you text on [iMessage](/imessage), SMS or [WhatsApp](/whatsapp), use inside [Slack](/slack), or hand a [phone call](/phone) to. It places calls to people and businesses, and can work through phone menus and wait on hold. It connects to Gmail, your calendar, Slack, Notion, Linear, HubSpot and 3,000+ other tools. Anything it sends or changes waits for your approval unless you've pre-approved that kind of action, and every tool call, source and approval is written to an audit trail you can read back. Arlo's [privacy policy](/privacy) says it doesn't use your messages, files, memories, calls or connected-account content to train AI models. Arlo has no shopping checkout or flight booking of its own, so for errands and travel either Instinct or Muse does more. Anyone can start Arlo today, with no invite or waitlist: it's free to start, no card, then monthly plans from $20. See [Arlo vs Instinct](/compare/arlo-vs-instinct) and [Meta Muse alternatives](/compare/muse-alternatives). ## Instinct vs Muse vs Arlo at a glance | | Instinct | Muse | Arlo | |---|---|---|---| | Can you start today? | Invite only, waitlist | Yes, if you're 18+ in the US or Canada | Yes, no invite | | Where you use it | iMessage, SMS, WhatsApp, phone calls; group chats in early access | Apps for iPhone, iPad, Android and Mac; web; WhatsApp | iMessage, SMS, WhatsApp, Slack, phone calls | | Built for | Personal life; licensed for personal use only | Personal life and small business | Work: founders, operators, teams | | Before it acts | Asks before purchases and when it needs input | Asks before some actions by default, including messages and purchases; can be set to always ask | Waits for your approval on anything it sends or changes, unless you've pre-approved that kind of action | | Payments | Link one-time cards; Shop Pay | Link one-time cards, saved cards, Shop Pay; blocked by Amazon | No card checkout | | Phone calls | You call it, it calls you; calls businesses in early access | Calls US businesses, in beta | Places calls to people and businesses; works phone menus and waits on hold | | Record of what it did | Records its terms say "may not always be accurate" | Audit trail | Audit trail of every tool call, source and approval | | Trains on your data | Yes, unless you opt out | Yes, unless you opt out | Its privacy policy says your content isn't used to train AI models | | Price | Free; no published pricing | Free up to a weekly limit; Power $20/mo, Max $100/mo | Free to start, then plans from $20 a month | Both agents change week to week. The Instinct and Muse columns reflect their public sites, terms and dated reporting as of October 2026. ## Frequently asked questions **What is the difference between Instinct and Meta Muse?** Both are personal AI agents that run errands on their own cloud computers. Instinct is one you text on iMessage, SMS or WhatsApp, or call, and it's invite-only. Muse is Meta's agent, with apps for phone, tablet, Mac and the web plus WhatsApp, and it's open to adults in the US and Canada. Instinct leans toward travel and coordinating with other people's Instincts; Muse toward shopping, its own apps and small business. **Is Instinct or Muse free?** Both are free as of October 2026. Instinct has no published pricing, and its terms allow fees later. Muse is free up to a weekly limit, with Power at $20 a month and Max at $100 a month for more usage. **Which is safer, Instinct or Muse?** Neither has a clean record. Instinct has acted without asking (an email, a login code) and misdescribed its own actions, and its terms disclaim its safeguards. Muse has had a SEV-2 flaw, a same-day Mac patch, a disputed Messages report and a rolled-back human-caller test, and TIME reported that it profiles users' contacts. Both train on your data unless you opt out. Whichever you use, check what it asks about before acting, and connect only the accounts it needs. **Do I need an invite for Instinct or Muse?** For Instinct, yes: each member gets five invites, and everyone else joins a waitlist. Muse needs no invite if you're 18 or over in the US or Canada. **Which one can make phone calls?** Both call businesses in a limited rollout: Instinct's Concierge in early access, and Muse for US businesses in beta. Only Instinct has a number you can call, and it can call you. **Which works on iMessage or WhatsApp?** Instinct works on iMessage, SMS and WhatsApp. Muse works in WhatsApp once you connect it, and isn't on iMessage or SMS. **Can I use Instinct or Muse for work?** Instinct's terms license it for personal use only. Muse for Small Business connects to Slack, Shopify and QuickBooks, but plans attach to one Meta account, with no team seats or admin roles. Arlo is built for work and runs inside a team's Slack as well as iMessage, SMS and WhatsApp. More options are in [Instinct alternatives](/compare/instinct-alternatives). --- # Best Lindy Alternatives for Teams in 2026 > The best Lindy alternatives are Zapier Agents and Poke, but one is a workflow builder and the other a consumer chat app. Arlo is the strongest pick for teams: an AI colleague in iMessage, Slack, WhatsApp, and calls, with governance and a full audit trace on every action. Updated: 2026-06-19 · Canonical: https://www.arlo.sh/compare/lindy-alternatives Most people land here after trying to build something in Lindy and realizing they're now maintaining workflows for a living. The two names you'll hear most as alternatives are Zapier Agents and Poke, and they pull in opposite directions: one is even more of a builder, the other is a consumer text app for one person. If what you actually want is an [AI colleague](/glossary/ai-colleague) that lives in the tools your team already uses and answers to a policy, [Arlo](/) is the one worth looking at. Here's the honest rundown first. ## What Lindy is, in brief As of June 2026, Lindy is a no-code AI agent builder for non-technical teams. You assemble multi-step agents from blocks on a drag-and-drop canvas, with multi-agent collaboration, a big template library, thousands of integrations, and a "Computer Use" feature for sites without an API. It's a solid platform if you genuinely want to design and own automation flows. The catch is right there in the design. Lindy makes you think in triggers and blocks. If you'd rather just message an assistant and have it work out the steps, you'll feel the friction fast. ## The main Lindy alternatives The three that come up most, kept neutral: - **Zapier Agents** — As of June 2026, Zapier layered AI Agents on top of its automation platform, with 7,000+ integrations, Copilot for turning plain English into Zaps, guardrails, memory, and human-in-the-loop approvals. Great if your team already lives in Zaps. But it's builder-first, same as Lindy. - **Poke** (by The Interaction Company) — As of June 2026, Poke is a consumer assistant you reach over iMessage, SMS, and Telegram for personal tasks like planning and managing your calendar. It's text-native, but it's built for individuals, not governed team workspaces. - **Martin** — As of June 2026, Martin pitches itself as a personal "AI chief of staff" that runs your email, calendar, and tasks over text and voice, with connections like Gmail and Slack. Also individual-focused, with a consumer subscription. Want a closer head-to-head? See [Arlo vs Lindy](/compare/arlo-vs-lindy). If the consumer-assistant route is what you're weighing, [Poke alternatives](/compare/poke-alternatives) and [Arlo vs Poke](/compare/arlo-vs-poke) go deeper. ## How the options compare | | Lindy | Zapier Agents | Poke | Arlo | |---|---|---|---|---| | Primary model | Visual workflow builder | Visual workflow builder | Text chat assistant | Conversational AI colleague | | Built for | Non-technical builders | Automation power users | Individuals | Teams | | Surfaces | Web app | Web app | iMessage, SMS, Telegram | iMessage/SMS, Slack, WhatsApp, phone calls | | Tools without an API | Computer Use | Limited | Limited | Secure browser session (logs in, clicks, fills forms) | | Governance / approvals | Available | Human-in-the-loop | Not team-oriented | Policy before every run; reviewer approves writes | | Audit trace | Varies | Audit logging | Not team-oriented | Full trace of every call, source, approval | Facts about Lindy, Zapier, Poke, and Martin are as of June 2026 and may change. Check each vendor for current details. ## Why teams choose Arlo Arlo isn't a canvas. You don't wire up flows; you just talk to it. It's a single AI colleague that works where your team already works: iMessage and SMS, Slack, WhatsApp, and live phone calls. Nothing new to download, no separate app to teach everyone. What makes it different for a team: - **You ask, it figures out the steps.** Plain language in, no blocks or triggers to babysit. - **It remembers.** Persistent memory plus a morning briefing of what changed, so context carries across days and across surfaces. - **It can reach almost anything.** Slack, Gmail, Notion, Linear, GitHub, and 3,000+ tools. For anything without an API, it logs in once through a secure browser session and clicks, reads, and fills forms the way a person would. Then there's governance, which is on by default rather than something you remember to turn on. Every connection runs through policy before a run starts. Drafts, sends, and tool writes hold for a reviewer you pick. And every call, source, and approval lands in an audit trace you can actually read back. That's the part a build-it-yourself platform or a consumer text app simply doesn't give a team. ## Which alternative fits you - Pick a **builder** (Lindy, Zapier) if you want to design and own the automation yourself. - Pick a **consumer assistant** (Poke, Martin) if you want personal help over text for one person. - Pick **Arlo** if a whole team needs a governed AI colleague across messaging and calls. See how it fits [founders](/use-cases/ai-assistant-for-founders), [realtors](/use-cases/ai-assistant-for-realtors), and [sales teams](/use-cases/ai-assistant-for-sales), or browse the [resources](/resources). ## Frequently asked questions **What are the best Lindy alternatives?** The two names you'll hear most are Zapier Agents and Poke, and they pull in opposite directions — one is even more of a builder, the other is a consumer text app for one person. Arlo is the third option: an [AI colleague](/glossary/ai-colleague) you message across iMessage and SMS, Slack, WhatsApp, and calls, with policy and a full audit trace. **What is the best Lindy alternative for a team that doesn't want to build workflows?** Arlo. Most people arrive here after realizing they're now maintaining workflows for a living. Arlo has no builder to open — you message it like a teammate and it reasons out the steps per request, under rules you set. **Is there a Lindy alternative that's cheaper?** It depends on how much you use. As of September 2026 Lindy's published per-user plans run roughly $29.99–$199.99/month with a 7-day trial. Arlo is free to start, then runs on monthly plans from $20 at 400 credits per $1, spent only on work it actually does, and whatever is left stays good for 12 months. **Can a Lindy alternative use software with no API?** Yes. Arlo logs in once through a secure browser session and clicks, reads, and fills forms the way a person would, so internal systems that never shipped an API are still in reach. Lindy addresses the same problem with computer-use / Autopilot. **Should I pick a workflow builder or a conversational assistant?** Pick a builder when the work is a stable, high-volume pipeline that runs identically every time — that's cheaper and more predictable. Pick a conversational colleague for judgment calls, ad-hoc requests, and cross-tool work no flowchart captured. See [AI assistant vs workflow automation](/compare/ai-assistant-vs-workflow-automation) for the full breakdown. **Which Lindy alternative has the strongest governance?** Arlo gates every send and tool write behind a reviewer by default and records every call, source, and approval in a [full audit trace](/glossary/ai-agent-audit-trail). Lindy offers role-based team accounts, with audit capability varying by plan. ## Try Arlo If you want an AI colleague your team can just message, across iMessage, Slack, WhatsApp, and calls, with policy and a full audit trace on every action, [get started with Arlo](/signup). --- # Martin alternatives — AI assistants for teams and personal life > The main Martin alternatives are Poke, Catch, Lindy, and Arlo. Martin, Poke, and Catch are personal AI assistants over text and voice; Lindy automates work across business apps. Arlo is an AI colleague across iMessage, Slack, WhatsApp, and phone calls with approval gates and a full audit trail. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/martin-alternatives If you're looking at Martin alternatives, start with what Martin is for. Martin is a voice-first personal assistant — you talk to it to manage your calendar, email, and reminders. Poke and Catch play a similar personal-assistant role over text. Lindy is a work-automation tool. Arlo is an AI colleague for a team — it works across every channel your team uses, makes phone calls itself, and holds every action behind an approval gate. Those are different products, so the right pick comes down to whether you're outfitting yourself or a team. Everything below is real to the best of our knowledge as of July 2026. Because we can't verify each competitor's live feature set, treat the specifics as "as of July 2026." After the rundown, I'll make the case for [Arlo](/) if you want one governed assistant spanning iMessage, Slack, WhatsApp, and live calls. ## The main Martin alternatives at a glance | Assistant | Where it works | Best for | | --- | --- | --- | | Martin | Voice, text, email | Voice-first scheduling and reminders | | Poke | iMessage, SMS, Telegram (WhatsApp in some markets) | Personal tasks over text | | Catch | Text and chat (personal assistant) | Personal tasks and daily help | | Lindy | Email, Slack, calendars, business tools | Workflow automation at work | | Arlo | iMessage/SMS, Slack, WhatsApp, live phone calls | An AI colleague across a team's tools, with governance | ## Martin Martin is the voice-first assistant. As of July 2026 it connects to your calendar, email, and messaging, and you can talk to it, text it, or email it. The pitch centers on scheduling, reminders, drafting quick messages, and daily updates — a personal assistant you speak to like you would a real one. Want the direct head-to-head? See [Arlo vs Martin](/compare/arlo-vs-martin). ## Poke Poke comes from The Interaction Company and the pitch is that using it feels like texting a friend. As of July 2026 it runs over iMessage, SMS, and Telegram, with WhatsApp in some markets. The sweet spot is everyday personal stuff — planning your day and juggling your calendar. For the direct comparison, see [Arlo vs Poke](/compare/arlo-vs-poke). ## Catch Catch is a personal AI assistant you message to manage your day and get quick help. As of July 2026 the pitch is a friendly, conversational helper for one person's tasks and reminders. If you want a personal assistant for your own to-dos over chat rather than voice, Catch is aimed at you. See [Arlo vs Catch](/compare/arlo-vs-catch) for the head-to-head. ## Lindy Lindy is built to get work done, not chat. As of July 2026 it's a workflow automator that plugs into email, Slack, your calendar, and a long list of business apps — triaging inboxes, sitting in on meetings, and chaining multi-step workflows on its own. If your problem is repetitive business processes rather than a chatty assistant, Lindy is the obvious pick here. ## Why Arlo is the best Martin alternative for teams Martin, Poke, and Catch are personal assistants — one person, their calendar, their reminders. That's a good fit for an individual. But bring an assistant into a team and the questions change: Can everyone reach it where they already work? Can you trust it to take action? Can you prove what it did afterward? Arlo is built for exactly those questions. - **It works everywhere, including the phone.** Arlo is in iMessage and SMS, Slack, WhatsApp, and live phone calls. Like Martin it handles a call, but it also lives in the team's channels — and context doesn't drop when a thread becomes a call. - **It remembers, and it briefs you.** Arlo holds onto what matters across conversations and sends a morning briefing on what changed overnight, so nobody re-explains themselves. - **It acts, and connects to real tools.** Arlo works across 3,000+ tools. When an app has no API, it signs in once through a secure browser session and operates it by hand — clicking, reading, and filling forms the way a person would. - **Governance is the default, not an add-on.** Every send and every tool write pauses for a reviewer you choose, and every action lands in a full [audit trail](/glossary/ai-agent-audit-trail). That's the line between an AI colleague you can put on real work and a personal assistant you just talk to. See how [approvals before sending](/use-cases/ai-assistant-with-approval-before-sending) work. - **It scales past one person.** Arlo runs on credits at 400 per $1, on monthly plans from $20 up to $50,000 as you scale. Admin policy and audit export are available on request for orgs that need them. ## How to choose Pick **Martin, Poke, or Catch** if you want a personal assistant for your own calendar, texts, and reminders. Pick **Lindy** if the goal is automating business workflows across apps. Pick **Arlo** if you want one [AI colleague](/glossary/ai-colleague) across iMessage, Slack, WhatsApp, and calls, with memory and governed actions your whole team can audit. If phone calls are the draw, see how Arlo works as an assistant that [makes phone calls](/use-cases/ai-executive-assistant-that-makes-phone-calls) or fits [in Slack](/use-cases/ai-assistant-in-slack). ## Frequently asked questions **What are the best Martin alternatives?** Poke, Catch, Lindy, and Arlo. Martin, Poke, and Catch are personal AI assistants over text and voice; Lindy automates work across business apps. Arlo is an [AI colleague](/glossary/ai-colleague) across iMessage and SMS, Slack, WhatsApp, and phone calls with approval gates and a full audit trail. **What is the best Martin alternative for a team?** Arlo. Martin is personal and voice-first for one user's day; Arlo is built for shared work, with 3,000+ integrations, a reviewer on every send, and a record of what happened. See the head-to-head at [Arlo vs Martin](/compare/arlo-vs-martin). **Which Martin alternative is best for voice?** Martin itself is the most voice-native of the group — it's built around talking. Arlo [places live phone calls](/use-cases/ai-executive-assistant-that-makes-phone-calls), and Catch places outbound calls, but for both, voice is one channel rather than the organizing idea. **Is there a free Martin alternative?** On this list, Poke is light-free with heavier use paid, and Lindy offers a 7-day trial on per-seat paid plans. Arlo is free to start too: connecting Slack grants your workspace $100 in credits (40,000), no card required. After that, plans start at $20 a month at 400 credits per $1 — $50 is roughly 625 messages — and anything unused stays good for 12 months rather than lapsing monthly. **Which Martin alternative connects to the most tools?** Arlo, at 3,000+ — and for software with no API it logs in once through a secure browser session and operates it by hand. As of June 2026 Martin connects calendar, Gmail, Slack, and WhatsApp; Catch lists 25+ integrations. **Why would I stay with Martin?** If you want a proactive personal assistant for your own day and you'd rather speak than type, Martin was designed for exactly that, and the governance layer in a team product would just be overhead. ## Try Arlo Arlo is the Martin alternative for teams that want one assistant everywhere they work — including the phone — with an approval gate on every send and a full audit trail. [Try Arlo](/signup) to put a colleague inside the tools you already use. --- # Meta Muse alternatives — AI assistants you can text instead > The main Meta Muse alternatives are Arlo (for work: text it on iMessage, SMS or WhatsApp, use it in Slack, and it places calls), Pally and Poke (personal assistants you text), Instinct (invite-only), Lindy (Slack-first) and OpenClaw (self-hosted). Muse is free up to a weekly limit, for adults in the US and Canada, and suits personal errands, shopping and small business. Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/muse-alternatives Muse, Meta's AI agent, launched on 8 September 2026 and has been the top free iPhone app in the US since 18 September. [TIME](https://time.com/article/2026/10/06/meta-muse-ai-agent-privacy/) reported on 6 October that it had 5 million downloads and 4 million active users. It's a personal agent that runs on its own cloud computer and does chores for you: email, forms, booking travel, lowering bills, buying things with a one-time card. Muse is a separate product from Meta AI, the assistant built into Facebook, Instagram, Messenger and WhatsApp. People go looking for alternatives for a few reasons. Some want an assistant that lives in iMessage instead of another app. Some want one for work. Some live outside the US and Canada, and some would rather not hand their accounts to Meta. This page covers each of those. This comparison is written by Arlo, which makes one of the alternatives, so keep that in mind when you read our section. The facts about Muse come from Meta's own pages and dated reporting as of October 2026. ## What Muse is (as of October 2026) - **Made by:** Meta. Meta Superintelligence Labs built it, and it launched in the US on 8 September 2026 ([Meta](https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/)) and in Canada on 18 September ([iPhone in Canada](https://www.iphoneincanada.ca/2026/09/18/metas-muse-ai-agent-is-now-available-in-canada/)). - **Who can use it:** people 18 and over in the US and Canada ([muse.ai/business](https://muse.ai/business)). Meta has given no timing for other countries. - **Where you use it:** the Muse app on iPhone, iPad, Android phones and Android tablets; the web at muse.ai; a Mac app, which can use your computer since 23 September ([TechCrunch](https://techcrunch.com/2026/09/23/everything-new-coming-to-metas-ai-agent-muse/)); and WhatsApp, once you connect it. It isn't on iMessage or SMS. - **How it works:** each user gets Muse Secure VM, a cloud computer with a browser you can watch and step into. It keeps working after you close the app. - **What it does:** email and calendar, forms, booking travel and dinners, negotiating bills, auditing and cancelling subscriptions, price tracking and shopping. For services with a public API, it builds its own connectors. - **Payments:** Link by Stripe issues a one-time card number for each purchase, or Muse pays with a card you've saved on the shop's site, and it asks before completing a purchase ([Meta Help Centre](https://www.meta.com/help/artificial-intelligence/1436362127544482/)). Shopify turned on Shop Pay checkout for Muse across Shopify stores on 21 September ([OfficeChai](https://officechai.com/ai/after-amazon-blocks-metas-muse-agent-shopify-says-itll-integrate-muse-to-help-people-shop-on-its-stores/)). Amazon has blocked Muse from Amazon.com since 20 September ([The Paypers](https://thepaypers.com/payments/news/amazon-blocks-metas-muse-ai-agent-from-shopping-on-its-platform)). - **Calls:** calling US businesses is still a beta, expanded on 16 September ([TechCrunch](https://techcrunch.com/2026/09/17/rival-ai-agents-instinct-and-metas-muse-both-add-the-ability-to-make-calls/)). A test in which contractors placed some calls was rolled back (see the safety section below). - **Small business:** Muse for Small Business, launched on 29 September, adds connectors for Slack, Shopify, QuickBooks, Canva and Meta ad accounts, on the same plans ([Meta](https://about.fb.com/news/2026/09/introducing-muse-small-business)). Plans attach to one Meta account, with no team seats or admin roles, and you can't use Muse inside Slack. - **Price:** free up to a weekly usage limit. Power costs $20 a month and Max $100 a month, for more usage ([Meta Help Centre](https://www.meta.com/help/subscriptions/1021145227643680/)). [TechCrunch](https://techcrunch.com/2026/09/08/meta-debuts-its-muse-ai-agent-will-consumers-trust-it/) reported at launch that Muse required a payment card to get started; [Engadget](https://www.engadget.com/2256577/how-to-get-started-with-meta-s-new-ai-agent-muse/) found a card optional and needed only for purchases. - **Approvals and data:** Meta says Muse asks before actions such as sending messages, making purchases or sharing information with a connected app, and it keeps an audit trail. Muse conversations aren't shared with Meta's ad systems. Training on your Muse data is on by default, and you can opt out. ## The alternatives at a glance | Assistant | Where you use it | Built for | Phone calls | To start | |---|---|---|---|---| | **Arlo** | iMessage, SMS, WhatsApp, Slack | Work | Places calls | Free to start, no card | | **Pally** | iMessage, Android RCS, SMS, Telegram | Personal life | Included, drawn from a weekly limit | Free for everyone | | **Poke** | iMessage, SMS, Telegram; WhatsApp in some markets | Personal life | — | Free for light use | | **Instinct** | iMessage, SMS, WhatsApp, calls | Personal life | Calls you; calls businesses in early access | Invite only, free | | **Lindy** | Slack first, plus iMessage and email | Teams, per seat | Joins meetings | 7-day trial | | **OpenClaw** | Messaging apps you connect it to | Technical users | Via a voice plugin | Free, self-hosted | | Muse | Muse apps (iPhone, iPad, Android, Mac), web, WhatsApp | Personal life and small business | Calls US businesses, in beta | Free up to a weekly limit | ## Arlo: if you want it in your texts, or you need it for work Arlo is an [AI colleague](/glossary/ai-colleague) you text on [iMessage](/imessage), SMS or [WhatsApp](/whatsapp), use inside [Slack](/slack), or hand a [phone call](/phone) to. It connects to Gmail, your calendar, Slack, HubSpot, Notion, Linear and 3,000+ other tools, and reaches sites with no API through a secure browser session you sign into once. Anything it sends or changes waits for your approval unless you've pre-approved that kind of action, and every tool call, source and approval goes into an audit trail you can read back. It places calls for you, to people as well as businesses, and can work through phone menus and wait on hold. Muse has approvals and an audit trail too. Where the two differ is where they live and who they're for. Arlo sits in the messaging apps you already use and is built for work that other people depend on: your company's inbox, your team's Slack, customers on the phone. Arlo is free to start with no card, then runs on monthly plans from $20. ## Pally: a personal assistant you text, now free for everyone Pally (pally.com), a Y Combinator company, is the closest thing to Muse's personal-errands pitch that you reach by text. You chat with it over iMessage, Android RCS, SMS or Telegram. It books flights, pays with a single-use card capped at the amount you approve, reads your WhatsApp inbox and sends replies there once you approve them, and places calls. Since 6 October 2026 Pally has been free for everyone, with every feature and one weekly usage limit. Its Pro and Max plans are gone, and calls are included, drawn from that limit. See [Pally alternatives](/compare/pally-alternatives). ## Poke: the polished texting option Poke lives in iMessage, SMS and Telegram, with WhatsApp in some markets, connects to your email, calendar and 30-plus personal apps, and runs errands through one-tap "Recipes". Cognition, the company behind the Devin coding agent, bought its maker in July 2026, and Poke keeps running. If you liked the idea of Muse but want it in your texts, Poke is worth trying. See [Arlo vs Poke](/compare/arlo-vs-poke). ## Instinct: if you can get an invite Instinct, the AI assistant from Spear Street Technology, is often compared with Muse. You text it on iMessage, SMS or WhatsApp, or call it, and it can call you. It raised a $1 billion Series C at a $10 billion valuation on 28 September 2026 ([TechCrunch](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/)), and it's still invite-only: each member gets five invites, and everyone else joins a waitlist ([Fortune](https://fortune.com/2026/09/30/noah-shinn-instinct-ai-assistant-meta-muse-alexandr-wang-tech-series-c-ai-agent-mark-zuckerberg/)). It's free, with no published pricing, and its calls to businesses are in early access. The head-to-head is [Instinct vs Meta Muse](/compare/instinct-vs-muse); for options you can start today, see [Instinct alternatives](/compare/instinct-alternatives). ## Lindy: if your work lives in Slack Lindy relaunched in August 2026 as an AI teammate that works in Slack first, and you can also reach it over iMessage and email. It connects to more than 1,000 tools, asks before anything with outside impact, and is sold per seat with a 7-day trial. Muse can connect to Slack, but it's a single-owner agent you use in its own app, with no team version or Slack app. Lindy and Arlo both work inside Slack. See [Arlo vs Lindy](/compare/arlo-vs-lindy). ## OpenClaw: if you'd rather nobody else held your accounts OpenClaw is open source and self-hosted. You run it on your own machine or server and connect it to WhatsApp, Telegram, Slack, Signal, iMessage and other apps. It browses, fills in forms and runs tasks, and your data stays on your hardware. It's free, but you set it up and maintain it yourself. For developers who don't want a large company holding their logins, it's the obvious choice. **Also new:** Hark Pro, from Brett Adcock's AI lab Hark, launched on 6 October 2026 as a free app for iPhone, Android and the web ([Hark](https://hark.com/articles/introducing-hark-pro)). We cover it in [Hark Pro alternatives](/compare/hark-alternatives). ## Is Muse safe to use? Muse's design takes safety seriously: an isolated cloud computer, approvals before sending or buying, one-time payment cards and an audit trail. Meta's [security blog](https://research.meta.ai/blog/security-and-safety-for-ai-agents-our-approach-with-muse) says a system called Sentinel lets read-only and low-risk actions go ahead without asking. Its first month also brought incidents worth knowing about: - **A Mac flaw, patched the same day.** A security researcher found a flaw in the Mac app, and Meta patched it quickly ([9to5Mac](https://9to5mac.com/2026/09/22/security-bite-the-last-24-hours-at-meta-were-not-a-musing/), 22 September). - **A SEV-2 flaw.** The Information and Reuters reported on 25 September that a researcher had found a flaw that could expose a user's cloud computer, emails and files. Meta rated it SEV-2 and is adding a clearer safety warning in Muse ([Reuters via The Star](https://www.thestar.com.my/tech/tech-news/2026/09/26/meta-bolsters-muse-safety-warning-after-security-vulnerability-found-the-information-reports)). - **The Messages dispute.** Inc. columnist Jason Aten said Muse for Mac synced his Apple Messages without permission. Meta disputed his account on 30 September, saying Messages access needs both Full Disk Access and Muse's Messages connector ([TechCrunch](https://techcrunch.com/2026/09/30/meta-disputes-claim-that-muse-read-a-users-private-messages-without-permission/)). On 2 October Apple said it will add macOS controls so that granting an app that level of access needs very explicit user action ([Reuters via Capital.com](https://capital.com/en-gb/news/apple-says-it-will-flag-ai-requests-for-mac-data)). - **Human callers.** Reuters reported on 22 September that Meta had tested a "human concierge" for half its employees, with contractors placing some Muse calls. A Meta VP said starting without proper disclosures "was a miss", and the test was rolled back ([BNN Bloomberg/Reuters](https://www.bnnbloomberg.ca/business/artificial-intelligence/2026/09/22/meta-testing-a-human-concierge-for-its-new-personal-ai-agent-muse-reuters-exclusive/)). - **Profiles of your contacts.** TIME reported on 6 October that Muse keeps continuously updated profiles of the people in a user's life, including people who don't use Muse. Meta didn't dispute it, and said each user's data stays isolated ([TIME](https://time.com/article/2026/10/06/meta-muse-ai-agent-privacy/)). By default Muse asks before only some actions. If you use it, Engadget suggests switching it to always ask. The same advice applies to any assistant that can act for you, Arlo included: know what it asks about, check that you can see a record of what it did, and connect only the accounts it needs. ## How to choose If the job is personal errands and shopping and you're in the US or Canada, Muse is free, capable and built for exactly that. If you run a small business on Shopify, Instagram and Meta ads, Muse for Small Business connects to all three. If you want your assistant in iMessage instead of another app, try Pally, which is now free for everyone, or Poke. If you have an Instinct invite, it's free too. If you want to run it yourself, use OpenClaw. If the work involves other people (a company inbox, a team in Slack, customers on the phone), look at Arlo, or at Lindy if your team buys per seat. ## Frequently asked questions **What are the best Meta Muse alternatives?** Arlo for work, over iMessage, SMS, WhatsApp, Slack and phone calls; Pally and Poke for personal errands by text; Instinct, if you can get an invite; Lindy for teams that work in Slack; and OpenClaw if you want to host an open-source assistant yourself. **Is Meta Muse free?** Yes, up to a weekly usage limit. As of October 2026 the paid plans are Power at $20 a month and Max at $100 a month, according to Meta's help page. TechCrunch reported at launch that Muse required a payment card to get started; Engadget found a card optional and needed only for purchases. Arlo is free to start with no card, then runs on monthly plans from $20. **Is Meta Muse safe?** It's built with safeguards: an isolated cloud computer, approvals before purchases, one-time card numbers and an audit trail. Its first month also brought a same-day Mac patch, a SEV-2 flaw that could expose a user's cloud computer and files, a disputed report that it read a columnist's Messages, a rolled-back test of human callers, and TIME's report that it keeps profiles of users' contacts. Training on your data is on by default, and you can opt out. **Is there a Muse alternative I can text in iMessage?** Yes. Arlo, Pally and Poke all work in iMessage, so you can text them like a contact, and so does Instinct if you have an invite. Muse works in its own apps, on the web, on the Mac and in WhatsApp. **Can I use Muse for work?** For a small business, partly. Muse for Small Business connects to Slack, Shopify, QuickBooks and Meta ad accounts, but plans attach to one Meta account, with no team seats or admin roles, and Muse has no Slack app. Arlo and Lindy are built for work and run inside Slack. **Is Muse available outside the US?** In Canada too. As of October 2026 Muse is available to people 18 and over in the US and Canada, and Meta has given no timing for other countries. Arlo, Pally and Poke can be used without a Meta account. **How does Muse compare with Instinct?** Both are personal agents that run errands on their own cloud computers. Muse is open to adults in the US and Canada and has its own apps. Instinct is invite-only, and you text or call it. See [Instinct vs Meta Muse](/compare/instinct-vs-muse). --- # Pally alternatives — AI assistants you can text in 2026 > As of October 2026 Pally is free for everyone, so the alternatives compete on fit. Arlo is for work: text it on iMessage, SMS or WhatsApp, use it in Slack, with approvals and an audit trail. Poke is for personal texting, Martin covers SMS, WhatsApp and calls, Tomo coaches goals and habits, alfred_ suits people who bill their time, and Instinct is personal and invite-only. Updated: 2026-10-07 · Canonical: https://www.arlo.sh/compare/pally-alternatives This page is about Pally (pally.com), the AI assistant you text — not Pallyy, the social-media scheduler. Pally is one of the best-rounded assistants you can text. You chat with it over iMessage, Android RCS and SMS, or Telegram; it books flights and hotels, checks out on a capped card, and makes calls. Since 6 October 2026 it has also been free for everyone, with every feature, so price is rarely the reason to leave. People look for alternatives because they want to text their assistant on WhatsApp, because they want it for work with a team, because they want something narrower, like a coach or an assistant built for one profession, or because Pally's privacy policy now lets it train on your data unless you opt out. This page is written by Arlo, so read it with that in mind. Facts about Pally, Instinct and Hark Pro are as of October 2026; the other products were last checked against their own sites in September 2026. We say plainly where each one beats us. ## What Pally is (as of October 2026) - **Who makes it:** Pally Technologies, a Y Combinator company (Summer 2025). - **Price:** free for everyone with every feature since 6 October 2026, with one usage limit that refills through the week. Pro and Max were discontinued and refunded, and the only paid plan, VIP, is invite-only at $1,000 a month. - **Channels:** iMessage, Android RCS and SMS, and Telegram, with group chats on iMessage and Telegram. Pally reads your WhatsApp inbox and can send replies there, but you can't chat with it on WhatsApp. - **Calls:** included for everyone, drawing on the weekly usage limit. Every account gets its own phone number and email address. - **Actions:** about two dozen named apps, plus MCP servers on request, and browser sign-in for almost any site. Pally books flights, hotels and rental cars, and pays with a capped single-use card (US only) or your saved card. - **People:** unlimited automations, free Teams for up to 10 people, and group chats. - **Approvals:** Pally shows you anything it proposes word for word; things you explicitly ask for can go without a second confirmation. - **Training:** since 6 October 2026 its [privacy policy](https://pally.com/privacy) says Pally may train its own models on your data unless you opt out. ## The alternatives at a glance | Assistant | Where you talk to it | Built for | Phone calls | |---|---|---|---| | **Arlo** | iMessage, SMS, WhatsApp, Slack | Work — founders, operators, teams | Places calls | | **Poke** | iMessage, SMS, Telegram; WhatsApp in some markets | Personal life | — | | **Martin** | SMS, WhatsApp, email, Slack, iOS app | Busy individuals | Yes | | **Tomo** | iMessage, plus a companion app | Personal goals and habits | — | | **alfred_** | iMessage, SMS, RCS | People who bill for their time | — | | **Instinct** | iMessage, SMS, WhatsApp, calls | Personal use; invite-only | In and out | | Pally | iMessage, Android RCS and SMS, Telegram | Personal life | Included (weekly usage limit) | ## Arlo — the Pally alternative for work Arlo is an [AI colleague](/glossary/ai-colleague) you text on [iMessage](/imessage), SMS or [WhatsApp](/whatsapp), use inside [Slack](/slack), or hand a [phone call](/phone) to. Both can act for you; the difference is what each is built around. Arlo connects to Gmail, your calendar, Slack, Notion, Linear, HubSpot and 3,000+ other tools, and reaches software with no API through a secure browser session you sign into once. Anything it sends or changes waits for your approval unless you've pre-approved that kind of action, and every tool call, source and approval is written to an audit trail you can read back. It works inside your team's Slack, and it places calls to people and businesses, through phone menus and hold. Arlo is free to start, with no card, then runs on plans from $20 a month. On price, Pally is the cheaper choice. If WhatsApp is where you want to talk to your assistant, Arlo does that and Pally doesn't. Full comparison: [Arlo vs Pally](/compare/arlo-vs-pally). ## Poke Poke lives in iMessage, SMS and Telegram, with WhatsApp in some markets, connects to your email, calendar and 30-plus personal apps, and runs life admin through one-tap "Recipes." Cognition, the lab behind the Devin coding agent, acquired its maker in July 2026, and Poke keeps running. If you like Pally's personal focus but want the most polished texting experience, Poke is the one to try. See [Arlo vs Poke](/compare/arlo-vs-poke). ## Martin Martin pitches itself as a personal AI like Jarvis. You reach it over SMS, WhatsApp, email and Slack or its iOS app; it manages your calendar and inbox, sets reminders and makes calls. It's a subscription with a free trial. If you want a Pally-style personal assistant you can message on WhatsApp, Martin is the closest match. See [Arlo vs Martin](/compare/arlo-vs-martin). ## Tomo Tomo is built around personal goals and habits: a companion you text in iMessage that joins group chats, texts you first, helps with your calendar and email, and keeps you accountable to what you said you'd do. It's a paid subscription. Pick it over Pally if you want accountability for your goals more than help with errands. ## alfred_ alfred_ is a texting assistant over iMessage, SMS and RCS for people who bill for their time — founders, lawyers, agents and accountants. It works across Gmail or Outlook, your calendar, tasks and notes, and asks you to approve every send, with an individual plan and a per-seat team plan. It doesn't list phone calls or WhatsApp. ## Instinct Instinct, the AI assistant from Spear Street Technology, is the viral personal assistant of 2026: you text it on iMessage, SMS or WhatsApp, or call it, and it runs a computer in the cloud to carry errands through. As of October 2026 it's invite-only with a waitlist, free during its beta, and licensed for personal use by its terms. If you can get in and the job is personal errands, it's worth trying. See [Instinct alternatives](/compare/instinct-alternatives). ## Hark Pro Hark Pro, from Brett Adcock's AI lab Hark, launched to everyone on 6 October 2026 ([Hark](https://hark.com/articles/introducing-hark-pro)). It's free, with paid tiers for more usage, and you use it through apps for iOS and Android and on the web; as of October 2026 nothing public offers a number you can text. See [Hark alternatives](/compare/hark-alternatives). ## How to choose - **Personal life, at no cost:** stay with Pally, or try Poke. - **Personal life, on WhatsApp:** Martin. - **Goals and habits rather than errands:** Tomo. - **Your inbox and calendar, if you bill for your time:** alfred_. - **Work that other people depend on** — your company's email, your team's Slack, customers on the phone — with approvals and a record of what happened: Arlo. Details reflect each product's public site on the dates above; check them for the current state. ## Frequently asked questions **What are the best Pally alternatives?** Arlo for work across iMessage, SMS, WhatsApp, Slack and phone calls; Poke for personal texting; Martin for SMS, WhatsApp and calls; Tomo as a goals and habits coach; alfred_ for people who bill for their time; and Instinct, which is personal and invite-only. **Is there a Pally alternative on WhatsApp?** Yes. As of October 2026 Pally reads your WhatsApp inbox, but you chat with it over iMessage, Android RCS and SMS, or Telegram. Arlo, Martin and Instinct all let you text the assistant itself on WhatsApp, and Poke does in some markets. **Is there a Pally alternative for work or teams?** Arlo. It works inside your team's Slack as well as your texts, connects to 3,000+ tools including the shared ones a company uses, holds what it sends or changes for your approval unless you've pre-approved that kind of action, and keeps an audit trail of every action and approval. Pally's free Teams links up to 10 people's personal Pallys, which suits coordination between individuals. **Which Pally alternative is free?** Pally itself is free for everyone with every feature as of October 2026, within one weekly usage limit, so no alternative here beats it on price. Poke can be used free for light use, Instinct is free during its invite-only beta, and Martin offers a free trial. Arlo is free to start, with no card, then runs on plans from $20 a month. **Does Pally train on your data?** Since 6 October 2026 Pally's privacy policy says it may train its own models on your data unless you opt out under Data retention in your settings. Arlo's [privacy policy](/privacy) says it doesn't use your messages, files, memories, calls or connected-account content to train AI models. **Which Pally alternatives can make phone calls?** Arlo, Martin and Instinct. Arlo places calls to people and businesses, works through phone menus and waits on hold. Instinct's calls to businesses are limited to some early-access users as of October 2026. Pally itself includes calls for everyone, drawn from its weekly usage limit. --- # Poke alternatives — AI assistants for work and personal life > The main Poke alternatives are Lindy, Martin, and Arlo. Poke and Martin are personal helpers over text and voice; Lindy automates work across business apps. Arlo is an AI colleague that works across iMessage, Slack, WhatsApp, and phone calls with governed actions and an audit trail. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/compare/poke-alternatives If you're shopping for Poke alternatives, you're really choosing between three different products that happen to share a category. Poke and Martin are personal helpers you talk to. Lindy automates work across business apps. Arlo is an AI colleague that lives across your whole team's tools and actually makes the phone call. Which one fits depends entirely on which of those jobs you're hiring for. Poke's details were checked as of August 2026; Lindy's and Martin's as of June 2026. After the rundown, I'll make the case for [Arlo](/) if you want one assistant spanning iMessage, Slack, WhatsApp, and live calls. For a wider ranked field, see the [best AI assistants you can text](/compare/best-ai-assistants-you-can-text). ## The main Poke alternatives at a glance | Assistant | Where it works | Best for | | --- | --- | --- | | Poke | iMessage, SMS, Telegram (WhatsApp in some markets) | Personal tasks over text | | Lindy | Email, Slack, calendars, business tools | Workflow automation at work | | Martin | Voice, text, email | Voice-first scheduling and reminders | | Arlo | iMessage/SMS, Slack, WhatsApp, live phone calls | An AI colleague across a team's tools, with governance | ## Poke Poke comes from The Interaction Company in Palo Alto — acquired by Cognition, the lab behind the Devin coding agent, on 23 July 2026 in a reported low-nine-figure deal. The whole pitch is that using it feels like texting a friend. As of August 2026 it runs over SMS, Telegram, and WhatsApp in some markets, and it was one of the first third-party AI agents you could reach through iMessage. The sweet spot is everyday personal stuff: planning your day, juggling your calendar, the kind of back-and-forth you'd do over text anyway. Pricing is negotiated. Light requests are free and heavier ones cost. If you're choosing for a team, the acquisition matters; we unpack it in [what the Cognition deal means](/blog/cognition-poke-what-it-means). Poke keeps running and picks up Cognition's models and infrastructure, but the stated reason for the deal was to bring Poke's conversational style into Devin — so the product's personality is now also an input to a developer-tools roadmap. That's a question to ask, not a reason to rule it out. Want the head-to-head? See [Arlo vs Poke](/compare/arlo-vs-poke). ## Lindy Lindy is built to get work done, not chat. As of June 2026 it pitches itself as a workflow automator that plugs into email, Slack, your calendar, and a long list of business apps. It'll triage an inbox, sit in on meetings and record them, and chain together multi-step workflows on its own. If your problem is repetitive business processes rather than wanting a chatty assistant in your texts, Lindy is the obvious pick here. ## Martin Martin is the voice-first one. It hooks into your calendar, email, and messaging, and you can talk to it, text it, or email it. As of June 2026 it's mostly described around scheduling, reminders, drafting quick messages, and daily updates. If what you actually want is something you can speak to like a personal assistant, that's the lane Martin was built for. ## Why Arlo is the best Poke alternative Every other assistant on this list picked one lane: personal texter, work automator, or voice scheduler. Arlo refuses to. It's meant to be one colleague for the entire workspace, and that range is why teams pick it over Poke. - **It works everywhere, including the phone.** Arlo is in iMessage and SMS, Slack, WhatsApp, and live phone calls. No new app to learn, and no dropped context when a thread turns into a call. - **It remembers, and it briefs you.** Arlo holds onto what matters across conversations and sends a morning briefing on what changed overnight, so you're not re-explaining yourself every time. - **It does the work.** Arlo connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools. When something has no API, it logs in once through a secure browser session and clicks, reads, and fills forms the way a person would. - **Governance is the default, not an add-on.** Every connection runs through policy before each turn. Drafts, sends, and writes pause for a reviewer you pick, and every tool call, source, and approval lands in an audit trace. That's the real line between Arlo and a personal text-bot. One is built for teams that need accountability. - **It scales past one person.** Arlo runs on credits at 400 per $1, on monthly plans from $20 up to $50,000. Admin policy, roles, SSO, and audit export are available on request — which is what an org wants before it lets any assistant act on its behalf. ## How to choose Pick **Poke or Martin** if you just want a personal assistant for your own texts, calendar, and reminders. Pick **Lindy** if the goal is automating business workflows across apps. Pick **Arlo** if you want one colleague across iMessage, Slack, WhatsApp, and calls, with memory and governed actions your team can audit. Comparing options for a founding team, a sales org, or a brokerage? See how Arlo fits as an [AI assistant for founders](/use-cases/ai-assistant-for-founders), [for sales](/use-cases/ai-assistant-for-sales), or [for realtors](/use-cases/ai-assistant-for-realtors). ## Frequently asked questions **What are the best Poke alternatives?** Lindy, Martin, and Arlo, and they pull in different directions. Poke and Martin are personal helpers you talk to over text and voice. Lindy automates work across business apps. Arlo is an [AI colleague](/glossary/ai-colleague) across iMessage and SMS, Slack, WhatsApp, and phone calls, with governed actions and an audit trail. **What is the best Poke alternative for a team?** Arlo, because the gap with Poke is not features but accountability. A personal assistant has no shared context between people, no approval step before something goes out in a teammate's name, and no record of what it did. Those become required the moment an assistant acts on a team's behalf. **Is there a free Poke alternative?** Poke itself is light-free with heavier use paid, and Lindy offers a 7-day trial on per-seat paid plans. Arlo is free to start too: connecting Slack grants your workspace $100 in credits (40,000), no card required. After that, plans start at $20 a month at 400 credits per $1 — $50 is roughly 625 messages — and anything unused stays good for 12 months rather than lapsing monthly. **Which Poke alternative can make phone calls?** Arlo [places live calls](/use-cases/ai-executive-assistant-that-makes-phone-calls) and keeps a transcript of each one. Martin is voice-first for spoken interaction. Most text-first assistants in this category do not place calls at all. **Why would I stay with Poke?** If the assistant serves exactly one person and the tasks are personal — your calendar, inbox, reminders, travel — Poke does that job well and the team controls in Arlo would only add friction. Switching pays off when the work becomes shared. See the head-to-head at [Arlo vs Poke](/compare/arlo-vs-poke). ## Try Arlo Arlo is the Poke alternative for teams that want one assistant everywhere they work, with governance and a full audit trail. [Try Arlo](/signup) to put a colleague inside the tools you already use. --- # Viktor alternatives — AI coworkers for teams in 2026 > The strongest Viktor alternatives in 2026 are Arlo (an AI colleague across iMessage, Slack, WhatsApp, and phone calls with approval-first governance), Lindy (build-your-own agents and workflows), Catch (a personal assistant across channels), Poke (a personal assistant in your texting apps), and ChatGPT. The right pick depends on team vs personal jobs — and the channels your work actually uses. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/compare/viktor-alternatives Viktor made the "[AI employee](/glossary/ai-employee)" category loud. It lives in Slack and Microsoft Teams, connects to thousands of tools, writes and runs code, and — after a reported $75M raise in May 2026 — it's the name most teams hear first. It's a good product. But it isn't the right shape for every team, and the reasons people go looking for an alternative are consistent: work that happens outside Slack (texts, phone calls), a want for approval-first governance rather than configurable trust, a personal rather than team-shaped job, or simple price-and-fit shopping. Here are the alternatives that actually deserve the shortlist, and the honest case for each. ## The short list | Tool | Best for | Channels | Governance | |---|---|---|---| | **Arlo** | Teams whose work spans chat, texts, and calls | iMessage/SMS, Slack, WhatsApp, phone calls | Approvals + audit trail on by default | | **Viktor** | Heavy in-Slack delegation, code execution | Slack, Teams | Configurable per-tool approvals | | **Lindy** | Building custom agents and workflows | Web app, email, calendar triggers | Per-workflow configuration | | **Catch** | One person who wants a proactive assistant | Slack, email, text, phone | Personal-scale | | **Poke** | Personal life admin in your texting apps | iMessage, WhatsApp, Telegram | Personal-scale | | **ChatGPT** | Open-ended thinking, drafts, research | Its own app | Session-scale | ## Arlo — the AI colleague that also texts and calls Arlo is the closest like-for-like replacement: an [AI coworker](/glossary/ai-coworker) your team talks to in the channels it already uses. The difference in reach is that Arlo treats **iMessage/SMS, WhatsApp, and live phone calls** as first-class surfaces alongside Slack — it can text you like a person, place real calls, and keep the transcript. It connects to 3,000+ tools, works no-API software through a secure browser session, keeps persistent memory with one morning briefing, and — the load-bearing difference — ships governance on by default: policy checks before every run, approval gates on every send and write, and a full [audit trail](/glossary/ai-agent-audit-trail) after. Arlo sells credits at 400 per $1 — the same rate at every amount, on monthly plans from $20 with top-ups up to $50,000 — and unlike most credit-based tools, unused credits roll over and last 12 months from purchase instead of resetting each month. See the full [Arlo vs Viktor breakdown](/compare/arlo-vs-viktor). ## Lindy — build your own agents Lindy approaches the problem from the automation side: you design agents ("Lindies") with triggers, steps, and tool actions, closer to a modern workflow builder than a coworker you address in chat. If your team likes designing systems and wants deterministic, repeatable flows, it's a strong fit. If you want to delegate messy, one-off work in plain language, the builder mindset can feel like overhead. Full comparison: [Arlo vs Lindy](/compare/arlo-vs-lindy). ## Catch — a personal assistant, done properly Catch is a proactive personal assistant you reach over Slack, email, text, or phone. It's aimed at one person's workload rather than a team's shared queue — closer to hiring yourself an assistant than adding a teammate to Slack. If what you actually miss about Viktor is "someone handling *my* stuff," Catch belongs on the list. See [Arlo vs Catch](/compare/arlo-vs-catch). ## Poke — personal, and honest about it Poke lives in iMessage, WhatsApp, and Telegram and runs your personal digital life — reminders, scheduling, email triage — with a recipe library for one-tap automations. It is deliberately not a team product, and it's good precisely because of that focus. If the job is personal, Poke is a strong consumer pick; if the job is team work, it isn't trying to be that. See [Poke alternatives](/compare/poke-alternatives) for that side of the map. ## ChatGPT — the default, for a different job Some teams evaluating Viktor discover what they wanted was better thinking, not delegated doing. ChatGPT remains the best general-purpose reasoning tool, with connectors and agentic features growing every quarter — but it's a destination you visit, not a coworker in your channels. The full picture: [Arlo vs ChatGPT](/compare/arlo-vs-chatgpt). ## How to choose Three questions cut the list fast. **Where does the work happen?** If it's Slack-only, Viktor and Arlo both fit; if it includes texts and calls, Arlo; if it's your own life, Poke or Catch. **Who is the AI working for?** A team queue points to Arlo or Viktor; one person points to Catch or Poke; a workflow diagram points to Lindy. **What has to be true before it touches customers?** If the answer involves approvals and an audit trail from day one, that's Arlo's home turf. Product details above reflect each company's public materials as of July 2026 — verify current state on their own sites before you commit. ## Frequently asked questions **What is the best Viktor alternative for teams?** Arlo is the closest team-shaped alternative: an AI coworker in Slack that also works over iMessage/SMS, WhatsApp, and live phone calls, with approvals and an audit trail on by default. **Is there a Viktor alternative that makes phone calls?** Arlo places live phone calls — confirming appointments, chasing vendors, asking questions — and keeps the transcript, so you can ask how it went. Phone calls aren't part of Viktor's stated product as of July 2026. **What's the cheapest way to decide between these tools?** Run the same three real tasks from your backlog through each tool and compare finished work. Most tools on this list offer some form of trial; Arlo is free to start and then runs on plans from $20 a month — $50 is roughly 625 messages — and what you don't spend on the bake-off is still there 12 months later. **Do any of these work without APIs?** Arlo can operate software that has no API by logging in once through a secure browser session and using the interface the way a person would. **Can I use more than one?** Plenty of teams do — a team-level colleague like Arlo for shared work, plus a personal assistant like Poke for individual life admin. The mistake is expecting one shape to do the other's job. --- # What Is an AI Agent? Definition, Types, and How They Work > An AI agent is software that perceives its environment, decides what to do, and takes multi-step actions to reach a goal, often semi-autonomously. Unlike a chatbot that only replies, an agent uses tools, remembers context, and does the work, ideally with human approval on risky steps. Updated: 2026-06-19 · Canonical: https://www.arlo.sh/glossary/ai-agent An AI agent is software that perceives its environment, decides what to do, and takes real actions to reach a goal, usually across several steps and with some autonomy. The difference from a chatbot comes down to one word: action. A chatbot answers your question. An agent goes and does the thing. ## What an AI agent actually does An agent runs a loop. It looks at where things stand, reasons about the goal, picks an action, takes it, checks the result, and goes again until it's done or stuck. That loop is the whole trick. It's what lets an agent book a meeting, update a CRM record, or dig an answer out of four different systems instead of just telling you how you might do that yourself. Three things separate a real agent from a fancy autocomplete: - **Tool use.** It calls APIs, searches the web, sends messages. The good ones can even log into a website with no API and click through it like a person. - **Planning.** It breaks a goal into steps and reroutes when a step fails, rather than giving up at the first error. - **Memory.** It carries context across turns and sessions, so you're not re-explaining yourself every morning. ## Chatbot vs AI assistant vs AI agent People use these words interchangeably, but they're really points on a line from "talks" to "does." | Capability | Chatbot | AI assistant | AI agent | |---|---|---|---| | Responds to questions | Yes | Yes | Yes | | Remembers past context | Rarely | Sometimes | Yes | | Uses external tools | No | Limited | Yes | | Takes multi-step actions | No | Some | Yes | | Works toward a goal autonomously | No | No | Yes | | Acts inside your real tools | No | Sometimes | Yes | An [AI assistant](/glossary/ai-personal-assistant) lands in the middle. It's helpful and remembers what you told it, but it mostly waits for the next instruction. An agent is built to take a goal and run with it. ## Levels of autonomy Autonomy isn't on/off. Think of it as a dial: - **Suggest** — the agent proposes; you do it. - **Draft** — the agent writes the email or fills the form, then waits for your yes. - **Act with review** — the agent executes, but anything risky pauses for a human. - **Act autonomously** — the agent finishes routine work end to end inside set limits. Most agents worth trusting in production sit in the draft and act-with-review bands. That's where you get speed without handing over the keys. Save full autonomy for the boring, low-stakes, hard-to-screw-up tasks. ## What makes an AI agent reliable Raw capability is the easy part now. The hard part is capability you can actually trust. Reliable agents tend to share four traits: - **Persistent memory**, so context and preferences survive between sessions. - **Guardrails and policy** that decide what the agent is even allowed to touch before it acts. - **Human approval** on sends, writes, and anything you can't take back. - **An audit trail** of every tool call, source, and approval, so you can see exactly what it did. Strip these out and an agent that "takes actions" is mostly a liability waiting to happen. Build them in and you've got something you can hand real work to. ## Risks to manage Agents fail in specific, predictable ways. They act on a wrong assumption, call the wrong tool, leak data between systems, or burn money looping on a task they can't finish. The fixes are the reliability traits above, plus tight scope. Give the agent a narrow job, make it ask before consequential steps, and keep a trace you can read after the fact. Autonomous enough to help, boxed in enough to stay safe. ## How this relates to an AI colleague An [AI colleague](/glossary/ai-colleague) is just an AI agent framed as a teammate. Same underlying tech, but it lives where your team already works, holds onto context, and runs under shared rules instead of sitting off in its own app. Next to a generic [AI executive assistant](/glossary/ai-executive-assistant), the colleague framing leans harder on durable memory and accountability than on knocking out one-off tasks. ## Arlo as a governed AI agent Arlo is an AI agent you talk to like a coworker, in iMessage, SMS, Slack, WhatsApp, and on live phone calls. It connects to Gmail, Notion, Linear, GitHub, and 3,000+ tools, and for anything without an API, it logs in once through a secure browser session and works the site like a person. It keeps persistent memory and sends you a morning briefing. Governance comes standard: every connection clears policy before a run, drafts and sends wait on a reviewer you pick, and every tool call, source, and approval lands in a full audit trace. Founders, [realtors](/use-cases/ai-assistant-for-realtors), and [sales teams](/use-cases/ai-assistant-for-sales) use it to hand off actual work, not just questions. Arlo runs on prepaid credits — 400 credits per $1 at every amount, on monthly plans from $20, and they last twelve months; admin policy, roles, SSO, and audit export are available on request. ## Frequently asked questions **What is an AI agent?** An AI agent is an AI system that pursues a goal over multiple steps — deciding what to do next, calling tools, reacting to what comes back — instead of returning a single response. The defining trait is that it acts, not just answers. **What is the difference between an AI agent and a chatbot?** A chatbot maps input to a reply. An agent plans, uses tools, and changes state in the world: it sends the email, opens the ticket, updates the record. That capability is exactly why review matters. **What is the difference between an AI agent and automation?** Traditional automation follows a fixed path you defined in advance and breaks when reality differs. An agent decides the path at runtime from the goal and what it observes, which handles messy, varying work but makes its behaviour less predictable — so consequential steps belong behind an approval. **Are AI agents safe to give tool access?** Only with scoping and review. The working pattern is [human-in-the-loop](/glossary/human-in-the-loop): policy resolves each connection before a run, consequential actions wait for a named approver, unanswered approvals expire rather than defaulting to send, and every run leaves an [audit trace](/glossary/ai-agent-audit-trail). **What can AI agents do today?** Reliably: triage and draft, research across sources, move data between tools, run recurring multi-step work, operate software through a browser when there is no API, and hold context across long-running work. Less reliably: anything needing accountability, relationships, or judgement about people. **Where do you run an AI agent?** Arlo runs where the work already happens — [iMessage and SMS](/imessage), [Slack](/use-cases/ai-assistant-in-slack), [WhatsApp](/whatsapp), and live phone calls — with one memory across all of them. ## Try Arlo Want an agent that acts, with guardrails already in place? [Start with Arlo](/signup) and put one governed AI colleague to work across the tools your team already uses. More in our [resources](/resources). --- # What Is an AI Agent Audit Trail? > An AI agent audit trail is a complete, tamper-evident record of what an AI agent did on your behalf — every tool it called, every source it read, and every approval it got — each step tied back to the request that started it. It's what lets you hand real work to an AI and still answer "what exactly happened, and who signed off?" Updated: 2026-07-04 · Canonical: https://www.arlo.sh/glossary/ai-agent-audit-trail Once an AI does more than chat — once it sends email, updates records, and takes actions in your real tools — a new question shows up: what did it actually do? An audit trail is the answer. It's the running record of every step an [AI agent](/glossary/ai-agent) took, so a run is never a black box you have to take on faith. An AI agent audit trail logs the whole chain of a task: the request that kicked it off, the tools the agent touched, the data it read, the actions it prepared, who approved them, and what finally went out. Read it back and you can reconstruct exactly what happened and why — the difference between trusting an assistant and being able to verify it. ## What a good audit trail records Not every "activity log" is an audit trail. A real one captures the full chain, not just the output: - **The request.** What was asked, by whom, and in which channel. - **Every tool call.** Which integrations the agent used, and what it did in each — read, draft, write, send. - **Sources.** The data and documents the agent read to reach its answer, so a claim can be traced to where it came from. - **Approvals.** Which actions paused for review, who signed off, and when. - **Outcomes.** What actually changed in your systems as a result. ## Why it matters An audit trail isn't paperwork for its own sake. It's what makes delegation safe: - **Trust through verification.** You can hand off real work because you can always check what the assistant did, instead of hoping it did the right thing. - **Debugging.** When something looks off, the trail shows exactly where a run went sideways — which step, which source, which tool. - **Accountability and compliance.** As AI regulation tightens through 2026, being able to show a reviewable record of automated actions — and the human approvals behind them — moves from nice-to-have to expected. - **Security.** A clear log of what was accessed and changed is the first thing you reach for if you ever need to review an assistant's behavior. ## Audit trail vs. chat history vs. logs These get conflated, but they answer different questions: | | What it shows | What it misses | | --- | --- | --- | | Chat history | The conversation you had | What the AI did in your tools | | System logs | Raw technical events | The human-readable chain and approvals | | Audit trail | The full task: request → tools → sources → approvals → outcome | Nothing critical — that's the point | ## How Arlo does it [Arlo](/glossary/ai-colleague) treats the audit trail as part of how a run works, not a report bolted on afterward. Every connection resolves through policy before a run starts. Anything that sends, writes, or spends waits for a reviewer through an [approval gate](/use-cases/ai-assistant-with-approval-before-sending). And every tool call, source, and approval stays tied to the original request in a trace you can read back. So whether Arlo acted in [Slack](/use-cases/ai-assistant-in-slack), over text, or [on a call](/use-cases/ai-executive-assistant-that-makes-phone-calls), the record of what it did is always there. For what this looks like as a working setup rather than a concept — evidence collection, policy attestation, and reviewable approvals — see [AI assistant for compliance](/use-cases/ai-assistant-for-compliance). ## Frequently asked questions **What is an AI agent audit trail?** It's a complete record of what an AI agent did on your behalf — every tool call, source, and approval — with each step tied back to the request that started it, so you can reconstruct exactly what happened. **How is it different from chat history?** Chat history shows the conversation. An audit trail shows what the agent actually did in your tools as a result — the actions, the data it touched, and who approved each risky step. **Why does an AI assistant need one?** Because once an assistant can take real actions, you need to be able to verify them. An audit trail is what makes it safe to delegate and what you'd reach for to debug or demonstrate compliance. **Does Arlo keep an audit trail?** Yes. Every tool call, source, and approval is logged and tied to the request that started it, alongside an approval gate on anything that sends, writes, or spends. ## Learn more An audit trail is one of the three things that separate an [AI colleague](/glossary/ai-colleague) from a chatbot — the others being memory and governed action. See how it works in practice with an [approval gate on every send](/use-cases/ai-assistant-with-approval-before-sending), or [try Arlo](/signup) and read back your own first run. --- # What is an AI chief of staff? > An AI chief of staff is an AI colleague that runs the connective tissue around a leader: the morning briefing, meeting prep, follow-ups that must not slip, and tasks that span email, calendar, chat, and records. Unlike a human chief of staff it doesn't exercise independent judgment — the strongest setups pair its coverage with approval gates, so leverage compounds without surprises. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/glossary/ai-chief-of-staff A chief of staff's real product is coverage: someone is watching everything around a leader — the commitments, the follow-ups, the prep, the loose threads — so the leader can spend attention where it compounds. An AI chief of staff applies an [AI colleague](/glossary/ai-colleague) to that same surface area: it briefs you each morning, preps you before each meeting, chases what you delegated, and carries cross-tool tasks to done, from the channels you already use. The term needs an honest boundary. A great human chief of staff exercises judgment *as you* — reads a room, kills a bad idea early, represents you in trade-offs. AI does not do that credibly in 2026, and products that pretend otherwise should worry you. What AI does better than any human is the coverage layer: never forgetting, never sleeping, and costing a rounding error. The full trade-off is mapped in [AI chief of staff vs a human](/compare/ai-chief-of-staff-vs-human). ## What an AI chief of staff actually runs - **The morning brief** — one message: what changed overnight, today's shape, the three things at risk of slipping. Arlo delivers it in your texts or Slack from persistent memory across every channel. - **Meeting prep** — who you're seeing, what happened last time, what's open, what you promised. Assembled from your email, calendar, notes, and CRM without being asked twice. - **Commitment tracking** — everything you said you'd do and everything others owe you, chased on schedule so nothing dies in a thread. - **Cross-tool execution** — "move the board meeting, tell the leads, update the doc" is one sentence to an AI chief of staff and four apps' worth of work it does for you. - **Escalation with context** — what needs your judgment arrives framed: the situation, the options, what it recommends, and one tap to approve. ## Where the human line sits Keep the judgment, delegate the coverage. An AI chief of staff shouldn't negotiate your term sheet or manage a conflict between two directors. It should make sure you walked into both conversations briefed, and that every follow-up from both landed. With [approval gates](/glossary/human-in-the-loop) on anything outbound and an [audit trail](/glossary/ai-agent-audit-trail) behind every action, the failure mode of delegation — surprises under your name — is engineered out rather than hoped away. ## Frequently asked questions **What does an AI chief of staff do?** It runs the coverage layer around a leader: daily briefings, meeting prep, commitment tracking, follow-ups, and cross-tool tasks — escalating judgment calls to you with context. **How is an AI chief of staff different from an AI executive assistant?** Largely scope. An [AI executive assistant](/glossary/ai-executive-assistant) centers on calendar, inbox, and logistics; a chief-of-staff setup adds the watching brief — commitments, prep, and cross-team follow-through. With Arlo it's the same colleague configured for a bigger surface. **Can an AI chief of staff replace a human one?** No — it replaces the coverage half of the job, not the judgment half. For leaders who'd never hire a human CoS, it delivers most of the felt value; for those with one, it makes them dramatically more effective. **What tools does it need access to?** Email, calendar, chat, and wherever commitments live (docs, CRM, tickets). Grant access scoped and approval-gated, and extend autonomy as trust is earned. **Where does Arlo fit?** Arlo is an AI colleague that runs exactly this surface — briefings, prep, follow-ups, cross-tool execution — over iMessage/SMS, Slack, WhatsApp, and live phone calls, with approvals and a full audit trail on by default. --- # What is an AI colleague? > An AI colleague is an AI assistant that works inside the tools your team already uses — chat, email, and calls — instead of a separate app. Unlike a basic chatbot, it keeps persistent memory, takes real actions in your software through governed, approved tool access, and logs everything it does. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/glossary/ai-colleague "AI colleague" is the label for a newer kind of assistant: one that lives where your work already happens and can actually get the work done instead of just talking about it. You don't open another chat window or learn another app. You message or call it the way you'd ping a teammate, and it answers — and acts — in iMessage, Slack, WhatsApp, or on the phone. The phrase caught on because "chatbot," and even "AI assistant," undersell what these tools now do. A colleague doesn't just draft a reply when asked. It remembers the account, knows the deadline, sends the message once you approve, and tells you afterward what it did. That shift — from a thing you talk to into a thing you delegate to — is the whole idea. ## The three things that make it a colleague A chatbot answers questions. An AI colleague finishes the job. Three properties make the difference, and most chatbots have none of them: - **Memory.** It remembers your preferences, your projects, and what you said last week, so you stop re-explaining yourself every morning. Context carries across conversations and across surfaces — a thread you start in iMessage still makes sense later in Slack or on a call. - **Action.** It connects to your real tools and does things in them: drafting and sending messages, updating records, booking meetings, running research. Not "here's what you could do," but the thing, done. - **Accountability.** Because it acts, it has to be governed. Sensitive steps wait for your sign-off, and every tool call lands in an [audit trail](/glossary/ai-agent-audit-trail) you can read back. Drop any one of these and you're back to a chatbot. Keep all three and you have something you can actually hand work to. ## What an AI colleague can do Think about the spread of work a good human assistant covers across a dozen apps. That's the target: - Triage your inbox and draft the replies. - Schedule meetings and chase reminders. - Read from and write to the tools you live in — your CRM, project tracker, docs, and spreadsheets. - [Make and take phone calls](/use-cases/ai-executive-assistant-that-makes-phone-calls) for you. - Research a prospect or a question and come back with a tight summary and its sources. - Run recurring work on a schedule and deliver a morning briefing of what changed. [Arlo](/) is one of these. It plugs into Slack, Gmail, Notion, Linear, GitHub, and a few thousand other tools. For apps with no API, it signs in once through a secure browser session and operates them the way you would, by hand. ## AI colleague vs. coworker vs. teammate vs. employee vs. agent These terms blur into each other and get used interchangeably. Most of them point at the same idea; a couple carry different baggage. Here's the honest map: | Term | What it emphasizes | Where it usually lives | | --- | --- | --- | | AI chatbot | Answering questions in a conversation. | Its own chat window. | | AI assistant | Helping you with tasks. | One app or chat surface. | | AI agent | Pursuing a goal on its own, often headlessly. | Behind an API or a workflow. | | [AI coworker](/glossary/ai-coworker) / AI teammate | The same idea as an AI colleague, framed around working alongside people. | The channels a team already uses. | | AI employee / digital worker | A heavier, usually enterprise framing — a "role" a system fills, sometimes pitched with little human oversight. | Enterprise automation and RPA. | | AI colleague | Assistant *and* agent together — memory, real action, and governance — living where you work. | iMessage, Slack, WhatsApp, and calls. | **AI coworker** and **AI teammate** are effectively synonyms for AI colleague — same memory-plus-action-plus-governance idea, just a different word for the working-alongside-you part. **AI employee** and **digital worker** lean enterprise and often imply a fully autonomous "role" with no one watching, which is exactly the thing to be skeptical of. The honest distinction across all of them isn't the label — it's whether the thing has **memory**, can **act** in your real tools, and is **governed** so a human stays in the loop. If you're weighing the words, compare an [AI agent](/glossary/ai-agent), an [AI executive assistant](/glossary/ai-executive-assistant), an [AI personal assistant](/glossary/ai-personal-assistant), and an [AI virtual assistant](/glossary/ai-virtual-assistant) — or browse the full [glossary](/glossary). Arlo is the colleague that spans all of them. ## Why "inside your tools" matters Location is the whole game. A tool you have to remember to visit is a tool you forget by Thursday. Put the assistant in the channels you already check all day — your texts and your team chat — and the friction disappears. Nothing new to install, no new habit to build. You just ask. See it in practice as an [AI assistant you text in iMessage](/imessage), [in Slack](/use-cases/ai-assistant-in-slack), or [on WhatsApp](/whatsapp) — one you can [text or call](/use-cases/ai-assistant-you-can-text-and-call) like a coworker — or read how it stacks up against [Poke](/compare/arlo-vs-poke). ## Is an AI colleague just hype? The idea does get oversold, so it's worth being precise about what's real. What's real today: persistent memory, action across the tools you already use, phone calls, and a clear audit trail. What isn't real — and what you shouldn't want — is a fully autonomous "employee" that needs no oversight and answers to no one. The colleague framing is honest *because* it keeps a human in the loop: the assistant moves fast, but you stay the reviewer on anything that sends, writes, or spends. Speed without that check isn't a colleague; it's a liability. ## How trust works Acting inside real software is only worth anything if it's safe to let it, so governance isn't a setting you turn on later. Access to each tool resolves through policy before every run. High-stakes moves — sends, payments, writes — pause for a reviewer you pick. A full trace links every step back to the request that kicked it off. You can hand off real work and still know exactly what happened. Arlo runs on prepaid credits: 400 credits per $1 at every amount, on monthly plans from $20, and they roll over and expire twelve months after purchase. Admin policy, roles, SSO, and audit export are available on request. ## How an AI colleague fits real roles The same colleague looks a little different depending on the job. [Founders](/use-cases/ai-assistant-for-founders) use it to clear operational work between fundraising and product; [sales teams](/use-cases/ai-assistant-for-sales) use it to follow up on leads fast and keep the CRM honest; [recruiters](/use-cases/ai-assistant-for-recruiters) and [realtors](/use-cases/ai-assistant-for-realtors) lean on the text-and-call speed. See more [use cases](/use-cases) for how it plays out role by role. ## Frequently asked questions **What is an AI colleague?** An AI colleague is an AI assistant that works inside your existing tools — chat, email, and calls — with persistent memory, the ability to take governed actions in your software, and a full audit trail. It's something you delegate to, not just talk to. **How is an AI colleague different from an AI assistant?** "AI assistant" usually means help with tasks inside one app or chat window. An AI colleague adds memory across conversations, real action across your whole stack, and governance — the difference between a helpful chat and a teammate who gets things done. **Is an AI colleague the same as an AI agent?** Not quite. An [AI agent](/glossary/ai-agent) pursues a goal on its own, often headlessly behind an API. An AI colleague is an agent *plus* an assistant: it acts, but it lives in the channels you use and answers to you, with approvals and an audit trail. **What can an AI colleague actually do?** Triage and draft email, schedule meetings, update your CRM and project tools, make phone calls, run research, and deliver a morning briefing — across thousands of integrations, including apps with no API. **Is it safe to let an AI colleague act in my tools?** Yes, when it's governed. With Arlo, connections resolve through policy before each run, anything that sends or writes waits for a reviewer you choose, and every action is logged in an audit trace. **How much does an AI colleague cost?** Arlo starts free — $100 in credits for connecting Slack — then sells prepaid credits: 400 credits per $1 at every amount, $50 is the minimum purchase and buys 20,000 credits, and credits roll over and expire twelve months after purchase. That $50 is roughly 625 back-and-forth messages or 18 apps built; admin controls, SSO, and audit export are available on request. ## Getting started The easiest way in is through a channel you already use. [Try Arlo](/signup), message it in iMessage, Slack, or WhatsApp, or give it a call, and hand off the first thing on your list. --- # What is an AI coworker? > An AI coworker is an autonomous AI that works alongside a team as a digital team member. It takes initiative, owns multi-step work end to end across your connected tools, and reports back — instead of waiting for a prompt like a chatbot. "AI employee," "digital coworker," and "AI teammate" all name the same idea. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/glossary/ai-coworker "AI coworker" is the label for AI that doesn't just answer — it works. Instead of a chat window you open when you have a question, an AI coworker sits alongside your team like a digital team member: it takes initiative, picks up a job, carries it across the tools you already use, and comes back with the result. The same idea travels under a pile of names — **AI employee**, **digital coworker**, **AI worker**, **AI teammate** — and they all point at one thing: software you delegate to, not just talk to. The term caught on because "chatbot" and even "AI assistant" undersell what these tools now do. A coworker doesn't wait to be told each step. Give it a goal and it owns the work from intake to resolution, then reports back — the same way you'd hand something to a person on your team and expect it to come back done. ## AI coworker, AI employee, digital coworker — one concept, many names These are near-synonyms, not distinct products. **AI coworker** and **AI teammate** emphasize working alongside people. **AI employee** and **digital worker** carry a heavier, usually enterprise framing — a "role" a system fills, sometimes pitched with little human oversight, which is the part to be skeptical of. Underneath the vocabulary, the test is the same: does the thing have memory, can it act in your real tools, and is it governed so a human stays in the loop? ## AI coworker vs. AI assistant or chatbot An [AI assistant or chatbot](/glossary/ai-personal-assistant) waits for a prompt and surfaces information. You ask; it answers. The moment the task needs a second step — sending the email, updating the record, chasing the reply — you're back at the keyboard. An AI coworker owns that whole arc: it does the thing, checks the result, and closes the loop. The shift is from a tool you operate to a teammate you hand work to. ## AI coworker vs. AI agent An [AI agent](/glossary/ai-agent) is the underlying building block — autonomous software that perceives, decides, and takes multi-step actions toward a goal. "AI coworker" is that same capability framed as a persistent teammate: an agent with a name, a role, durable memory, and a place on your team. Put simply, every AI coworker is powered by agent technology, but not every agent is a coworker. A headless agent behind an API does a job and disappears. A coworker sticks around — it remembers the account and the deadline, and shows up in the channels your team already lives in. ## AI coworker vs. AI colleague "AI coworker" and "AI colleague" are effectively the same concept — a teammate-shaped AI with memory, action, and governance. [AI colleague](/glossary/ai-colleague) is simply Arlo's preferred framing for it, and the deeper reference page for the whole family. If you're comparing the words themselves, start there; there's no meaningful difference in what the thing actually does. ## What makes an AI coworker actually usable Autonomy alone isn't enough — an autonomous tool you can't trust is a liability, not a teammate. Three properties turn the idea into something a team can rely on: - **Persistent memory.** It remembers your preferences, your projects, and what you said last week, so you stop re-explaining yourself every morning. Context carries across conversations and across surfaces. - **Governed action.** Because it acts in real software, sensitive steps — anything that sends, writes, or spends — pause for [approval before sending](/use-cases/ai-assistant-with-approval-before-sending). You move fast; you stay the reviewer. - **A full audit trail.** Every tool call, source, and approval lands in an [audit trail](/glossary/ai-agent-audit-trail) you can read back, so you always know exactly what your coworker did and why. It also has to live where the team already works. A coworker you have to remember to visit is one you forget by Thursday. The useful ones show up in your texts, your team chat, and on the phone — no new app, no new habit. ## Where the term comes from The "AI coworker / AI employee / digital worker" framing is now common across the market — from Salesforce's Agentforce and Coworker.ai to a wave of enterprise "digital worker" platforms — each staking out some version of AI-that-does-the-job. They differ mainly in how much human oversight they assume. The honest version keeps a person in the loop on anything consequential; that's what separates a genuine teammate from an unaccountable black box. ## Frequently asked questions **What is an AI coworker?** An AI coworker is an AI you delegate work to the way you would a colleague — reachable in the channels your team already uses, carrying memory of past decisions, and able to take real action in your tools rather than only answering questions. **How is an AI coworker different from a chatbot?** A chatbot returns text. An AI coworker finishes the task: it reads the tools behind the request, drafts the output, and — once a reviewer approves — actually sends the email, opens the ticket, or updates the record. **Is an AI coworker the same as an AI agent?** Closely related. "[AI agent](/glossary/ai-agent)" describes the technical capability to plan and use tools; "AI coworker" describes the working relationship — where it sits, how you reach it, what it is trusted with, and who reviews it. **Can an AI coworker replace a hire?** It absorbs the recurring, delegable slice of a role — triage, follow-up, scheduling, data entry, chasing — not the judgement, relationships, or accountability. Teams typically use one to stop a hire being necessary sooner rather than to remove one. **How do you keep an AI coworker from doing something wrong?** By scoping what it may touch and reviewing what it does: policy resolves each connection before a run, consequential actions wait for a named approver, and every run leaves an [audit trace](/glossary/ai-agent-audit-trail). Autonomy is extended per category as each one proves boring. **What does an AI coworker cost?** Arlo starts free — $100 in credits for connecting Slack — then sells prepaid credits: 400 credits per $1 at every amount, $50 is the minimum purchase and buys 20,000 credits, and credits roll over and expire twelve months after purchase. At about 32 credits a back-and-forth message, $50 is roughly 625 messages or 18 apps built. ## Meet Arlo Arlo is an AI colleague you talk to like a coworker — in iMessage, SMS, Slack, WhatsApp, and on live phone calls. It connects to 3,000+ tools, and for anything without an API it logs in once through a secure browser session and works the site like a person. It keeps persistent memory and sends you a morning briefing. Governance comes standard: every send and tool write waits for your approval, and every action lands in a full audit trace. It runs on prepaid credits — 400 credits per $1 at every amount, on monthly plans from $20, and they last twelve months; admin controls, SSO, and audit export are available on request. Want a coworker that acts, with guardrails already in place? [Try Arlo](/signup) and hand off the first thing on your list. If you're still mapping the terms, compare an [AI colleague](/glossary/ai-colleague), an [AI agent](/glossary/ai-agent), and an [AI executive assistant](/glossary/ai-executive-assistant). --- # What is an AI employee? Definition, examples, and limits > An AI employee is an AI system a team treats like a hire: it receives work in plain language, carries tasks to completion across the company's tools, and is accountable for the result. It differs from a chatbot (answers only) and workflow automation (fixed steps). Products like Arlo and Viktor define the category; the strongest pair autonomy with approvals, scoped access, and an audit trail. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/glossary/ai-employee "AI employee" is the 2026 term for an AI system that a team treats like a hire rather than a tool: you delegate work to it in plain language, it carries the task through to a finished result across your company's software, and it's expected to be accountable for what it did. You @mention it in Slack or text it the way you'd message a teammate — and the output isn't an answer, it's completed work. The term is marketing-flavored — companies like Viktor built their positioning on it, and Arlo's own term of art is [AI colleague](/glossary/ai-colleague) — but the underlying category is real, and it's worth defining precisely because the gap between products that *chat about* work and products that *do* work is the widest gap in AI right now. ## What makes something an AI employee Four properties separate the category from its neighbors: 1. **Delegation, not prompting.** You hand it outcomes ("chase these five invoices," "get the QBR deck drafted"), not step-by-step instructions. 2. **Real action in real tools.** It works inside the company's actual stack — email, CRM, docs, tickets — rather than producing text for a human to paste. 3. **Persistence.** It remembers context across tasks and days; you don't re-explain the company every morning. 4. **Accountability.** What it did is inspectable — who asked, what it touched, what went out — the way you could ask an employee "walk me through what happened." ## AI employee vs the neighbors | Term | What it does | Where it falls short of "employee" | |---|---|---| | Chatbot / assistant | Answers questions in a session | Talks; doesn't do | | [AI agent](/glossary/ai-agent) | Pursues a goal autonomously, often headless | Capable, but unmanaged — no channel, no accountability | | Workflow automation | Repeats fixed steps on triggers | Breaks on anything it wasn't scripted for | | **AI employee / colleague** | Takes delegated work to completion in your tools | The real thing — when governed | ## Examples, honestly labeled As of July 2026 the recognizable products in and around the category: **Arlo** (an AI colleague across iMessage/SMS, Slack, WhatsApp, and live phone calls, with approvals and a full audit trail on by default), **Viktor** (an AI employee in Slack/Teams known for code execution and heavy delegation — [compared here](/compare/arlo-vs-viktor)), and adjacent products like Lindy (build-your-own agents) and personal assistants like Poke that apply the same idea to one person's life rather than a team's work. ## The limits that matter An AI employee is not a person, and pretending otherwise is how teams get burned. It doesn't own outcomes — you do; it executes. It's only as good as the access you give it, which is why the serious products are built around **governance**: scoped, [approval-gated](/glossary/human-in-the-loop) access with an [audit trail](/glossary/ai-agent-audit-trail), so autonomy is extended deliberately rather than assumed. And it fails differently than a person — confidently and fast — which is exactly why the approval gate exists on anything consequential. ## Frequently asked questions **What does "AI employee" actually mean?** An AI system a team delegates real work to — received in plain language, completed in the company's tools, inspectable afterward. It's a category label, not a legal or HR concept. **Is an AI employee the same as an AI agent?** An [AI agent](/glossary/ai-agent) is the underlying capability — autonomous goal pursuit. An AI employee is an agent packaged for a team: reachable in your channels, persistent, and governed. Every AI employee is an agent; most agents are not AI employees. **What's the difference between an AI employee and an AI colleague?** Mostly emphasis. "AI employee" stresses delegation and headcount-like output; "AI colleague" — Arlo's term — stresses working alongside you in your channels with approvals and accountability. The products overlap heavily. **Can an AI employee replace a human hire?** It absorbs the coordination layer of roles — follow-ups, triage, updates, scheduling, reporting — which can defer a hire or return hours to a team. Judgment, relationships, and ownership stay human. **What should I check before giving one real access?** Three things: scoped permissions (it can touch only what you granted), approval gates on consequential actions, and a full audit trail. If a product can't show all three, don't connect it to customers. --- # What is an AI executive assistant? > An AI executive assistant is software that does the scheduling, inbox triage, coordination, and communication a human EA does. It works inside your existing tools, keeps persistent memory, takes real actions in your software through approved access, and logs everything it does. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/glossary/ai-executive-assistant An AI executive assistant is software that does the scheduling, inbox triage, coordination, and communication work a human EA handles. The thing that sets it apart is where it lives: inside the chat, email, and calendar you already use, not in yet another app you have to open. It remembers your context across days, takes real actions in your software once you approve them, and logs what it did. A chatbot answers questions. This one gets things off your plate. ## What an AI executive assistant does A good EA is mostly logistics and follow-through. They keep the calendar clean, work the inbox down, draft the replies you'd write anyway, chase the people who owe you answers, and make sure nothing falls through. An AI version covers the same ground, except it's working the whole day without getting tired of it. - **Scheduling** — finding times, booking meetings, untangling conflicts. - **Inbox** — triaging email, drafting replies, surfacing the handful of things that actually need you. - **Coordination** — chasing responses, nudging stalled projects, tracking who promised what. - **Communication** — sending messages on your behalf, and in some cases handling live phone calls. - **Briefings** — a quick summary of what moved overnight so you're not starting cold. The line between this and a basic assistant tool is action. A chatbot tells you what to do. An AI executive assistant goes and does it in your real systems, after you've signed off. ## Who it's for This is for people whose time is the bottleneck and who either can't justify a full-time human EA or can't find a good one. Founders, executives, salespeople, real estate agents, small teams. If your day is mostly calendar and inbox, that's exactly the load this takes off you. See how it plays out for [founders](/use-cases/ai-assistant-for-founders), [realtors](/use-cases/ai-assistant-for-realtors), and [sales teams](/use-cases/ai-assistant-for-sales). ## How it differs from a chatbot and a human EA A generic chatbot sits in its own window, forgets you between sessions, and only produces text. A human EA remembers everything and gets work done, but they cost a salary, keep set hours, and can't be in two places at once. The pitch for an AI executive assistant is the memory and action of a person, available the whole time, right inside your tools. (When the scope grows past logistics into briefings, commitments, and cross-team follow-through, the role becomes an [AI chief of staff](/glossary/ai-chief-of-staff).) | | Human EA | AI executive assistant | Basic chatbot | |---|---|---|---| | Lives in your tools | Partially | Yes — chat, email, calls | No — separate app | | Persistent memory | Yes | Yes | No | | Takes real actions | Yes | Yes, with approvals | No — text only | | Availability | Set hours | Continuous | On demand | | Audit trail | Informal | Full, logged | None | | Cost | Full salary | A fraction of a salary | Low / free | ## What to look for Plenty of tools call themselves assistants. Most are chatbots with a nicer label. Here's what actually separates the real thing: - **Persistent memory.** It should know your preferences, your people, and the work in flight, and tell you what changed without being asked. - **Real tool access.** It should connect to your calendar, email, and the rest of your stack and take action. The good ones can even drive tools that have no API by working in a secure browser session. - **Governance and approvals.** Drafts, sends, and writes should run through policy and wait for a reviewer you pick before anything leaves the building. - **An audit trail.** Every action, source, and approval logged, so you can check the work instead of trusting it blind. Miss those and you've bought a chatbot. The related idea of an [AI colleague](/glossary/ai-colleague) takes the same thinking past one person's logistics into a shared teammate for the whole workspace. ## Arlo as an AI executive assistant [Arlo](/) works inside iMessage, SMS, Slack, WhatsApp, and live phone calls, so there's no new chat app to learn. It keeps persistent memory and sends a morning briefing of what changed. It connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools, and for the ones with no API it logs in once in a secure browser session and works like a person would. Governance isn't an add-on. Every connection resolves through policy before each run, sends and writes wait for a reviewer you choose, and every tool call, source, and approval lands in a full audit trace. Arlo runs on prepaid credits — 400 credits per $1 at every amount, on monthly plans from $20, and they last twelve months; admin policy, roles, SSO, and audit export are available on request. For how it stacks up against other assistants, see [Arlo vs Poke](/compare/arlo-vs-poke) and [Poke alternatives](/compare/poke-alternatives). ## Frequently asked questions **What is an AI executive assistant?** An AI executive assistant handles the coordination layer of an executive's day — triaging email, scheduling and rescheduling, drafting replies and follow-ups, preparing briefings, and chasing the things that go quiet — and takes those actions in your real tools rather than just advising you to. **How is it different from a human executive assistant?** A human EA brings judgement, discretion, and relationships. An AI EA brings availability and volume: it works at any hour, never drops a thread, and is priced like software rather than a salary. The realistic pattern is an AI EA absorbing the mechanical majority so a human EA (or the executive) spends attention on what actually needs it. **Can an AI executive assistant make phone calls?** Some can. Arlo places live calls — confirming an appointment, chasing a vendor, passing a message — and keeps the transcript, so you can ask how it went. Most AI assistants are text-only. **Is it safe to give an AI assistant access to my inbox and calendar?** The safeguard is scoping plus review, not trust. Connections resolve through policy before each run, sends and writes wait for a reviewer you choose, and every tool call, source, and approval lands in a full [audit trace](/glossary/ai-agent-audit-trail). **Where do you reach it?** Arlo works where you already are — [iMessage and SMS](/imessage), [Slack](/use-cases/ai-assistant-in-slack), [WhatsApp](/whatsapp), and live phone calls — with one memory across all of them. **What does an AI executive assistant cost?** Arlo starts free — $100 in credits for connecting Slack — then sells prepaid credits: 400 credits per $1 at every amount, $50 is the minimum purchase and buys 20,000 credits, and credits roll over and expire twelve months after purchase. That $50 is roughly 625 back-and-forth messages — against a human EA salary, the comparison is not close on price alone. ## Try Arlo Want an assistant that lives where you already work and only acts with your approval? [Try Arlo](/signup). --- # What is an AI personal assistant? > An AI personal assistant is software that manages personal and work tasks for you: reminders, messages, calls, research, and running your apps. Older voice assistants like Siri or Alexa mostly answer questions. Modern ones take real actions inside your tools, remember your preferences, and report back when the work is done. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/glossary/ai-personal-assistant An AI personal assistant is software that handles your everyday tasks. It sets reminders, sends messages, makes calls, does research, and runs the apps you already use. Siri and Alexa were the first wave, but they mostly answered questions and flipped a few smart switches. The newer ones actually do things, remember what you told them last week, and tell you when a task is finished. That last part is the whole game. A voice assistant reads you the weather. An action-taking assistant reads your inbox, drafts a reply, books the call, looks up the vendor, edits the doc, and then tells you what it did. ## What an AI personal assistant can do Most capable assistants cover the same ground: - **Reminders and scheduling** — timed nudges, calendar events, and follow-ups so nothing slips. - **Messages** — drafting, sending, and triaging your texts and email. - **Calls** — placing or answering phone calls to confirm appointments or chase down an answer. - **Research** — searching, comparing options, and handing you the short version. - **Apps** — operating the software you live in, from email and calendars to project tools. The good ones meet you where you already are, in your messages and on the phone. The mediocre ones make you open yet another app and remember to check it. ## Voice assistant vs. action-taking AI personal assistant Three things separate the old guard from the new: memory, the ability to act, and how deeply they plug into your tools. | Capability | Voice assistant (Siri/Alexa style) | Action-taking AI personal assistant | |---|---|---| | Primary mode | Answers questions, controls devices | Completes multi-step tasks end to end | | Memory | Little or none between sessions | Persistent memory of your preferences and context | | Tool access | Limited, fixed integrations | Connects to email, calendar, and many apps | | Apps without an API | Cannot use them | Logs in and operates them like a person | | Accountability | No record of actions | Reports back and keeps a trail of what it did | A voice assistant is reactive and forgets everything between requests. An action-taking assistant remembers and follows through. ## Why memory and real tool access matter Strip those two things away and you're left with a chatbot that can describe your to-do list but not touch it. Memory is what saves you from re-explaining yourself. The assistant knows your preferences, the projects you have open, and what you talked about yesterday, so it can give you a morning briefing of what changed and pick up mid-thread. Real tool access is what lets it finish the job instead of narrating it. It connects to the services you use, and for the ones with no API, it logs in once in a secure browser session and clicks, reads, and fills out forms the way you would. This is also why governance starts to matter. Once an assistant can act on your behalf, you want a say in it. Drafts and sends should be able to wait for your approval, and every action should be written down so you can check exactly what happened. ## How an AI personal assistant fits with related roles The label blurs into a few others. An AI personal assistant — this page — skews toward everyday personal and work tasks. An [AI executive assistant](/glossary/ai-executive-assistant) leans harder into scheduling, inbox, and coordination at a professional level. An [AI colleague](/glossary/ai-colleague) is the broader idea of an assistant that works inside your team's shared tools with governed, audited actions. Picking one? It helps to compare approaches. See [Arlo vs. Poke](/compare/arlo-vs-poke) and [Poke alternatives](/compare/poke-alternatives). ## Arlo as a text- and call-based AI personal assistant [Arlo](/) lives in the channels you already use: iMessage and SMS, Slack, WhatsApp, and live phone calls. There's no separate app to learn. It keeps persistent memory and sends a morning briefing of what changed. It connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools, and for software with no API, it logs in once in a secure browser session and operates it directly. Because Arlo actually acts, governance is part of the design. Connections resolve through policy before each run, sends and tool writes can wait for a reviewer you choose, and every tool call, source, and approval lands in a full audit trace. Arlo runs on prepaid credits — 400 credits per $1 at every amount, on monthly plans from $20, and they last twelve months; admin policy, roles, SSO, and audit export are available on request. It tends to land best with [founders](/use-cases/ai-assistant-for-founders), [realtors](/use-cases/ai-assistant-for-realtors), and [sales teams](/use-cases/ai-assistant-for-sales) who run their day from their messages and the phone. If you're weighing which of the two you actually need, [AI colleague vs AI personal assistant](/compare/ai-colleague-vs-personal-assistant) sets the distinction out side by side. ## Frequently asked questions **What is an AI personal assistant?** An AI personal assistant handles the admin of your own life and work — messages, scheduling, reminders, bookings, follow-ups, research — through conversation, and takes the action rather than handing you instructions. **How is it different from Siri or Google Assistant?** Voice assistants execute single commands against a narrow set of built-in functions. An AI personal assistant holds context across a conversation, remembers what happened last week, works across your connected tools, and completes multi-step tasks that take more than one action. **Can an AI personal assistant book things and send messages for me?** Yes — that is the point of one. With Arlo, anything consequential waits for your approval first: you see what it is about to send and approve or deny it, and an approval that is never answered expires instead of going out. **Do I need to install an app?** Not with Arlo — you text a number from [iMessage or SMS](/imessage), or call it. It also works in [WhatsApp](/whatsapp), and in [Slack](/use-cases/ai-assistant-in-slack) if your work lives there. **Is a personal AI assistant private?** It depends on the product, so check what any assistant retains and who can see it. With Arlo, connections resolve through policy before each run and every run is inspectable in a full [audit trace](/glossary/ai-agent-audit-trail), so what it accessed is a matter of record rather than trust. **What does an AI personal assistant cost?** Arlo starts free — $100 in credits for connecting Slack — then sells prepaid credits: 400 credits per $1 at every amount, $50 is the minimum purchase and buys 20,000 credits, and credits roll over and expire twelve months after purchase. At about 32 credits a back-and-forth message, $50 is roughly 625 messages or 69 researched answers with sources. ## Try Arlo Want an assistant that texts, calls, and actually gets work done? [Try Arlo](/signup) and start handing off tasks today. --- # What is an AI phone assistant? > An AI phone assistant is an AI that handles real phone calls on your behalf: it places outbound calls (confirmations, follow-ups, bookings), receives inbound ones (fielding questions, capturing details, escalating what needs you), and reports back with a summary and updated records. It differs from voice control like Siri, which dials for you but leaves the conversation to you. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/glossary/ai-phone-assistant An AI phone assistant is an AI that has phone conversations *for* you, not just *with* you. You delegate a call the way you'd delegate an email — "confirm Thursday's booking," "call the supplier about the missing tracking" — and the assistant dials, speaks with whoever answers, navigates the back-and-forth, and comes back with the outcome. The good ones also work the other direction: answering your line, fielding the common questions, capturing details, and escalating the calls that genuinely need you. It's worth defining because the phone remains where an enormous share of real-world coordination actually happens — clinics, vendors, restaurants, county offices, customers of every local business — and it's the surface most AI products silently skip. Listing who can and can't call is short work: [we keep the honest list here](/compare/ai-assistants-that-can-make-phone-calls). ## What one does, concretely - **Outbound**: places the call, states who it is and why it's calling, holds a natural conversation — including holds, IVR menus, and "let me check" pauses — and pursues the task to an outcome. - **Inbound**: answers, handles the questions it can, captures name/number/reason for the rest, and escalates by text or transfer when a human is needed. - **Afterward** — the part that separates a toy from a tool: a transcript you can go back to, and your records (calendar, CRM, notes) updated to match what was agreed on the call. ## AI phone assistant vs voice assistant vs IVR | | Voice assistant (Siri etc.) | Phone tree / IVR | AI phone assistant | |---|---|---|---| | Who has the conversation | You | Nobody — menus | The AI, on your behalf | | Direction | Dials for you | Inbound only | Outbound + inbound | | Outcome captured in your tools | — | — | ✓ | | Example | "Call Mom" | "Press 2 for billing" | "Call the venue and confirm Thursday" | ## How Arlo treats the phone For Arlo, the phone is a first-class channel of an [AI colleague](/glossary/ai-colleague), not a bolt-on — [here is what that looks like as a product](/phone). The same colleague you text in iMessage or @mention in Slack places calls, with one memory across all of it. That continuity is the practical part: you text a request, Arlo calls, and you can ask how it went in the same thread. Governance applies to calls like any consequential action: approval gates where it matters, and every call in the [audit trail](/glossary/ai-agent-audit-trail) with a transcript you can check. The [phone-calling executive assistant guide](/use-cases/ai-executive-assistant-that-makes-phone-calls) shows the end-to-end pattern. ## Frequently asked questions **What is an AI phone assistant?** An AI that conducts real phone calls on your behalf — placing and receiving them, holding the conversation, and reporting the outcome with your records updated. **Is Siri an AI phone assistant?** No — Siri dials numbers and can screen or hold, but you still have the conversation. An AI phone assistant has the conversation itself. **Do AI phone calls sound robotic?** Modern AI voices are close to human in pacing and tone. The bigger quality bar is behavioral: honest identification, staying on task, and graceful escalation to you. **Should the AI say it's an AI on a call?** Yes — honest identification is both good practice and, in a growing number of places, a legal expectation. A trustworthy product doesn't rely on passing as human. **Which products actually do this?** As of September 2026: Arlo places calls for work, with an audit trail; Google's Ask for Me calls local businesses; Meta's Muse and Instinct call businesses in beta; and Catch, Pally and Martin place calls too. Most AI assistants — including ChatGPT and Slack-only AI employees — don't touch the phone. --- # What Is an AI Virtual Assistant? Definition and Guide > An AI virtual assistant is software that does the admin work a human VA would, like scheduling, email triage, reminders, research, and data entry, but automatically. The good ones have persistent memory, real access to your tools, and governance so a person approves sensitive actions. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/glossary/ai-virtual-assistant An AI virtual assistant is software that does the same admin and support work a human virtual assistant (VA) would, like scheduling, email triage, reminders, research, and data entry, only it runs automatically. You're not handing tasks to a remote person. You're handing them to a program that connects to your tools and acts for you. The ones worth using add three things: persistent memory, real access to your accounts, and governance so a reviewer signs off on anything sensitive before it happens — at which point the more useful word for it is an [AI colleague](/glossary/ai-colleague). ## What an AI virtual assistant does Same recurring admin load a human VA carries, run by software: - Scheduling meetings and untangling calendar conflicts - Triaging, drafting, and replying to email - Setting reminders and chasing follow-ups - Researching people, companies, and options before a decision - Updating data across apps - Summarizing long threads, documents, and calls What sets it apart from a chatbot is action. A chatbot answers your question and stops. An AI virtual assistant logs into your real accounts and actually does the thing. ## AI virtual assistant vs. human VA vs. voice assistant People blur these three. They solve different problems. | | AI virtual assistant | Human VA | Voice assistant (Siri, Alexa) | |---|---|---|---| | Availability | 24/7, instant | Working hours | 24/7 | | Cost | Subscription or prepaid credits | Hourly or salaried | Bundled with device | | Task depth | Multi-step workflows across tools | Multi-step, judgment-heavy | Single commands | | Memory | Persistent across tasks | Personal memory | Little to none | | Tool access | APIs plus browser actions | Whatever they can log into | Limited integrations | | Best for | Repeatable admin and ops work | Nuanced, relationship-heavy work | Quick device commands | A human VA still beats software on judgment, relationships, and the fuzzy tasks where someone has to read between the lines. A voice assistant is for quick one-off commands, not work you can leave running. The AI virtual assistant lives in between. It does real multi-step work continuously and it remembers the context from one task to the next. ## Where an AI virtual assistant helps most Anywhere the same admin work keeps coming back, it pays for itself: - Founders trying to protect their time and keep the inbox and calendar from eating the day. See our [AI assistant for founders](/use-cases/ai-assistant-for-founders) guide. - Sales teams researching prospects and keeping the CRM honest ([AI assistant for sales](/use-cases/ai-assistant-for-sales)). - Realtors running showings, follow-ups, and a steady stream of client questions ([AI assistant for realtors](/use-cases/ai-assistant-for-realtors)). - Ops roles stuck with routine data entry and updates that span four apps. If you need support closer to the executive level, here's how this compares to an [AI executive assistant](/glossary/ai-executive-assistant). ## What to look for Plenty of products call themselves AI virtual assistants and can't safely do much. Three things separate a real one from a demo. Persistent memory comes first. It should know your preferences, your people, and what you're in the middle of, instead of meeting you fresh every morning. Real tool access is second, and it's where most products quietly fall short. It has to connect to the apps you already live in, including the ones with no public API, by logging in once and working through a secure browser session. Governance is the one people skip and the one that bites. Sensitive sends and writes should route through a reviewer you pick, with a full record of every tool call, source, and approval behind them. An assistant that can touch your accounts with nobody watching is a liability. One that asks before it does the things that matter is a colleague. ## How Arlo works as an AI virtual assistant [Arlo](/) is an AI virtual assistant that works where you already are: iMessage and SMS, Slack, WhatsApp, and live phone calls. Nothing new to install. You text it the way you'd text a person, and it goes and does the work. Arlo keeps persistent memory and sends a morning briefing of what changed overnight. It connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools, and for apps with no API it logs in once through a secure browser session and clicks, reads, and fills out forms like a person would. Governance is on by default. Every connection runs through policy before each run, drafts and sends wait for a reviewer you choose, and every action lands in a full audit trace. For teams, Arlo runs on prepaid credits — 400 credits per $1 at every amount, on monthly plans from $20, and they last twelve months; admin policy, roles, SSO, and audit export are available on request. ## Frequently asked questions **What is an AI virtual assistant?** An AI virtual assistant is software that takes on the delegable work a remote human assistant would handle — inbox triage, scheduling, follow-ups, data entry, research — through conversation, acting in your connected tools instead of only drafting suggestions. **How is an AI virtual assistant different from a human VA?** A human VA brings judgement and can own ambiguous work end to end. An AI VA answers instantly, works at any hour, and scales with usage rather than headcount — but needs its scope defined and its consequential actions reviewed. **What can an AI virtual assistant actually do?** With Arlo: triage and reply to email, schedule and reschedule, open and update records in connected tools, run recurring work on a schedule, place phone calls, deliver a morning briefing, and operate systems that have no API through a secure browser session. **Can it work with tools that have no integration?** Yes. Where there is no API, Arlo signs in once through a secure browser session and clicks, reads, and fills forms the way a person would — under the same approvals and audit trail as any other action. **Is it safe to let a virtual assistant act on my accounts?** The control is scoping and review: policy resolves each connection before a run, consequential sends and writes wait for a reviewer you name, and every run lands in a full [audit trace](/glossary/ai-agent-audit-trail). **What does an AI virtual assistant cost?** Arlo starts free — $100 in credits for connecting Slack — then sells prepaid credits: 400 credits per $1 at every amount, $50 is the minimum purchase and buys 20,000 credits, and credits roll over and expire twelve months after purchase instead of at the end of the month. That $50 is roughly 625 back-and-forth messages or 69 researched answers — set against an hourly or salaried human VA. ## Try Arlo Want an AI virtual assistant that works inside your messages and on your calls, with memory and governance built in? [Sign up for Arlo](/signup), or browse more [resources](/resources) to see how it fits your team. --- # What is human-in-the-loop AI? (HITL, explained) > Human-in-the-loop (HITL) AI is a design where a person stays at the decision points of an AI system: the AI does the work, but consequential actions — sends, purchases, changes, calls — wait for human approval, and everything is reviewable afterward. It's how teams trust AI with real access: autonomy is extended deliberately, category by category, instead of granted all at once. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/glossary/human-in-the-loop Human-in-the-loop (HITL) is the design principle that keeps a person at the decision points of an AI system. The AI reads, drafts, plans, and prepares; a human approves the moments that matter — the email before it sends, the call before it's placed, the record before it changes hands. It's the difference between delegating work and abdicating it. The term came from machine learning (humans labeling data and correcting models in training loops), but in 2026 its everyday meaning has shifted to **agentic systems**: now that [AI agents](/glossary/ai-agent) can genuinely act — send, buy, book, call, modify — the question "where does the human sit?" became the central trust question of the category. HITL is the answer that works. ## The three loops 1. **Approval (before)** — consequential actions queue for a human yes. The practical art is scoping: gate by consequence (outbound, irreversible, money, customers), not by blanket rule, or you've rebuilt the work you delegated. 2. **Exception (during)** — the AI proceeds on routine work but escalates when confidence drops or stakes spike: the angry customer thread, the invoice that doesn't match, the call going sideways. 3. **Review (after)** — everything is inspectable after the fact via an [audit trail](/glossary/ai-agent-audit-trail): what ran, what it touched, who approved. Review is what lets approval gates loosen safely over time. Mature systems use all three, and move work gradually from loop 1 toward loop 3 as trust is earned — which is the actual, unglamorous mechanism by which "AI adoption" happens without incident. ## What this looks like in practice In Arlo, HITL is the default state, not a feature flag. Every draft, send, and tool write waits on [a reviewer you pick](/use-cases/ai-assistant-with-approval-before-sending); approvals arrive where you already are (a tap in iMessage or Slack); policy checks run before every action; and every run lands in a full audit trail. As specific categories prove boring — the routine reschedule, the internal update — you loosen their gates deliberately. Autonomy becomes something the system *earns per category*, never something you granted in one nervous click. ## The honest trade-off HITL costs latency: an approval queue is only as fast as its human. Badly scoped, it becomes rubber-stamping — all the clicks of doing the work with none of the attention. The design answer is asymmetry: approvals only where consequences are real, exception-based escalation everywhere else, and enough context in each approval request (what, why, to whom, with one tap to act) that reviewing takes seconds, not re-research. ## Frequently asked questions **What does human-in-the-loop mean in AI?** A design where humans hold the decision points of an AI system — approving consequential actions before they happen, handling escalations, and reviewing what ran afterward. **Why does human-in-the-loop matter for AI agents?** Because agents act. An agent with tool access can send, change, and spend; HITL is how you get that leverage without accepting silent failure as a cost of doing business. **Is human-in-the-loop slower?** The approval step adds a tap where consequences are real; everything else runs at machine speed. Well-scoped gates cost seconds and prevent the class of mistakes that costs days. **What's the difference between human-in-the-loop and human oversight?** Oversight is often after-the-fact review alone. HITL puts the human *before* consequential actions too — approval, not just autopsy. **How does Arlo implement it?** Approval gates on every send and write by default, a reviewer you choose, policy checks before every run, and a full audit trail after — with autonomy extended per category as trust is earned. --- # AI agent for ads — Google, Meta, LinkedIn, X > An AI agent for ads runs the daily grind of your ad accounts: it pulls performance across Google, Meta, LinkedIn, and X, flags wasted spend and creative fatigue, drafts budget and campaign changes, and applies them only after you approve. Arlo does this from a text or Slack thread — a spend brief every morning, one tap per change, and a full audit trail of everything it touched. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-agent-for-ads Ad accounts punish inattention. Budgets drift, a broken conversion tag burns a week of spend, creative fatigues quietly, and the platform's own "recommendations" are optimized for the platform. Managing it well is a daily job — checking, comparing, adjusting — that most teams do weekly at best. An AI agent for ads does the daily job daily: it watches your accounts, tells you what actually changed, drafts the fixes, and — this part is non-negotiable — waits for your approval before touching spend. That last clause is why the category needs care. Plenty of tools will "optimize" your campaigns autonomously; anyone who has run real budgets knows the horror stories. The right shape for an ads agent is [human-in-the-loop](/glossary/human-in-the-loop): machine diligence, human sign-off on every dollar that moves. (Ads are also one lane of the broader [AI marketing agent](/use-cases/ai-marketing-agent) role — same colleague, wider brief.) ## What Arlo does with your ad accounts - **The morning spend brief** — one message in iMessage or Slack: yesterday's spend and results by platform, what moved, what looks off. No dashboard safari. - **Waste detection** — campaigns past their frequency ceiling, keywords eating budget without converting, audience overlap, a landing page that started 404ing. Flagged in plain language, with the dollar amount at stake. - **Drafted changes, not surprise changes** — "shift $40/day from campaign A to B," "pause these three ad sets," "raise the cap on the winner." Each arrives as a proposal with the reasoning; you approve with a tap and Arlo applies it. - **Cross-platform answers** — "what's our blended CAC this week?" is one question, not four exports. Arlo connects to Google, Meta, LinkedIn, and X ads accounts among 3,000+ integrations, and can pull a proper report on request. - **Weekly rollups for whoever asks** — a clean performance summary to your channel, your client, or your inbox, on schedule. - **The audit trail** — every report pulled, every change applied, every approval, in one [reviewable trace](/glossary/ai-agent-audit-trail). When someone asks "who changed the budget?", there's an answer. ## Why approval-first matters more for ads than anywhere else An ads mistake is a special kind of expensive: it spends real money, silently, at machine speed. That's why Arlo's defaults are conservative — read-and-report freely, but **every write to an ad account waits on your approval**. Over time you can loosen specific, boring categories ("always pause anything spending with zero conversions for 3 days") while keeping human sign-off on budget moves. Autonomy is earned per category, never assumed — the opposite of handing your account to a black-box optimizer. ## The weekly loop, concretely What "managing your ads" means in practice, one week at a time: - **Daily (2 minutes of yours):** the morning spend brief lands in your thread — spend and results by platform, anything anomalous in dollars. Most days you read it and move on. - **When something's off:** Arlo flags it the day it happens — "campaign B's CPA doubled yesterday; the landing page is returning 404s" — with the proposed fix attached, not just the alarm. - **Midweek:** approval cards for whatever the data justified: the budget shift, the negative keywords, the paused ad set. Each one carries the why, the expected impact, and the rollback. You approve in taps. - **Friday:** the weekly rollup — week over week, cost per goal-event, what was changed and what it did, what's deliberately being left alone. Forwardable to a client or a cofounder as-is. - **Monthly:** a step back — creative fatigue audit, audience saturation, the experiment queue for next month, and anything structural (tracking drift, overlap) that crept in. The division of labor stays fixed: Arlo does the watching, the math, and the drafting; you do the judgment, one tap at a time. ## The numbers that matter, by objective An ads agent is only as good as the metric it optimizes. The honest mapping: | Your objective | Optimize toward | Treat as noise | |---|---|---| | Lead gen / demos | Cost per qualified conversation | Raw clicks, CTR bragging | | E-commerce | Blended CAC and contribution margin | Platform-attributed ROAS alone | | Early-stage learning | Replies and conversations started | Vanity conversions on tiny samples | | Brand / launch | Reach and frequency inside caps | Any "engagement" metric | If a platform metric and your business metric disagree, the business metric wins — and if conversion volume is too small to conclude anything, the right report says exactly that. ## Agent vs the platform's built-in automation | | Platform auto-rules / "AI" campaigns | Arlo | |---|---|---| | Whose interest it optimizes | The platform's | Yours | | Sees across Google + Meta + LinkedIn + X | — | ✓ | | Explains changes in plain language | — | ✓ Reasoned proposals | | Waits for your approval | — | ✓ On by default | | Reports where you actually are (text/Slack) | — | ✓ | | Audit trail of every change | Partial, per platform | ✓ Unified | Platform automation has its place — smart bidding inside a campaign works. The agent sits above it: cross-platform eyes, your priorities, your sign-off. ## Getting started Connect your ad accounts, tell Arlo your targets ("keep blended CAC under $85, flag anything unusual daily"), and keep every write gated for the first two weeks. You'll get the morning brief immediately; the trust to loosen gates comes from watching its proposals be right. Arlo starts free with $100 in credits for connecting Slack, then runs on prepaid credits — $50 is the minimum top-up, and credits last 12 months — against what one silent week of wasted spend costs. ## Frequently asked questions **Can an AI agent really manage my ad campaigns?** Yes, with the right guardrails: it watches performance daily, flags waste, and drafts changes — but applies them only after you approve. Arlo ships that approval-first model by default. **Which ad platforms does it work with?** Arlo connects to major ad platforms — Google, Meta, LinkedIn, X — among 3,000+ integrations, and can operate dashboards with no API through a secure browser session. **Will it change budgets without asking?** No. Every write to an ad account waits on your approval by default. You can later authorize narrow, routine actions (like pausing zero-conversion ad sets) while keeping sign-off on budget moves. **How is this different from Google's automated recommendations?** Platform recommendations optimize inside one platform, with the platform's incentives. An agent works across all your platforms, optimizes for your targets, explains itself, and answers to you. **Can it produce client-ready reports?** Yes — scheduled cross-platform rollups delivered to a channel, a thread, or an inbox, plus on-demand answers like "what's driving CAC up this week?" **Can it write the ad creative too?** It drafts copy variations across distinct angles — pain-led, outcome-led, proof-led — in your brand voice, for your approval before anything runs. Creative claims are never invented: every factual statement in an ad comes from you or a verifiable source. **Which metrics should an AI ads agent optimize?** The ones tied to your objective, not the platform's: cost per qualified conversation for lead gen, blended CAC for e-commerce, replies for early-stage learning. A good agent tells you when your sample is too small to optimize anything honestly. --- # AI agent for customer success — renewals never surprised > An AI agent for customer success runs the coverage layer of CS: watching account health signals across your tools, flagging the accounts going quiet before renewal does, prepping QBRs from real usage and history, chasing follow-ups, and keeping the CRM true. Arlo does this from Slack, WhatsApp, or a text thread — with customer-facing sends approval-gated and every action audit-trailed. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-agent-for-customer-success Churn is rarely a surprise to the data — only to the team. The account that stopped logging in six weeks ago, the champion who changed jobs, the support thread that ended unresolved, the renewal in 45 days with no touch in 90: the signals sat in four different tools while everyone was busy with the accounts that were talking. An AI agent for customer success is the fix for exactly that shape of failure: it watches everything, so humans can go deep on the accounts that need them. ## What Arlo covers - **Health signals, unified** — usage drops, support sentiment, invoice friction, champion changes picked up from email and LinkedIn-shaped signals in your tools. Each account gets a watching brief; you get pinged when a pattern turns, not a dashboard to remember. - **Renewal runway** — every renewal gets a countdown with staged touches: the 90-day check-in drafted, the 45-day value recap assembled, the 14-day silence escalated to the owner as a genuine risk. - **QBR prep, assembled** — usage trends, delivered value, open items, and the plan — pulled from your actual tools into your deck or doc format, for the CSM to edit judgment into rather than build from scratch. - **The quiet-account chase** — check-ins drafted for accounts that stopped responding, and when email goes nowhere, [a real phone call](/compare/ai-assistants-that-can-make-phone-calls) — then the outcome logged where the team can see it. - **CRM truth as a side effect** — every thread, touch and call outcome you ask it to log lands on the record, so handoffs and forecasts run on reality. The [sales-side agent](/use-cases/ai-sales-agent) works the same muscle pre-close. - **Voice-of-customer rollups** — what accounts are actually asking for, aggregated from threads and tickets into a monthly note product will actually read. ## The line CS work must respect Customer trust is the asset, so the defaults are conservative: **anything customer-facing waits on the CSM's approval** — every check-in, recap, and call — until specific routine categories are deliberately loosened. Escalations (the angry thread, the churn signal on a top account) go to a human immediately with context, never handled solo. Everything lands in the [audit trail](/glossary/ai-agent-audit-trail), which is also what makes handoffs between CSMs stop losing history. ## Frequently asked questions **What does an AI agent for customer success do?** Watches account health across tools, flags risk early, preps QBRs, runs renewal touch sequences for approval, chases quiet accounts, and keeps the CRM true. **Will it talk to my customers directly?** Only with approval. It drafts; your CSM signs off — and high-stakes moments always escalate to a human rather than being handled autonomously. **Can it really predict churn?** It surfaces the patterns that precede churn — usage decline, silence, support friction, champion loss — early enough to act. Prediction is pattern-flagging plus human judgment, honestly divided. **Does it work with our CS stack?** Arlo connects to CRMs, support desks, email, and analytics among 3,000+ integrations, with a secure browser session for tools that have no API. **What does it cost?** Arlo starts free — $100 in credits for connecting Slack — then sells prepaid credits: 400 credits per $1 at every amount, on monthly plans from $20, good for 12 months — against the cost of one renewal that walked out silently. --- # AI agent for data entry — even in tools with no API > An AI agent for data entry moves information between systems like a careful person: reading source documents, filling forms, updating records, and keeping spreadsheets in sync. Arlo works across 3,000+ integrations — and for software with no API, it logs in through a secure browser session and uses the interface directly. Changes are reviewable, batch-approved where it matters, audit-trailed. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/use-cases/ai-agent-for-data-entry Data entry survived two decades of automation promises for one stubborn reason: the systems that need connecting weren't built to connect. The portal your supplier insists on. The government form. The legacy tool that runs the industry and last shipped an update in 2019. Integration platforms cover the modern half of the stack; a human covers the rest, one copy-paste at a time — usually a human whose actual job is something else. An AI agent closes that gap from the human side: instead of waiting for an API that will never ship, it uses the software the way a person does. ## What Arlo does - **Between-system transfer** — read the invoice PDF, the email thread, or the export; enter it into the CRM, the tracker, the books. The judgment calls a person would make (dedupe, format, "which field does this go in") follow rules you set once. - **No-API tools, handled** — Arlo logs in once through a secure browser session and clicks, reads, and fills forms directly. The supplier portal, the legacy ERP screen, the government filing — reachable without anyone's afternoon. - **Spreadsheets kept true** — the ops sheet, the pipeline tracker, the inventory count: updated as a side effect of the work happening in threads and tools, not reconstructed weekly by hand. - **Batch jobs with review** — "enter these 80 conference leads" comes back as a completed batch with a summary of exceptions ("3 duplicates skipped, 2 missing emails flagged") rather than 80 silent writes. - **Requested from a thread** — forward the attachment in iMessage or Slack with "get these into the system" and the confirmation lands back in the same thread. ## Accuracy, verifiable Human data entry fails by fatigue; automation fails by confidently doing the wrong thing at scale. Arlo's design attacks both: extraction is grounded in the actual source document, ambiguous rows get **flagged instead of guessed**, bulk writes can wait on [batch approval](/glossary/human-in-the-loop), and every change is recorded in a full [audit trail](/glossary/ai-agent-audit-trail) — so "what changed, from what source, approved by whom" always has an answer. Spot-check the first batches, then loosen review as the error rate proves itself. ## Frequently asked questions **Can an AI agent do data entry in software without an API?** Yes — Arlo logs in through a secure browser session and operates the interface like a person: reading screens, filling forms, submitting records. No integration required. **How accurate is it?** Entries are grounded in the source document, ambiguous data is flagged for a human instead of guessed, and batch summaries surface exceptions. You review early batches and loosen oversight as accuracy proves out. **What kinds of data entry can it handle?** Invoices and orders into finance tools, leads into CRMs, listings and inventory updates, form filings, spreadsheet syncs — anything with a readable source and a destination system. **Does it see my passwords?** No — Arlo authenticates once through a secure browser session without exposing credentials, and touches only the systems you've granted. **Is my data used to train AI models?** Arlo is built for governed business use with scoped access and audit trails; see the privacy policy for the current data-handling commitments. --- # AI agent for lead generation — research, enrich, follow up > An AI agent for lead generation works the gap between "we need pipeline" and a booked meeting: researching accounts that fit your ICP, enriching and qualifying inbound signups, drafting personalized first touches, routing hot leads to the right rep instantly, and chasing every lead before it cools. Arlo runs this from your Slack or text thread, with approval gates on all outreach. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/use-cases/ai-agent-for-lead-generation Lead generation fails in the gaps, not the tools. The signup that waited two days for a reply. The conference list nobody worked. The "hot" lead that was actually a student, and the student-looking one that was actually a VP with a personal email. Teams buy databases and sequencers and still leak pipeline, because the connective work — research, enrichment, qualification, timing — is manual, boring, and always loses to whatever's on fire. That connective work is exactly what an [AI agent](/glossary/ai-employee) absorbs. Not another data source; the worker between your data sources. ## What Arlo runs - **Account research on demand** — "find 30 companies that look like our best customers" returns a researched list with the why, not just a filter export: hiring signals, stack hints, recent news, the right person to reach. - **Inbound enrichment and qualification** — every signup gets researched the minute it lands: company, role, fit against your ICP. The VP with a Gmail address gets caught; the fake gets filtered. Hot ones ping the right rep in Slack immediately — [speed to lead](/use-cases/ai-assistant-for-sales) stops being aspirational. - **Personalized first touches, approval-gated** — drafted from the research, in your voice, referencing something true. You approve before anything sends; quality never outruns judgment. - **The chase** — leads cool on a clock. Arlo tracks every open thread, drafts the follow-up when the window hits, and [picks up the phone](/compare/ai-assistants-that-can-make-phone-calls) when email goes nowhere. - **List hygiene as a side effect** — the CRM reflects reality because updating it is part of every action, not a Friday chore. - **Reporting that answers the question** — "what did lead gen produce this month, by source?" asked in your thread, answered from your actual tools. ## The honest limits An agent doesn't fix a bad ICP, a weak offer, or a product nobody wants — garbage in, faster garbage out. And bulk cold spam remains a reputational tax no matter who writes it; Arlo's approval-first design is a feature here, not a speed bump. Where the agent shines is conversion of interest you've already earned and disciplined pursuit of accounts you've deliberately chosen — the two places human teams leak the most. ## Frequently asked questions **What does an AI agent for lead generation actually do?** Research target accounts, enrich and qualify inbound leads, draft personalized outreach for approval, route hot leads instantly, and chase every open thread on schedule — with your CRM updated throughout. **Can it find leads by itself?** It researches companies and people matching the profile you define, using your tools and the open web. You set the ICP; it does the digging and shows its reasoning. **Will it send cold email automatically?** No — every outreach draft waits for your approval by default. Arlo is built for quality-gated outreach, not bulk sends. **How fast does it respond to inbound?** Enrichment and routing happen the moment a lead lands; the rep gets pinged in Slack or by text with context. The first-touch draft is ready before a human would have opened the notification. **What tools does it work with?** CRMs, email, enrichment sources, and forms among 3,000+ integrations — plus a secure browser session for tools with no API. --- # AI agent for project management — status without the chasing > An AI agent for project management does the coordination work around projects: keeping the tracker true to reality, chasing owners before deadlines slip, assembling the status update nobody wants to write, and surfacing blockers early. Arlo does this across Linear, Notion, Slack, and email — from the thread you're already in, with approvals on anything outbound. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-agent-for-project-management Project management tools don't manage projects — people do, and people spend most of that effort on a loop of asking: what's the status, who owns this, is the deadline real, why is the tracker three days behind reality. The tracker is a mirror that someone has to keep polishing. An AI agent for project management takes the polishing — and the chasing, the assembling, and the reporting — so the humans keep only the deciding. ## What Arlo runs - **A tracker that tells the truth** — when work happens in Slack threads, PRs, and docs, Arlo reflects it into [Linear](/use-cases/ai-assistant-for-linear) or your tracker of choice: statuses, links, owners. The board stops being archaeology. - **Deadline chasing, depersonalized** — "the launch checklist has three items due Friday with no movement" arrives as a nudge to each owner, politely, from Arlo. Nobody has to be the nag; the nag is infrastructure. - **The status update, assembled** — weekly or per-milestone: what shipped, what slipped, what's blocked, what's next — drafted from the actual tracker and threads, for your one-tap approval before it goes to stakeholders. - **Standup prep** — each morning, per person or per team: yesterday's movement, today's plan, blockers surfaced from the tools rather than memory. - **Blocker surfacing** — the dependency that quietly stalled, the review sitting unclaimed for four days, the vendor who never replied. Flagged when it's a nudge, not a crisis — and [chased by phone](/use-cases/ai-assistant-you-can-text-and-call) when a call is what it takes. - **Meeting-to-tracker closure** — decisions and action items from meetings land as tracked items with owners, instead of dying in notes. ## Where it lives In the channels where the project actually happens. Ask in the project's Slack channel ("what's at risk for the 15th?"), text from your phone between meetings, or let the standing briefs arrive on schedule. Arlo carries one memory across [Notion](/use-cases/ai-assistant-for-notion), Linear, email, and chat — so "the migration project" means the same thing everywhere. ## Trust and the outbound line Internal reads and tracker updates are the safe zone, and Arlo runs them freely once you've granted access. Anything that leaves the team — the stakeholder update, the vendor chase, the customer-facing date — waits on [your approval](/glossary/human-in-the-loop) by default, and every action lands in the [audit trail](/glossary/ai-agent-audit-trail). You get relentless coordination without surprise communication. ## Frequently asked questions **What does an AI agent for project management actually do?** Keeps the tracker true, chases owners before deadlines slip, assembles status updates, preps standups, and surfaces blockers — across Linear, Notion, Slack, and email. **Does it replace a project manager?** It replaces the chasing-and-assembling layer of the job. Scope calls, prioritization, and stakeholder judgment stay human — with far better information under them. **Which tools does it work with?** Linear, Notion, Slack, Gmail and 3,000+ others — plus a secure browser session for tools with no API. **Will it message stakeholders on its own?** No — outbound updates are drafted for your approval by default. Internal tracker updates and nudges run under the access you granted. **What does it cost?** Arlo starts free — $100 in credits for connecting Slack — then runs on monthly plans from $20: 400 credits per $1 at every amount, good for 12 months — so what you spend tracks the coordination load rather than the calendar. --- # AI agent for social media — drafts, scheduling, and listening > An AI agent for social media runs the account-management grind: drafting posts in your brand voice, keeping a content calendar full, watching mentions and competitor moves, flagging the replies worth answering, and reporting what actually performed. Arlo does this from your Slack or text thread, and nothing publishes without your approval — brand risk stays gated while the volume work disappears. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/use-cases/ai-agent-for-social-media Social media management is a volume job with a judgment problem. The volume: posts to draft, a calendar to fill, mentions to monitor, competitors to watch, metrics to report. The judgment: one bad post is a screenshot forever. Most tools solve volume by removing judgment — auto-posting queues, AI captions nobody reviewed. The right shape is the opposite: an agent that does all the volume work and routes every public action through [your approval](/glossary/human-in-the-loop). ## What Arlo handles - **Drafts in your brand voice** — posts, threads, captions, variations per platform. Arlo remembers your voice, your topics, and which past posts worked; you edit taste, not grammar. - **A calendar that stays full** — it proposes a week of content from your pipeline (launches, blog posts, customer wins), you approve the batch in one pass, and gaps get flagged before they become silence. - **Mention and reply triage** — the support question, the angry thread, the influencer compliment: surfaced in your channel with a drafted response for each, ranked by what deserves attention. - **Competitor and trend watch** — when a competitor's post takes off or your space starts arguing about something, you get a short note with a take — while it's still relevant. - **Performance, translated** — not a dashboard export: "carousel posts are outperforming links 3:1 this month; here's next week's plan weighted accordingly." - **From the thread you live in** — brief, approve, and read reports in iMessage or Slack. The agent goes to the platforms so you don't have to. ## Where the approval gate earns its keep Public posts are irreversible in the way that matters — deletion doesn't beat screenshots. So Arlo's default is absolute: **nothing publishes without sign-off**. In practice approval is a one-tap batch each morning, which costs a minute and buys you what auto-posting tools can't: certainty that a human saw it first. Every publish, edit, and reply lands in the [audit trail](/glossary/ai-agent-audit-trail) — useful the day someone asks "who approved that?" Social is usually one lane of a broader [AI marketing agent](/use-cases/ai-marketing-agent) setup — same colleague, same gates. ## Frequently asked questions **What can an AI agent do for social media?** Draft posts in your voice, keep the calendar full, triage mentions and replies, watch competitors, and report performance in plain language — with every public action approval-gated. **Will it post without my approval?** No. Publishing waits on your sign-off by default, typically a one-tap batch review. That gate is deliberate and stays on. **Which platforms does it support?** Arlo connects to major social platforms among 3,000+ integrations, and for tools without an API it can work the dashboard through a secure browser session. **Can it handle replies and DMs?** It triages them — surfacing what matters with drafted responses — and sends only after your approval. The angry-customer thread always escalates to you. **Is this a replacement for a social media manager?** It's the execution half of one. Strategy, taste, and community judgment stay human; the drafting, scheduling, monitoring, and reporting grind goes to the agent. --- # AI assistant for agencies — every client, one colleague > An AI assistant for agencies runs multi-client operations at team scale — keeping each client's comms, context, and data separate while a 5–20 person team shares governed access. Arlo remembers each account, drafts client updates for your approval, logs work across client tools, and keeps a per-client audit trail. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-agencies Agencies don't run on one client's problems. They run on twelve, all at once, held apart by whoever's paying attention that day. The status update to the retail account, the dev handoff for the fintech build, the PR client who wants a recap by Friday — the work is fine. The overhead of keeping it all straight, across a team, without the wires crossing, is what wears you down. An AI assistant for agencies takes that layer off the team. [Arlo](/) runs multi-client operations across iMessage and SMS, Slack, WhatsApp, and live phone calls, keeps each account separate, and routes anything client-facing through your approval first. This is not the solo strategy setup — that's our [AI assistant for consultants](/use-cases/ai-assistant-for-consultants). An agency is multi-client and multi-person: five to twenty people sharing one operation, where one client's data must never surface in another's thread, and every account needs its own accountable record. ## Client isolation by default The fastest way to lose a client is to leak the wrong context into their conversation. Arlo holds per-client memory that stays walled off: the retail account's roadmap never bleeds into the fintech thread, and a draft for one client is built only from that client's context. Each account has its own scope, its own connected tools, and its own history, so the assistant reasons inside one client's world at a time. That isolation is what makes shared access safe. A 5–20 person team can all delegate to the same assistant without anyone worrying that an account manager's request pulls up another team's client data. ## Per-client context and memory Running fifteen engagements means fifteen mental models nobody can hold at once. Arlo keeps persistent memory per account: the scope, the stakeholders, the decisions made, the deadlines, and where each thread actually stands. Every morning it posts one briefing of what moved overnight across every client — so the team walks in already knowing which accounts need attention. When someone says "the campaign brief we sent them," Arlo already knows the client, the document, and what was last agreed. Nobody re-briefs it or scrolls back through three weeks of a Slack channel before replying. ## Client updates drafted for approval Client-facing comms are where trust lives, and one wrong send to the wrong account costs more than it saves. So nothing reaches a client unreviewed. Arlo drafts the update, the recap, the status note — and holds it for a named reviewer before it goes out. - A client asks "where are we on the launch?" and Arlo assembles the status from Linear, Notion, and GitHub, then drafts the reply for an account lead to approve. - A weekly client roll-up gets drafted per account, ready for the owner to sign off and send. - A rescheduling request comes in over text; Arlo proposes times against the team calendar and confirms once approved. You're still the name on every message. Arlo does the work; your team signs off. ## Logging work across every client tool The follow-through is the part that slips. Arlo connects to Gmail, Notion, Linear, GitHub, and 3,000+ other tools, so work lands in the systems each client runs on. It opens the ticket, updates the record, logs the hours, files the recap — scoped to the right account every time. Some client tools have no API: the legacy portal, the homegrown dashboard. Arlo can log into those once in a secure browser session and read or update them the way a person would, under the same approvals and audit trail as everything else. ## Governed access across the team When five to twenty people share an assistant, governance can't be optional. Every connection resolves through policy before a run starts, every send or write waits for the reviewer you choose, and every account keeps its own [audit trail](/glossary/ai-agent-audit-trail) — every tool call, source, and approval on the record, per client. | Action | What happens | | --- | --- | | Draft a client update | Held for a named reviewer before sending | | Update a client's tool | Routed through policy, scoped to that account | | Onboard a new client's toolset | Connected once, scoped, then available to the team | | Every tool call and source | Logged in that client's audit trail | That per-client record is what makes agency work accountable: if a client asks what happened on their account, there's a full trace to show them. Arlo runs on monthly plans from $20 — 400 credits per $1, and they last 12 months; admin policy, roles, SSO, and audit export are available on request for firms that need to govern the whole team. ## Onboarding a new client Winning an account shouldn't mean a week of setup. Point Arlo at the new client's tools — their Slack, their project tracker, their inbox — each connection scoped by policy up front. From there the team delegates to it the same way they'd loop in a coworker, and the new account gets its own isolated memory and audit trail from day one. It's the same one-colleague model behind our [AI assistant in Slack](/use-cases/ai-assistant-in-slack): one [AI colleague](/glossary/ai-colleague), every surface, governed by default. Explore other setups in [use cases](/use-cases). ## Frequently asked questions **What does an AI assistant do for an agency?** It runs multi-client operations at team scale — keeping each client's communications, context, and data separate while a 5–20 person team shares governed access. It remembers each account, drafts client updates for approval, logs work across client tools, and keeps a per-client audit trail. **How does an AI assistant keep client data separate?** Client isolation is the default rather than a setting. Each engagement carries its own context and memory, and connections resolve through policy before a run, so work on one account doesn't leak into another. **Will it send things to clients without me seeing them first?** No. Client-facing sends are held for a named reviewer. Arlo drafts the update, and it waits — nothing reaches a client until a person on your team approves it. See [approval before sending](/use-cases/ai-assistant-with-approval-before-sending). **Can the whole agency team use one assistant?** Yes. Arlo is added once and the team reaches it in [Slack](/use-cases/ai-assistant-in-slack), WhatsApp, or iMessage — no per-person app to install. Access to each client's tools is governed rather than shared wholesale. **Can it prove what it did on a client account?** Yes. Every tool call, source, and approval lands in a per-client [audit trail](/glossary/ai-agent-audit-trail) you can read back — which is what makes it defensible when a client asks who did what, and when. **What happens when we onboard a new client?** You connect that client's tools and the assistant starts building context for the account from the first conversation, without mixing it into an existing engagement. ## Try Arlo Run your whole client roster through one governed AI colleague that keeps every account separate, remembers each engagement, and holds every client-facing send for approval. [Get started](/signup). --- # AI Assistant for Compliance and Governance > A governed AI assistant for compliance gates every action behind human approval and records it in a full audit trail — every tool call, source, and sign-off on the record. Arlo resolves policy per run, holds writes and sends for a named reviewer, and exports the trail, so accountable teams can safely let AI act. Updated: 2026-07-13 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-compliance The thing that stops accountable teams from letting AI act isn't capability — it's not knowing what it will do, or being able to prove what it did. A governed AI assistant answers both. Every action is gated by a human approval, and every action is recorded in a full audit trail. That combination — approval before, record after — is what turns "we can't let AI touch that" into "we can, and here's the proof." [Arlo](/) is built this way by default: nothing that leaves a mark goes out until a person signs off, and nothing happens without a trace. This is governance as a function, not a vertical. Whether you're in a regulated industry or just an accountability-heavy org where a wrong send is expensive, the shape is the same: a human check on consequential actions and a defensible record of every one. ## Every action gated by approval The moment an assistant can send, write, or spend, the question stops being "is it smart?" and becomes "what happens when it's wrong?" Arlo's answer is that consequential actions don't complete on their own. Anything that commits you pauses for the reviewer you name: - **Sends** — an email, a message, a reply on your behalf. - **Writes** — updating a record, editing a page, opening or closing a ticket. - **Spend and bookings** — anything that costs money or commits an external party. Read-only work — research, summaries, drafts you haven't asked it to send — flows without interruption. The gate lands only where the risk is, so review is a checkpoint, not a bottleneck. This is the same [approval-before-sending](/use-cases/ai-assistant-with-approval-before-sending) model, made the default for governed work. ## Policy resolves per run Governance isn't a switch you flip after the fact — it shapes the run from the start. Before Arlo does anything, every tool connection resolves through policy: what it may touch and how far it can go. So the boundaries are decided up front, not discovered after an action already fired. 1. **Policy first.** Each connection resolves through policy before the run begins. 2. **Act up to the line.** Arlo reasons, drafts, and does read-only work on its own. 3. **Pause at the gate.** Any send, write, or spend surfaces for the named reviewer. 4. **Record it.** The action, its sources, and the approval all land in the trail. ## A full, exportable audit trail You can't govern what you can't see. Every run produces an [audit trail](/glossary/ai-agent-audit-trail): every tool call, every source it read, every human approval, tied back to the original request. Nothing is summarized away — the record is the actual sequence of what happened and why. | | Ungoverned AI assistant | Arlo | | --- | --- | --- | | Sends and writes by default | Acts on its own | Waits for approval | | Who reviews risky actions | No one | A named reviewer | | Policy on tools it can touch | Rare | Resolved before every run | | Record of what happened | None | Full audit trail | | Export for review or audit | No | Yes, on request | That trail is what makes AI usable where accountability matters. When someone asks what the assistant did — a manager, a client, an auditor — there's a complete, exportable record to show, not a black box. ## Role-based access for accountable teams Governance at team scale means the right people hold the right controls. The baseline runs by default for every workspace, from the first dollar — per-run policy checks, approvals on consequential actions, and the full trace. For organisations that need them on top of that, admin policy, role-based access, SSO, and audit export are available on request, so an administrator sets what the assistant may do, roles determine who can approve what, and the trail exports for review. ## Governance that supports compliance workflows Rules like the EU AI Act phasing in through 2026 increasingly expect a human check on consequential automated actions and a record that one was in place. Arlo's approval gate plus audit trail is exactly that: a human in the loop before anything commits, and a defensible record after. To be clear about what this is: these are governance controls that support your compliance workflows — a human-in-the-loop gate and a complete audit trail — not a claim of specific certifications. Arlo gives your team the mechanism to keep a person on every consequential action and to prove it happened. How that maps to your obligations is yours to define; Arlo makes it enforceable and recordable. This is the [AI colleague](/glossary/ai-colleague) model with governance built in — across [Slack](/use-cases/ai-assistant-in-slack), iMessage and SMS, WhatsApp, and live phone calls, working through 3,000+ tools and no-API browser use, all under the same gate and trail. See how it fits a single workday as an [AI executive assistant](/glossary/ai-executive-assistant), or browse more [use cases](/use-cases). ## Frequently asked questions **What makes an AI assistant "governed"?** Three mechanisms working together: policy resolves each tool connection before a run begins, consequential actions pause for a named reviewer instead of completing on their own, and every tool call, source, and approval lands in an exportable audit trail tied back to the original request. **Which actions require approval and which don't?** Sends, writes, and spend require it — an email or message on your behalf, updating a record or ticket, anything that costs money or commits an external party. Read-only work flows without interruption: research, summaries, and drafts you haven't asked it to send. The gate lands where the risk is, so review is a checkpoint rather than a bottleneck. **What happens if an approval is never answered?** It expires, and an expired approval cannot execute. The request fails closed rather than defaulting to send — so an unattended gate is a blocked action, not a silent one. **Can a reviewer approve something once and not be asked again?** Yes, within a boundary you set. An approval can carry a standing grant scoped to a specific tool and action, with a required expiry date. The grant covers that scope until it expires or is revoked; a grant with no future expiry is rejected outright. **Does a reviewer approve or edit the action?** Approve or deny. The reviewer sees the tool, the account, and a preview of what would be sent, then decides. There is no edit-in-place step — if the draft is wrong, you deny it and ask for a different one. **Is Arlo compliant with the EU AI Act or SOC 2?** These are governance controls that support your compliance workflows — a human-in-the-loop gate and a complete audit trail — not a claim of specific certifications. Rules like the EU AI Act phasing in through 2026 increasingly expect a human check on consequential automated actions and a record that one was in place. Arlo gives you the mechanism and the record; how that maps to your obligations is yours to define. **Can we export the audit trail?** Yes. Audit export, along with admin policy, role-based access, and SSO, is available on request. Governance itself — per-run policy checks, approvals, and the full trace — runs by default for everyone, from the first dollar. ## Try Arlo Give an assistant real reach without giving up control. Every action gated by approval, every action on the record. [Try Arlo](/signup) and let AI act where you couldn't before. --- # AI Assistant for Consultants > An AI assistant for consultants runs client communication, scheduling, follow-ups, research, and deliverable drafting across iMessage, Slack, WhatsApp, and phone calls. It remembers each client's context, handles time-sensitive coordination, keeps client tools current, and routes every send through your approval first. Updated: 2026-06-19 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-consultants The billable work isn't what burns you out. It's everything around it: the reschedules, the "just circling back" emails, the prep before a strategy call, the status update buried in a client's Slack. An AI assistant for consultants takes that layer off your plate. [Arlo](/) handles client comms, scheduling, follow-ups, research, and first-draft deliverables across iMessage and SMS, Slack, WhatsApp, and live phone calls, and nothing leaves without your sign-off. If you run a solo practice or a small firm, you already know the bottleneck isn't how hard you work. It's how thin your attention gets spread across clients, channels, and deadlines. Arlo lives inside the tools you already use, so the coordination stops piling onto you. ## One assistant across every client channel Clients reach you however they want. One texts your cell. One only talks in their company Slack. Another only replies on WhatsApp. A prospect calls while you're in a meeting. Normally you're the glue holding all that context together. Arlo is a single assistant across iMessage and SMS, Slack, WhatsApp, and phone, so the glue isn't you anymore. - A client texts "can we move Thursday?" and Arlo checks your calendar, proposes times, and confirms. - A client Slack thread needs a status reply, so Arlo drafts one for you to approve. - A call comes in while you're heads-down. Arlo can pick it up, capture the ask, and brief you after. It's the same one-colleague model behind our [AI assistant for founders](/use-cases/ai-assistant-for-founders): one memory, every surface. ## Memory of each client's context Running ten engagements at once means keeping ten mental models from bleeding into each other. That's the part that quietly eats your week. Arlo holds persistent memory per client: scope, the people who matter, decisions made, deadlines, and where each thread actually stands. Every morning it sends a briefing of what moved overnight across all of them. So when a client says "the migration plan we discussed," Arlo already knows the engagement, the document, and what you last agreed to. You're not re-briefing it or scrolling back through three weeks of messages before you reply. ## Client comms, scheduling, and follow-ups Logistics is most of the overhead in this job, and it's the easiest thing to drop. Arlo runs the loops that keep slipping: - **Scheduling.** Proposes and confirms times against your calendar, reschedules around conflicts, sends the reminders. - **Follow-ups.** Drafts the recap after a call, chases open items, nudges a client on a pending decision before it goes cold. - **Intake.** Turns a request that came in over text or a call into a tracked task instead of a message you'll forget by Friday. Arlo connects to Gmail, Notion, Linear, GitHub, and 3,000+ other tools, so the follow-through actually lands in the systems you run your practice on. ## Research, prep, and drafting deliverables The prep that eats your mornings is exactly the part Arlo is good at. Have it research a client's market before a strategy session, pull recent moves from a competitor, or assemble the background for a proposal. It'll draft the meeting notes, the summary, the first pass at a deliverable. You start from an editable draft instead of a blank page. Some client tools have no API. The legacy portals, the homegrown dashboards, the thing their ops team built in 2014. Arlo can log into those once in a secure browser session and read or update them the way a person would. ## Governed sends keep client trust intact Consulting runs on trust, and one wrong message to the wrong client costs more than it saves. So nothing goes out unreviewed. Governance is on by default: | Action | What happens | | --- | --- | | Draft a client email | Waits for your approval before sending | | Update a client's tool | Routed through policy, held for a reviewer you choose | | Make a call or send a text | You approve before it reaches the client | | Every tool call and source | Kept in a full audit trace | You're still the name on every message. Arlo does the work and you sign off. Arlo runs on monthly plans from $20 — 400 credits per $1, good for 12 months; admin policy, roles, SSO, and audit export are available on request for firms that need them. ## How it compares to doing it manually | Task | Manual | With Arlo | | --- | --- | --- | | Cross-channel client comms | You context-switch all day | One assistant, one memory | | Tracking each client's status | Spreadsheets and memory | Persistent context plus a morning briefing | | Follow-ups after calls | Often slip | Drafted and chased for you | | Research and prep | Hours per engagement | Pulled together on request | | Sends and updates | Manual, every time | Drafted, you approve | There's related setups for [founders](/use-cases/ai-assistant-for-founders), [sales](/use-cases/ai-assistant-for-sales), and [realtors](/use-cases/ai-assistant-for-realtors), plus the [glossary](/glossary/ai-colleague) and [resources](/resources) if you want to go deeper. Running several clients under one roof rather than solo? The multi-client version — per-client context, governed access, and a separate audit trail per account — is [AI assistant for agencies](/use-cases/ai-assistant-for-agencies). ## Frequently asked questions **What does an AI assistant do for a consultant?** It runs client communication, scheduling, follow-ups, research, and deliverable drafting across iMessage and SMS, Slack, WhatsApp, and phone calls — remembering each client's context, handling time-sensitive coordination, keeping client tools current, and routing every send through your approval first. **How does it keep each client's context straight?** It holds persistent memory per engagement, so it knows the account, the deadline, and what you agreed last month without being re-briefed. Work stays scoped to the client it belongs to. **Will it email my clients directly?** Only after you approve. Arlo drafts the message and holds it for you — nothing client-facing sends on its own. That's what makes it safe to point at real client comms; see [approval before sending](/use-cases/ai-assistant-with-approval-before-sending). **Can it work in whatever channel each client uses?** Largely, yes. Arlo covers iMessage and SMS, [Slack](/use-cases/ai-assistant-in-slack), [WhatsApp](/whatsapp), and live phone calls, with one memory across all of them — so a thread that starts as a text can finish on a call without re-explaining. **Can it help with research and deliverables, not just admin?** Yes — sourced research, call prep, and first drafts of deliverables. Read-only work like research and drafting runs without an approval gate, so it isn't slowed down by review. **Is it worth it for a solo consultant?** $50 of credits is the smallest amount Arlo sells and it lasts most solo practices months, so the honest test is cheap. It earns its keep when the coordination load — scheduling, follow-ups, keeping client tools current — starts eating the hours you'd rather bill. ## Try Arlo Run your whole client roster through one governed AI colleague that remembers every engagement and works where your clients already are. [Get started](/signup). --- # AI Assistant for Customer Support: Triage, Draft, Approve > An AI customer support assistant triages inbound tickets, drafts a first reply for a human to approve before it sends, escalates hard threads, and logs every resolution. Arlo does this across Slack, WhatsApp, and iMessage with an approval gate on every send and a full audit trail. Updated: 2026-07-13 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-customer-support Support teams don't need another chatbot standing between customers and answers. They need help writing the reply faster, catching the thread that's about to blow up, and keeping a clean record of what was said and by whom. An AI assistant for customer support does exactly that: it triages what's coming in, drafts the response a human signs off on, and logs the resolution so the next person isn't guessing. Arlo is an [AI colleague](/glossary/ai-colleague) for support work, not an autonomous bot that answers customers on its own. It reads the incoming thread, drafts the first reply, and then stops. Nothing goes to a customer until a human approves it. That approval gate is the whole point: support is where a wrong send does real damage, so Arlo is built to draft, not to send behind your back. ## What an AI support assistant actually does The job isn't "deflect tickets." It's clearing the busywork around each one so agents spend their time on judgment, not typing: - **Triages inbound** by reading the message, tagging urgency, and pulling the relevant history so an agent opens the thread already oriented. - **Drafts a first reply** grounded in your docs and past resolutions, ready for a human to edit and approve. - **Escalates** a thread that's out of scope or getting heated, with a summary of what's happened so far. - **Summarizes a ticket** into a few lines when someone needs to hand it off or loop in a manager. - **Logs the resolution** so the outcome is recorded and searchable, not buried in a channel. ## Concrete support scenarios What this looks like in a real workspace: - A customer message lands in `#support`. Arlo summarizes the issue, checks your runbook, and posts a draft reply. An agent reads it, tweaks one line, and approves. It sends. - A thread turns angry. Arlo flags it, writes a two-line recap of the history, and escalates to the on-call lead so nobody's re-reading twenty messages under pressure. - A teammate asks "what's the status on the Acme refund thread?" Arlo pulls the thread, the account context, and the last action taken, and hands back a tight summary. - End of a hard case: Arlo logs the resolution — what the customer wanted, what was done, what to watch for — so the record is there when they write back next month. ## Approvals and the audit trail Governance is what separates a support assistant you can trust from one you can't. Arlo holds every send and every tool write [for a reviewer you choose](/use-cases/ai-assistant-with-approval-before-sending) — no reply reaches a customer, no record gets changed, until a human confirms it. And every draft, source, approval, and action lands in a full [audit trail](/glossary/ai-agent-audit-trail), so when compliance or a manager asks "who said what, and why," the answer is one click away instead of a reconstruction. | Capability | In-app helpdesk AI | AI support assistant (Arlo) | | --- | --- | --- | | Draft replies from your docs | Yes | Yes | | Send only after human approval | Often auto-sends | Yes, every send | | Works across Slack and iMessage | No, one product | Yes | | Context survives a channel switch | No | Yes, with memory | | Full audit trail on every action | Partial | Yes | | Acts in tools with no API (browser) | No | Yes, logs in once and clicks | ## How Arlo differs from Zendesk or Intercom AI Be clear-eyed about this. If you live entirely inside one helpdesk, the AI baked into Zendesk or Intercom is right there and tightly wired to that product's ticket model. Arlo isn't trying to replace the helpdesk. It's a governed, cross-channel colleague that works where your team already talks — Slack, WhatsApp, iMessage — and holds every customer-facing send for a human. An in-app bot is designed to answer autonomously inside its own walls. Arlo is designed to draft, wait for approval, and leave an audit trail, which is what teams that care about accountability actually want. ## Context that follows the thread Support conversations don't stay in one place. A ticket starts in [Slack](/use-cases/ai-assistant-in-slack), the lead picks it up by text, and a hard case ends on a [phone call](/use-cases/ai-executive-assistant-that-makes-phone-calls). Arlo is one assistant across iMessage and SMS, Slack, WhatsApp, and live calls, and its memory carries the context across all of them, so nobody loses the thread when it changes channels. A morning briefing rounds up what moved overnight so the team starts the day oriented. ## Tools with and without an API Plenty of support stacks include an internal portal or a legacy system that never shipped an API. Arlo handles those by signing in once through a secure browser session and operating them by hand — reading a record, updating a status, filling a form — under the same approval gate and audit trail as everything else. It connects to 3,000+ tools directly, and where no integration exists, it clicks through like a person would. ## Frequently asked questions **What does an AI customer support assistant do?** It triages inbound tickets, drafts a first reply for a human to approve before it sends, escalates the threads it shouldn't handle alone, and logs every resolution. Arlo does this across Slack, WhatsApp, and iMessage with an approval gate on every send and a full audit trail. **Will an AI assistant reply to customers without a human checking?** Not with Arlo. The draft is prepared and held for a reviewer — the reply reaches the customer only after a person approves it. That's the difference between speeding your team up and letting a model talk to your customers unsupervised. **How is this different from Zendesk or Intercom AI?** Those are AI features inside a helpdesk you already run, and they work well there. Arlo isn't a helpdesk — it's a colleague that works across the channels your team actually talks in, acts in the other tools a resolution touches, and keeps one governed record of what it did. **What happens with a ticket it can't resolve?** It escalates rather than guessing. Hard threads route to a person with the context already gathered, so the human starts from a summarized thread instead of reading from scratch. **Can it act in the tools a resolution actually touches?** Yes — across 3,000+ integrations, and for internal systems with no API it logs in once through a secure browser session and operates them by hand. Consequential writes still wait for approval. **Does it keep a record for quality review?** Yes. Every tool call, source, and approval lands in a [full audit trail](/glossary/ai-agent-audit-trail) tied to the original ticket, so you can reconstruct exactly what was said, what was read, and who signed off. ## Try Arlo Give your support team a colleague that triages inbound, drafts the reply, and logs the resolution — with a human approving every send and a full audit trail behind every move. [Get started](/signup) and connect your first channel in minutes. --- # AI assistant for e-commerce teams — ops without the tab-hopping > An AI assistant for e-commerce runs the operational glue of a store: supplier follow-ups by email or live phone call, customer-service triage and drafts, daily numbers in one morning brief, and records kept current. Arlo does this from iMessage, Slack, or WhatsApp, works no-API portals through a secure browser session, and asks approval before anything ships. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-ecommerce Running a store is a loop of small operational jobs that never end: where's the PO, why is this SKU low, who answers the "where's my order" emails, what did we actually sell yesterday. None of it is strategy. All of it eats the day, across eight tabs — storefront, email, 3PL portal, supplier spreadsheets, ad dashboards, a chat full of pings. An [AI colleague](/glossary/ai-colleague) built for real work collapses that loop into a text thread. You ask; it does; it reports back; and the tools stay updated as a side effect. ## What Arlo runs for a store - **Supplier and 3PL follow-ups** — the chase is the job: the late PO, the missing tracking, the freight quote nobody answered. Arlo chases by email, and when email stalls, [it calls](/compare/ai-assistants-that-can-make-phone-calls) — then posts the outcome in your thread and updates the record. - **Customer email triage** — refund requests, where-is-my-order, the angry one that needs you. Arlo [sorts, drafts in your brand voice, and holds every send for approval](/use-cases/ai-assistant-for-email) until you loosen the gate for routine cases. - **The morning brief** — one message: yesterday's sales, the SKUs that crossed reorder points, shipments that slipped, and the two customer threads worth your attention. Pulled from your tools while you slept. - **Portals with no API** — half of e-commerce ops lives in supplier portals and legacy dashboards no integration covers. Arlo logs in once through a secure browser session and clicks, reads, and files the way a person would. - **[Records as a side effect](/use-cases/ai-agent-for-data-entry)** — inventory notes, supplier threads, ticket statuses, the ops spreadsheet: updated when the work happens, not reconstructed on Friday. ## Where it lives Where your team already talks. Text Arlo from iMessage on the warehouse floor, @mention it in the ops Slack channel, or ask on a call. One persistent memory spans every channel, so "the supplier from last month's shortage" needs no re-explaining — and the [morning brief](/use-cases/ai-assistant-for-small-business) lands in whichever thread you actually read. ## Governance, because this touches customers and money E-commerce automation has a failure mode: the bot that emailed the wrong customer, refunded the wrong order, or spammed a supplier. Arlo's answer is structural — policy checks before every run, **approval gates on anything outbound or irreversible**, and a full [audit trail](/glossary/ai-agent-audit-trail) of every call, email, and change. You extend autonomy category by category as it earns trust, and you can always read exactly what it did. ## Frequently asked questions **What can an AI assistant do for an e-commerce business?** Chase suppliers and 3PLs by email or live phone call, triage and draft customer email, deliver a daily numbers brief, work no-API portals through a secure browser session, and keep records updated — with approvals on anything outbound. **Does it integrate with my store stack?** Arlo connects to 3,000+ tools, and for portals with no API it operates the interface directly through a secure browser session. Check the in-product integrations list for your specific stack. **Can it talk to customers directly?** It drafts customer replies in your brand voice and holds them for approval by default. Many teams keep approval on for customer-facing email permanently and let routine internal work run gated-free. **Can it really call a supplier?** Yes. Arlo places live calls, has the conversation, and keeps the transcript, so you can ask how it went and have it update your records. **What does it cost?** Arlo starts free — connecting Slack grants your workspace $100 in credits — and then runs on a monthly plan with 10% bonus credits. The rate is 400 credits per $1 at every amount, on monthly plans from $20; $50 buys 20,000 credits, roughly 625 messages or 69 researched answers, and every capability including live supplier calls is there from the first dollar. Credits roll over and expire 12 months after purchase. Admin controls, SSO, and audit export are available on request. Weigh that against the hours of ops chasing it absorbs weekly. --- # AI assistant for email — triage, drafts, and follow-through > An AI assistant for email triages your inbox, surfaces the few messages that need you, drafts replies in your voice, chases follow-ups that would otherwise die, and files the rest — with approval before anything sends. Arlo treats email as one channel of a colleague that can also update the CRM, ping Slack, or pick up the phone. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-email Email is where work goes to queue. The median inbox mixes three genuinely important threads with forty that need a two-line reply, a forward, or nothing — and the owner pays a tax on all of them, every day, deciding which is which. An AI assistant for email removes that tax: it reads what arrives, surfaces what actually needs you, drafts what doesn't, and keeps the follow-ups alive that you'd otherwise lose. The distinction that matters in 2026 is between **inbox tools** and **colleagues that handle email**. Inbox tools live inside Gmail or Outlook and stop at the inbox's edge. An [AI colleague](/glossary/ai-colleague) like Arlo treats email as one surface of a bigger job — because half of "email work" isn't email at all. The reply needs a number from the CRM. The intro needs a calendar hold. The vendor thread needs someone to just call. ## What Arlo does with your inbox - **Triage** — reads incoming mail and sorts it into needs-you, needs-a-reply-it-can-draft, and noise. What needs you arrives as a short note in the channel you actually check: iMessage, Slack, or WhatsApp. - **Drafts in your voice** — replies sound like you because Arlo remembers how you write, who the sender is, and what happened in the thread last month. - **Approval before send** — nothing leaves without the reviewer you picked signing off. [Approval-first sending](/use-cases/ai-assistant-with-approval-before-sending) is the default, not a setting. - **Follow-ups that don't die** — the intro you never answered, the invoice at day 30, the trial that went quiet. Arlo tracks open loops and nudges or chases on schedule. - **Beyond the inbox** — the same request can read Gmail, update the CRM, drop a summary in Slack, and schedule the call. One colleague, 3,000+ tools, one [audit trail](/glossary/ai-agent-audit-trail). ## A day with it Morning: one briefing in your texts — what came in overnight, what Arlo already drafted, the two threads that genuinely need you. You reply "send 1 and 3, hold 2." Midday: a customer asks for pricing; Arlo drafts from the current sheet, you approve from your phone. Afternoon: it flags that the Hendricks invoice hit day 30 and asks whether to chase by email or just call — and if the answer is call, [it makes the call](/use-cases/ai-executive-assistant-that-makes-phone-calls) and reports back. None of this required opening the inbox. ## Email tools vs an email colleague | Capability | Inbox-only AI tools | Arlo | |---|---|---| | Triage + labels | ✓ | ✓ | | Drafts in your voice | ✓ | ✓ | | Works from iMessage/Slack | — | ✓ | | Actions outside the inbox (CRM, docs, tickets) | — | ✓ | | Places phone calls when email stalls | — | ✓ | | Approval gates + audit trail | Varies | ✓ On by default | | Persistent memory across channels | — | ✓ | Dedicated inbox tools are good at the inbox; if that's the whole job, they're worth a look — we keep an honest map in [Fyxer alternatives](/compare/fyxer-alternatives). The case for a colleague is everything the inbox touches. ## Getting started Connect Gmail, tell Arlo what matters ("investor threads and anything from these five domains come to me; draft the rest"), and keep approvals on for everything the first week. Trust is earned by watching the drafts. Most people loosen the gates on routine categories within days and keep them forever on anything customer-facing — which is exactly how it should work. When the inbox in question is a shared support queue rather than your own, the triage rules change — see [AI assistant for customer support](/use-cases/ai-assistant-for-customer-support). ## Frequently asked questions **What is an AI assistant for email?** An AI that triages your inbox, drafts replies in your voice, chases follow-ups, and files noise — ideally with approval before anything sends and an audit trail after. Arlo does this as one channel of a broader AI colleague. **Will it send emails without my permission?** Not by default. Arlo holds every send for the reviewer you pick until you deliberately loosen the gate for specific, routine categories. Every send is logged either way. **How is this different from Gmail's built-in AI?** Native suggestions help you write faster inside the inbox. An AI colleague works the inbox for you — from your texts or Slack — and carries tasks beyond it: records, schedules, calls. **Can it handle more than one email account?** Arlo connects to Gmail among 3,000+ tools; for multiple accounts and other providers, check the in-product integrations list for current coverage of your setup. **Does it work with Outlook?** Arlo connects to 3,000+ tools including major email providers; check the integrations list in the product for current coverage of your stack. --- # AI assistant for founders: one colleague across your tools > An AI assistant for founders is an AI colleague that works inside the tools you already use — iMessage, Slack, WhatsApp, and phone calls — to handle email triage, scheduling, investor and customer follow-ups, research, and cross-tool tasks. It keeps persistent memory and asks for approval before anything risky goes out. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-founders An AI assistant for founders is an AI colleague that lives in the tools you already use — iMessage and SMS, Slack, WhatsApp, and live phone calls — and handles the operational work that eats a founder's day. Email triage, scheduling, investor and customer follow-ups, research, tasks that touch four apps at once. It remembers context across conversations and asks before it sends anything that matters. You get most of what an executive assistant gives you, without the salary or the onboarding. ## The founder's real problem Early-stage founders are the bottleneck for everything. You're doing sales, recruiting, product, and fundraising at the same time, and each one lives somewhere different: Gmail, Slack, Notion, Linear, your calendar, a CRM. No single task is the problem. It's the switching, and the follow-ups that quietly die in between. A human EA fixes some of this. But most founders can't justify the cost or the ramp early on, and managing one is its own job. An [AI executive assistant](/glossary/ai-executive-assistant) picks up the repetitive coordination right away, inside the channels you already check, with no new app to learn. ## What it actually handles - **Email triage** — surfaces what needs you, drafts replies in your voice, ignores the rest. - **[Scheduling](/use-cases/ai-scheduling-assistant)** — finds times, sends invites, reschedules without the back-and-forth. - **Investor follow-ups** — tracks every conversation and nudges you before a warm intro goes cold. - **Customer follow-ups** — chases trials, renewals, and open threads. - **Research** — digs into a prospect or a market question and comes back with a sourced answer. - **Cross-tool tasks** — takes one request and runs it across Gmail, Slack, Notion, Linear, and more. ## What it looks like day to day You text from the airport: "Move my 2pm with the design candidate to Thursday and let them know." It checks your calendar, proposes a time, drafts the email, and waits for your nod. After a pitch, you message: "Send the deck and a recap to the three angels I met today." It pulls the contacts, writes a personal note for each, and queues them for approval. In Slack you ask where the Acme renewal stands. It reads the thread, checks the CRM record, and tells you the state and the next step. And every morning you get a briefing of what moved overnight: new replies, calendar shifts, threads sitting on your side. You start the day already caught up instead of spending the first hour finding out what you missed. ## Connected to your whole stack An assistant is only useful if it reaches the tools where the work actually happens. Arlo connects to Slack, Gmail, Notion, Linear, GitHub, and 3,000+ others. For software with no usable API, it logs in once in a secure browser session and clicks, reads, and fills forms like a person, so even the scrappy niche tool nobody integrates with is in reach. | Job | Without an AI assistant | With an AI colleague | | --- | --- | --- | | Email triage | Scan a full inbox by hand | Only what needs you, drafts ready | | Scheduling | Back-and-forth email threads | Times proposed, invites queued | | Follow-ups | Kept in your head, often missed | Tracked and nudged automatically | | Cross-tool tasks | Switch between five apps | One message, handled end to end | | Daily status | Pieced together from each tool | One morning briefing | ## Nothing risky goes out unreviewed Speed without a check is a liability when you're talking to investors and customers. Every connection runs through policy before each turn, and drafts, sends, and tool writes wait for a reviewer you pick. Every tool call, source, and approval lands in a full audit trace. So an investor email never goes out in your name unless you said yes first. ## How it beats a chatbot A chatbot is a window you have to remember to open, with no memory and no way to act. An [AI colleague](/glossary/ai-colleague) works inside your texts and Slack, remembers your context across conversations, takes real governed actions in your software, and logs everything it did. For a founder running on too little time, that's not a small upgrade. ## Frequently asked questions **What does an AI assistant do for a founder?** It handles the operational layer — email triage, scheduling, investor and customer follow-ups, research, and cross-tool tasks — inside iMessage and SMS, Slack, WhatsApp, and phone calls. It keeps persistent memory across all of it and asks for approval before anything risky goes out. **Do I have to learn a new app?** No. You reach it where you already are: text it, message it in Slack or WhatsApp, or call it. There's no dashboard to adopt and no new interface for the team to learn. **How is this different from just using ChatGPT?** A chatbot answers; a colleague acts. Arlo connects to 3,000+ tools and, for software with no API, logs in once through a secure browser session and operates it directly. It also remembers context between conversations rather than starting fresh each time. **What stops it from emailing an investor something wrong?** The approval gate. Every send and tool write pauses for you before it executes, so drafts are prepared automatically but nothing leaves without a decision. Read-only work like research and prep runs uninterrupted. **Is it affordable pre-revenue?** Yes — Arlo starts free with $100 in credits for connecting Slack, and after that it's prepaid credits you spend only as work happens. $50 is the minimum top-up and buys 20,000 credits: roughly 625 messages or 69 researched answers, which for most founders is enough to find out whether it actually removes work. Credits last 12 months, so a slow month costs you nothing. **Can it grow into a team tool later?** Yes — that's the design. The same colleague serves a team as you hire, with governed access per person and an audit trail on every action, so you don't migrate off it when you grow. ## Try Arlo If you're a founder buried in tools and follow-ups, [Arlo](/) is one AI colleague across your whole workspace. Founders often lean on the same colleague for [sales](/use-cases/ai-assistant-for-sales) and [recruiting](/use-cases/ai-assistant-for-recruiters), or compare Arlo with [Lindy](/compare/arlo-vs-lindy) and [Poke](/compare/arlo-vs-poke) first. [Try Arlo](/signup) and hand off the operational work, with approvals so you stay in control. --- # AI assistant for HubSpot — CRM updates from chat > Arlo is an AI assistant for HubSpot that advances deals through stages, updates amounts and close dates, logs calls, emails, and notes to the right records, and answers pipeline questions from wherever you work. You ask in Slack, iMessage, WhatsApp, or on a call; every write waits for your approval and lands in an audit trail. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-hubspot A HubSpot pipeline is only as honest as the last person who bothered to update it. Deals sit in the wrong stage, amounts go stale, and the call you had Tuesday never gets logged — so the forecast the whole team reads on Friday is fiction. Arlo connects to HubSpot as an [AI colleague](/glossary/ai-colleague) that does the CRM hygiene for you: it moves deals, logs activity, updates properties, and answers pipeline questions from a message, so what's in HubSpot actually matches what happened. You don't open the CRM to keep it current. You tell Arlo in [Slack](/use-cases/ai-assistant-in-slack), iMessage, WhatsApp, or on a live call — "move the Acme deal to Contract Sent, $48k, closing end of month" — and it does the work in HubSpot and reports back. ## What Arlo does in HubSpot - **Advances deals through your pipeline.** Move a deal to the next stage, update the amount, push or pull a close date, and flag deals that have gone quiet past your stall threshold. - **Logs activity to the right record.** Record a call, log an email, or drop a timestamped note on the contact, company, or deal — with the follow-up captured so nothing slips. - **Keeps contact and company records clean.** Update job titles, lifecycle stage, deal owner, or any custom property, and reconcile a duplicate before it splits your reporting. For bulk imports from spreadsheets and portals, there's a dedicated [AI agent for data entry](/use-cases/ai-agent-for-data-entry). - **Manages tasks and lists.** Create a follow-up task with a due date, add a contact to an active list or sequence audience, and clear the tasks that are already done. - **Answers forecast and pipeline questions.** "What's weighted pipeline for Q3?", "which deals slipped their close date this week?", or "who hasn't been touched in 14 days?" — Arlo reads HubSpot and hands back a tight answer. ## Real sales-ops scenarios This is what delegating HubSpot to a colleague looks like in practice: - You finish a discovery call. Arlo logs the call to the deal, updates the stage to Qualified, sets next steps, and drafts the recap email — all queued for your approval before anything sends or saves. - Monday standup asks "what's at risk?" Arlo pulls every open deal past its expected close date with no activity in two weeks and posts the shortlist to your sales channel. - A prospect replies after weeks of silence. Arlo reopens the deal, updates the close date, creates a follow-up task, and drafts the reply so you're editing instead of digging. - End of quarter, you need the number cleaned up. Arlo walks the deals stuck in early stages, flags the ones with stale amounts, and proposes the property fixes for you to confirm in one pass. ## Closing the loop across your stack The value isn't just editing one record — it's the chain across tools. Arlo can log a deal in HubSpot, draft the follow-up email in Gmail or Outlook, and post the win to your [Slack](/use-cases/ai-assistant-in-slack) channel, all from one request. After Arlo [makes a phone call](/use-cases/ai-executive-assistant-that-makes-phone-calls) on your behalf, you can ask it to log the outcome to the matching HubSpot deal. HubSpot is one of 3,000+ tools it connects to, and its [persistent memory](/glossary/ai-colleague) means a deal you discussed yesterday still has context today, plus a morning briefing of what moved in the pipeline overnight. ## Governed by default Your CRM is too load-bearing to hand an assistant an unsupervised pen. The HubSpot connection resolves through policy before each run, anything that writes to a contact, company, or deal [waits for a reviewer you choose](/use-cases/ai-assistant-with-approval-before-sending), and every stage change, logged call, and property edit lands in a [full audit trail](/glossary/ai-agent-audit-trail). If Arlo can't reach HubSpot's API for something niche, it signs in once through a secure browser session and operates the interface by hand — under the same approvals and audit trail as everything else. You get a CRM that stays current without giving up control of what changes. See how this fits a whole [sales team](/use-cases/ai-assistant-for-sales), compare [Arlo vs Lindy](/compare/arlo-vs-lindy), or browse more [use cases](/use-cases). ## Frequently asked questions **What can an AI assistant do in HubSpot?** Advance deals through stages, update amounts and close dates, log calls, emails, and notes to the right record, clean up contact and company properties, manage tasks and lists, and answer pipeline and forecast questions — asked from Slack, iMessage, WhatsApp, or a phone call. **Will it change my CRM records without asking?** No. Anything that writes to a contact, company, or deal waits for a reviewer you choose. Reads are free — asking "what's weighted pipeline for Q3?" returns an answer immediately — but every write pauses first. **How do I connect Arlo to HubSpot?** Connect HubSpot once from your integrations, and the connection resolves through policy before each run. It takes a couple of minutes and doesn't require a per-person install for the rest of the team. **What if HubSpot's API doesn't cover what I need?** Arlo signs in once through a secure browser session and operates the interface by hand — under the same approvals and audit trail as an API call. So niche screens that were never exposed to the API are still reachable. **Can it work across HubSpot and my other tools in one request?** Yes. It can log the deal in HubSpot, draft the follow-up in Gmail or Outlook, and post the win to your [Slack](/use-cases/ai-assistant-in-slack) channel from a single ask. After it [makes a call](/use-cases/ai-executive-assistant-that-makes-phone-calls) for you, you can ask it to log the outcome to the matching deal. **Can I see what it changed in HubSpot?** Yes. Every stage change, logged call, and property edit lands in a [full audit trail](/glossary/ai-agent-audit-trail) tied to the request that caused it. ## Try Arlo with HubSpot Put an AI colleague on your HubSpot pipeline that advances deals, logs activity, and answers forecast questions on request — with approvals and an audit trail on every write. Arlo starts free — $100 in credits for connecting Slack — then runs on prepaid credits at 400 per $1. [Get started](/signup) and connect HubSpot in a couple of minutes. --- # AI assistant for Linear — issues and standups from chat > Arlo is an AI assistant for Linear that files and triages issues, sets labels, priority, and assignees, links issues to PRs, and summarizes what shipped this cycle — from wherever you work. You ask in Slack, iMessage, WhatsApp, or on a call; every write waits for your approval and lands in an audit trail. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-linear Linear is where the engineering work is tracked, but the tracking itself is friction — filing the issue, tagging it, dropping it in the right project, and keeping status honest through the cycle. Bugs get reported in a Slack thread and die there. Arlo connects to Linear as an [AI colleague](/glossary/ai-colleague) that handles the overhead: it turns a message into a well-formed issue, triages the backlog, keeps priorities straight, and tells you what actually shipped — without you tabbing over to the board. You describe the work in [Slack](/use-cases/ai-assistant-in-slack), iMessage, WhatsApp, or on a call — "file a P1 for the checkout 500, assign it to me, tag it auth" — and Arlo creates it in Linear and reports back. ## What Arlo does in Linear - **Files issues from context.** Turn a bug report in a Slack thread into an issue with a clear title, reproduction steps, description, labels, and the right project — no copy-paste, no lost detail. - **Triages the backlog.** Set priority, assign an owner, apply labels, move an issue to the right team or project, and close duplicates so triage doesn't rot. - **Plans and runs cycles.** Pull unestimated issues into the current cycle, flag work that won't fit the sprint, surface issues with no assignee, and rebalance load across the team. - **Links issues to code.** Connect an issue to its GitHub PR, update status when the PR merges, cross-reference related issues, and keep the "in progress → in review → done" flow honest instead of manual. - **Summarizes for standups and reviews.** "What shipped this cycle?", "what's blocked?", or "what changed on the launch project since Friday?" — Arlo reads Linear and hands back a tight rollup. ## Real engineering-team scenarios What delegating Linear to a colleague looks like day to day: - A customer bug lands in `#support`. You mention Arlo; it opens a Linear issue with the repro steps from the thread, tags it, sets priority, drops it in the right project, and links back to the conversation — all held for a reviewer to confirm. - Cycle planning Monday. Arlo lists every unestimated issue in the backlog, flags the three that clearly won't fit this sprint's capacity, and proposes what to pull in, so planning is editing a draft instead of staring at the board. - Every weekday at 9am, Arlo posts a standup summary to `#engineering`: what moved to Done since yesterday, what's in review, and which P1s are still open — you approve the format once and it runs on its own. - A PR merges. Arlo moves the linked issue to Done, updates the cycle burndown context, and notes it in the shipped list so the cycle review writes itself. - A stakeholder wants a status update. You ask on a call; Arlo reads the project, tells you which milestones are on track and which are slipping, and offers to file the follow-up issues right there. ## Working across your stack The point isn't one issue — it's the chain. Arlo can open a Linear issue, comment on the matching GitHub PR, and post the update to [Slack](/use-cases/ai-assistant-in-slack) in a single request, because it's one assistant across the whole toolchain of 3,000+ integrations. Its [persistent memory](/glossary/ai-colleague) means an issue you discussed yesterday still has context today, and the morning briefing tells you what changed in your cycles and projects overnight, wherever you read it. ## Governed by default An assistant reshuffling your board unsupervised is a nightmare, so every write runs on a leash. The Linear connection resolves through policy before a run, anything that creates, moves, or edits an issue [waits for a reviewer you choose](/use-cases/ai-assistant-with-approval-before-sending), and every status change, label edit, and assignment lands in a [full audit trail](/glossary/ai-agent-audit-trail). For anything Linear's API doesn't expose, Arlo signs in once through a secure browser session and operates the interface by hand — under the same approvals. You get fast issue hygiene without ceding control of the sprint. Compare [Arlo vs Lindy](/compare/arlo-vs-lindy), or browse more [use cases](/use-cases). If the problem is less about the tracker and more about the chasing around it — status nobody updated, cross-team dependencies, the standup that became a meeting — see [AI agent for project management](/use-cases/ai-agent-for-project-management). ## Frequently asked questions **What can an AI assistant do in Linear?** File issues from context with repro steps and labels, triage the backlog by setting priority, assignee, and project, plan and rebalance cycles, link issues to their GitHub PRs, and summarize what shipped, what's blocked, and what changed — asked from Slack, iMessage, WhatsApp, or a call. **Can it turn a Slack thread into a Linear issue?** Yes — that's one of the most common uses. Mention Arlo in the thread and it opens an issue with the reproduction steps carried over, tags it, sets priority, drops it in the right project, and links back to the conversation, held for a reviewer to confirm. **Will it change issues without approval?** No. Every write — filing, re-prioritizing, reassigning, closing — waits for a reviewer you choose. Reads are immediate, so asking "what's blocked?" answers straight away. **Can it post our standup automatically?** Yes. It can post a daily summary to your engineering channel — what moved to Done, what's in review, which P1s are still open. You approve the format once and it runs on a schedule. **Does it keep Linear in sync with GitHub?** It can connect an issue to its PR and move the issue's status when the PR merges, so the "in progress → in review → done" flow stops being manual. **Can I see everything it changed?** Yes. Every issue write and status change lands in a [full audit trail](/glossary/ai-agent-audit-trail) tied to the request that caused it. ## Try Arlo with Linear Put an AI colleague on your Linear workspace that files issues, triages the backlog, plans cycles, and writes your standup — with approvals and an audit trail on every write. Arlo starts free — $100 in credits for connecting Slack — then runs on prepaid credits at 400 per $1. [Get started](/signup) and connect Linear in a couple of minutes. --- # AI assistant for Notion — meetings into databases > Arlo is an AI assistant for Notion that turns meeting notes into database rows, updates properties, relations, and rollups, drafts and edits pages, and retrieves decisions from your wiki — from wherever you work. You ask in Slack, iMessage, WhatsApp, or on a call; every write waits for your approval and lands in an audit trail. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-notion Notion holds the team's knowledge, but knowledge only helps if someone keeps it filed and can find it later. Meeting notes never make it into the tracker, database properties go half-filled, and "what did we decide about pricing?" turns into ten minutes of scrolling. Arlo connects to Notion as an [AI colleague](/glossary/ai-colleague) that maintains the workspace for you: it files notes as structured rows, updates properties and relations, retrieves decisions from your wiki, and drafts pages — all from wherever you already talk to it. You don't open a tab and click around. You ask Arlo in [Slack](/use-cases/ai-assistant-in-slack), iMessage, WhatsApp, or on a call — "add this call to the CRM database" or "what did we decide in the Q3 planning doc?" — and it acts in Notion and reports back. ## What Arlo does in Notion - **Turns meetings and messages into database rows.** File a meeting's notes as a new row in your meetings or projects database, with the right properties set — owner, status, date, related project — instead of a loose page nobody finds. - **Updates properties, relations, and rollups.** Change a status select, set a date, link a row to its parent project via a relation, and let the rollups recompute — so your dashboards stay accurate without manual reconciliation. - **Retrieves decisions and knowledge.** Ask a question and Arlo pulls the answer from the right page or database in your wiki, with a link to the source — "what's our refund policy?", "who owns the launch?", "what did we decide about pricing?" - **Drafts and edits pages.** Write a spec, a PRD, or a doc from a template, update an existing page, or restructure a wiki section — then hold the change for approval before it saves. - **Manages databases and pipelines.** Add items to a tracker, advance a task through a status board, filter a database for what's overdue, and keep the workspace tidy. ## Real knowledge-management scenarios What delegating Notion to a colleague looks like in practice: - A meeting wraps. Arlo takes the notes, creates a row in your meetings database with attendees, decisions, and action items as properties, links it to the related project, and drafts follow-up tasks — all held for your approval. - Someone new asks "how do we handle refunds?" Arlo searches the wiki, finds the policy page, and answers with a direct quote and a link, instead of the person guessing or interrupting a teammate. - You're spinning up a project. Arlo duplicates your project template, fills in the title, owner, and timeline properties, sets the relation to the parent initiative, and seeds the sub-pages so the workspace is ready instead of blank. - Quarterly cleanup. Arlo scans a database for rows with empty status or stale dates, flags the ones that break your rollups, and proposes the fixes for you to confirm in a single pass. ## Working across your stack The value compounds when Notion isn't the only tool. Arlo can pull a decision from a Notion doc, file the resulting Linear issue, and draft the email — in one request — because it's one assistant across 3,000+ tools. A decision made in [Slack](/use-cases/ai-assistant-in-slack) or on a [phone call](/use-cases/ai-executive-assistant-that-makes-phone-calls) can land in the right Notion page automatically. Its [persistent memory](/glossary/ai-colleague) means context carries across conversations, and the morning briefing surfaces what changed in your workspace overnight, wherever you read it. ## Governed by default Arlo doesn't get free rein over your wiki. The Notion connection resolves through policy before each run, anything that writes to a page or database [waits for a reviewer you choose](/use-cases/ai-assistant-with-approval-before-sending), and every read and write lands in a [full audit trail](/glossary/ai-agent-audit-trail). For the internal tools that never shipped an API, Arlo signs in once through a secure browser session and operates them by hand — under the same approvals. You get a colleague in your workspace without surrendering control of what changes. Browse more [use cases](/use-cases), or see how Arlo works [in Slack](/use-cases/ai-assistant-in-slack). ## Frequently asked questions **What can an AI assistant do in Notion?** Turn meetings and messages into database rows with the right properties set, update properties, relations, and rollups, retrieve decisions from your wiki with a link to the source, draft and edit pages, and keep trackers and status boards current — asked from Slack, iMessage, WhatsApp, or a call. **Can it answer questions from our Notion wiki?** Yes. Ask something like "what's our refund policy?" or "what did we decide about pricing?" and it pulls the answer from the right page or database and cites the source page, rather than guessing or paraphrasing from memory. **Will it edit pages without me seeing the change first?** No. Page edits and database writes are held for approval before they save. Retrieval and search are reads, so answering a question doesn't require a gate. **Can it file meeting notes into the right database automatically?** Yes — it creates the row with attendees, decisions, and action items as properties, links it to the related project via a relation, and drafts the follow-up tasks, all held for your approval. **Will its updates break my rollups?** It's built to avoid that — it sets relations and properties so rollups recompute correctly, and during cleanup it can scan a database for rows with empty status or stale dates and propose the fixes for you to confirm in one pass. **Can I see what it changed in Notion?** Yes. Every property update, new row, and page edit lands in a [full audit trail](/glossary/ai-agent-audit-trail) tied to the request that caused it. ## Try Arlo with Notion Put an AI colleague on your Notion workspace that files meetings as rows, retrieves decisions, and keeps databases current — with approvals and an audit trail on every write. Arlo starts free — $100 in credits for connecting Slack — then runs on prepaid credits at 400 per $1. [Get started](/signup) and connect Notion in a couple of minutes. --- # AI Assistant for Real Estate Agents and Realtors > An AI assistant for realtors handles lead response, showing scheduling, follow-ups, and CRM updates over the channels agents actually use: iMessage, SMS, and live phone calls. Arlo replies to leads in seconds, books showings, and updates your tools by logging in through a secure browser, even ones with no API. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-realtors An AI assistant for realtors handles the work that doesn't fit between appointments: answering leads, booking showings, chasing follow-ups, and keeping the CRM current. The catch is that real estate doesn't happen at a desk, so the assistant has to meet you where the deal happens. Arlo works inside iMessage, SMS, and live phone calls. It can answer a new lead in seconds while you're mid-showing, then log the result in your tools once you're free. Real estate is a speed business. A lead that waits an hour for a reply has usually already texted three other agents. Meanwhile the busywork stacks up between appointments until you're doing CRM entry at 11pm. An [AI personal assistant](/glossary/ai-personal-assistant) built for this rhythm closes those gaps without you opening another app. ## Respond to leads in seconds, by text Most leads land when you're nowhere near a keyboard. Arlo catches new inquiries and replies over text right away. It answers the obvious first questions, qualifies how serious the person is, and pins down what they're actually looking for. The whole thing happens over iMessage and SMS, where buyers and sellers already are, not behind some portal they have to log into. You set the rules. Arlo can send replies on its own, or draft them and wait for your sign-off, so the tone and any commitments match how you run your business. ## Schedule showings without the back-and-forth Booking one showing can mean a dozen messages bouncing between the buyer, the seller, and your calendar. Arlo runs that coordination for you: - Proposes times that fit your calendar and the listing's availability - Confirms with the buyer over text and holds the slot - Sends reminders before the appointment, so fewer no-shows - Reschedules when plans change and tells everyone involved It remembers the whole thread. So when a buyer texts "can we move it to Saturday?" four days later, Arlo already knows which showing they mean. ## Keep follow-ups from slipping Deals get won in the follow-up. The check-in after a showing, the nudge on a stalled offer, the note three months after closing that turns into a referral. These are the easiest things to forget and the most expensive to drop. Arlo tracks them and brings them to you in a morning briefing of what changed and what needs you — the same pattern as [an AI assistant that follows up for you](/use-cases/ai-assistant-that-follows-up-for-you), applied to a pipeline. | Task | Without an AI assistant | With Arlo | |------|------------------------|-----------| | New lead reply | Minutes to hours later | Seconds, by text | | Showing scheduling | Manual back-and-forth | Coordinated and confirmed | | Post-showing follow-up | Easy to forget | Tracked and surfaced | | CRM updates | Done late or skipped | Logged automatically | ## Update your CRM, even one with no API Plenty of real estate CRMs and MLS systems have no usable API. That's the real reason so much data entry stays manual. Arlo gets around it by using these tools the way you do. It logs in once through a secure browser session, then clicks, reads, and fills forms like a person would. After a showing or a call, it updates the contact record, logs the activity, and moves the deal stage. You don't touch the keyboard. All of that runs under governance. Connections resolve through policy before each run, writes can wait for your approval, and every click, source, and approval lands in an audit trace you can go back and read. ## Coordinate over calls while you're on the move Realtors live on the phone, so Arlo does too. It can place live calls to confirm an appointment or pass along a message, then keep the transcript so you can ask how it went. Pair that with text and a lead can go from first contact to a booked showing without you stopping the car. ## Why text and calls matter here A tool that only lives in a web dashboard asks you to change how you work, and most agents won't. Arlo fits the way real estate already runs: fast replies by text, decisions made on the phone, systems updated quietly in the background. It connects to the apps you already use plus the 3,000+ tools available through integrations, and reaches the holdouts with no API through a browser session. What you get is one [AI colleague](/glossary/ai-colleague) covering the stretch between a lead arriving and a deal closing. It's there when you can't be, and it keeps a record of everything it did. ## Frequently asked questions **What does an AI assistant do for a real estate agent?** It handles lead response, showing scheduling, follow-ups, and CRM updates over the channels agents actually use — iMessage, SMS, and live phone calls. It replies to new leads in seconds, books showings, and keeps your tools current. **Can it reply to leads by text?** Yes, and that's the point — speed of first response is what wins the lead. It answers over iMessage and SMS in seconds rather than whenever you next check your phone. **Will it text a client something I haven't seen?** Client-facing sends wait for your approval. It drafts the reply immediately so you're approving rather than composing, but it doesn't message a lead on its own. **Can it update my CRM if my CRM has no API?** Yes. For tools that never exposed an API, it signs in once through a secure browser session and clicks, reads, and fills forms the way you would — which covers a lot of real-estate software. **Can it handle phone calls?** Yes. It [places calls](/use-cases/ai-executive-assistant-that-makes-phone-calls) and keeps a transcript of each one, which is useful when you're driving between showings and can't type. **Does it work for a solo agent or only a brokerage?** Both. A solo agent starts free with $100 in credits for connecting Slack, then buys credits and spends only what the work costs — a slow month costs nothing; the approval gate and audit trail matter more once a brokerage has several agents acting under one name. ## Try Arlo If lead response, scheduling, and CRM updates are eating your day, hand them off. Weighing tools? Compare [Arlo vs. Poke](/compare/arlo-vs-poke), or see the [AI assistant for sales](/use-cases/ai-assistant-for-sales) playbook. [Try Arlo](/signup) and put it to work on your pipeline. --- # AI assistant for recruiters — outreach and scheduling > An AI assistant for recruiters handles candidate outreach and replies over text, books interviews, runs follow-ups, and updates your ATS or CRM. Arlo works in iMessage, SMS, Slack, and on live calls, updates no-API tools through secure browser sessions, and routes candidate messages through approvals first. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-recruiters The first recruiter to reply usually gets the candidate. The one who sends a sloppy message usually loses them. That's the whole job in two sentences, and it's why an AI assistant has to be fast on the draft and careful on the send. [Arlo](/) drafts quickly, then waits for a human to approve anything that reaches a candidate. It runs inside iMessage, SMS, Slack, and live phone calls, so you're not logging into one more dashboard to do it. ## Move first on candidate outreach Good candidates answer whoever texts them first. Arlo writes the first-touch message from a role brief and the candidate's background, and it replies to inbound texts while you're stuck in an interview or away from your desk. It holds a consistent voice across dozens of live threads, which matters more than it sounds when you're running forty conversations at once and half of them blur together. And because it works over [iMessage and SMS](/glossary/ai-personal-assistant), candidates get a real text, not a link to a portal they'll never log into. ## Schedule interviews without the back-and-forth Coordinating a candidate, a hiring manager, and sometimes a four-person panel is the time sink everyone underestimates. Arlo proposes times that fit the hiring team's calendar, confirms with the candidate over text, and books the slot. It handles reschedules and reminders on the same thread, and it does the time-zone math so a candidate in Berlin never gets an invite set to your local clock. ## Follow up so no candidate goes dark Pipelines leak at the follow-up stage, almost always. Arlo keeps persistent memory on every candidate and nudges at the right moment: a check-in after an interview, a re-engagement text to someone who went quiet three weeks ago, the reply routed straight back to you. Every morning it sends a briefing of what moved across your pipeline overnight, so the first thing you read isn't a cold inbox. ## Update the ATS, even one with no API Data entry eats hours, and the worst culprits are the old applicant-tracking systems that never shipped a usable integration. Arlo logs in once through a secure browser session and works the screen like a person would: clicking, reading, filling forms. - Move candidates between stages and log interview notes. - Update the CRM after a call so the record matches what actually happened. - Reach across the 3,000+ tools Arlo connects to, plus any web app behind a login. ## Approvals keep candidate messages safe Candidate communication is reputation-critical, so Arlo treats every send as a governed action. Drafts, sends, and tool writes wait for a reviewer you pick. Policy is set per run, and every tool call, source, and approval lands in a full audit trace you can pull up later. | Task | Arlo drafts | You approve | Channel | |------|-------------|-------------|---------| | First-touch outreach | Yes | Yes | iMessage / SMS | | Interview confirmation | Yes | Optional | iMessage / SMS | | ATS stage update | Yes | Per policy | Secure browser | | Post-call recap to manager | Yes | Optional | Slack | ## Prep for calls and recruit on the move Arlo can join a live phone call and brief you before one. Text it from Slack or your phone between meetings: "what's the latest with the senior backend candidate?" You get a grounded answer pulled from memory and your connected tools, not a guess. It captures call outcomes and next steps so you're not typing up notes afterward, and the whole thing feels like texting a coworker who already did the reading. Recruiting and sales run the same play: fast, personal, high-volume outreach with a clean handoff. If your team does both, the [AI assistant for sales](/use-cases/ai-assistant-for-sales) page covers the pipeline side, and the [AI colleague](/glossary/ai-colleague) glossary explains how Arlo fits across a team. More in [resources](/resources). Hiring doesn't end at the offer. For onboarding, policy questions, and the people-ops ticket queue that follows, see [AI HR agent](/use-cases/ai-hr-agent). ## Frequently asked questions **What does an AI assistant do for a recruiter?** It handles candidate outreach and replies over text, books interviews, runs the follow-ups that otherwise slip, and updates your ATS or CRM. Arlo works in iMessage and SMS, Slack, and on live calls, and reaches no-API tools through a secure browser session. **Will it message candidates without my approval?** No. Candidate-facing messages route through an approval first — Arlo drafts, you approve. That matters more in recruiting than most functions, because a clumsy automated message to a candidate costs you the candidate. **Can it text candidates rather than email them?** Yes, over iMessage and SMS, which is where response rates actually are. Speed of first contact is the main advantage — it can reply within seconds of a candidate responding. **Can it update our ATS if there's no API?** Yes. For an ATS that never shipped an API, it signs in once through a secure browser session and updates records by hand, under the same approvals and audit trail as an API call. **Can it schedule interviews across several calendars?** Yes — it handles the back-and-forth of finding a slot that works and books it, instead of you brokering times between a candidate and a hiring panel. **Does it help with call prep?** Yes. It can pull together the candidate's background and the role context before a screen, and it can join you on the move via [phone calls](/use-cases/ai-executive-assistant-that-makes-phone-calls). ## Try Arlo Put an AI colleague on your recruiting pipeline. It texts candidates, books interviews, and updates your ATS, with every message reviewed before it sends. [Get started](/signup). --- # AI assistant for sales teams > An AI assistant for sales works inside your reps' chat, email, and calls to follow up on leads fast, log CRM activity, draft outreach, book meetings, and research prospects before calls. It takes real actions through governed, approved tool access, so managers get speed plus an audit trail of every send. Updated: 2026-06-19 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-sales An AI assistant for sales lives in the tools your reps already use — iMessage, SMS, Slack, WhatsApp, and live phone calls — and does the work that surrounds a deal. It replies to leads, keeps the CRM honest, drafts outreach, books meetings, and reads up on a prospect before a call. The part managers care about: nothing gets sent or written until an approver you pick signs off, and every action lands in a full audit trace. Reps don't lose deals because they're bad at selling. They lose them to slow follow-up, a CRM that's half-updated, and an hour of admin a day that should have been selling time. [Arlo](/) does that work from the same thread a rep is already texting in, so they don't have to stop and switch apps. This guide covers rolling the assistant out across a sales team. For the pipeline-mechanics view — follow-up cadences, CRM hygiene, approval-gated outreach deal by deal — see the companion [AI sales agent](/use-cases/ai-sales-agent) guide. ## What an AI assistant does for sales reps The repetitive, time-sensitive stuff around every deal: - **Lead follow-up speed** — drafts the first reply to a new inbound in seconds, then either nudges the rep to send or sends once approved. - **CRM logging** — turns a quick text into notes, call outcomes, and next steps written straight to the record. - **Outreach** — writes emails and messages grounded in the prospect's context and your past conversations, not a generic template. - **Scheduling** — proposes times, books the meeting, and handles the inevitable reschedule. - **Pre-call prep** — pulls account history, recent news, and open threads so nobody walks in cold. - **Deal momentum** — flags deals that have gone quiet and reminds reps of what they promised. ## Realistic scenarios A lead fills out a form at 9 p.m. Arlo drafts a tailored reply and queues it. The rep taps approve from the couch, and the lead hears back before a competitor wakes up. A rep texts a two-line summary after a discovery call. Arlo logs it, moves the deal stage, and files the follow-up task. No CRM tab opened. It's the morning of a renewal call. Arlo has already sent a briefing: the account's history, last quarter's usage, and the three questions left hanging in the last thread. A deal has been silent for ten days. Arlo notices, flags it, and drafts a re-engagement message for the rep to look over. ## Governance built for sales managers Speed without oversight is how the wrong message reaches the wrong account, and in sales that costs real money. Arlo is built so you keep the reins. - **Approval before send** — drafts, emails, and CRM writes wait for a reviewer you choose. Nothing goes out unreviewed unless you decide it should. - **Policy on every run** — each tool connection resolves through policy before the assistant acts. - **Full audit trace** — every tool call, data source, and approval is logged, so you can see what was sent, to whom, and why. So you can hand reps an AI assistant without losing sight of what goes out under the team's name. ## AI assistant vs. doing it manually | Task | Manual | With an AI assistant | | --- | --- | --- | | New lead reply | Minutes to hours, if remembered | Drafted in seconds, sent on approval | | CRM logging | Skipped or batched late | Logged from a quick text | | Call prep | Ad-hoc, time permitting | Briefing ready before the call | | Stalled deals | Caught on a pipeline review | Surfaced proactively | | Manager visibility | Spot checks | Full audit trace of every action | ## How it connects to your sales stack Arlo plugs into Slack, Gmail, Notion, Linear, GitHub, and 3,000+ other tools. For the sales systems that don't have a clean API — and there are always a few — it logs in once through a secure browser session and reads records, fills forms, and updates fields the way a person would. The CRM you already run stays in scope. Memory is what makes it feel like an [AI colleague](/glossary/ai-colleague) rather than a macro. Arlo remembers accounts, past conversations, and commitments, and sends a morning briefing of what moved overnight. ## Frequently asked questions **What does an AI assistant do for a sales team?** It works inside your reps' chat, email, and calls to follow up on leads fast, log CRM activity, draft outreach, book meetings, and research prospects before calls — taking real actions through governed, approved tool access rather than just suggesting them. **How does a sales manager keep control of what reps' assistants send?** Every send and CRM write waits for a reviewer, and every action lands in an audit trail tied to the rep and the request. Managers get the speed without losing the record of what went out in the company's name. **Will it keep the CRM honest?** That's one of its better uses — logging calls, emails, and notes to the right record automatically, so pipeline data reflects what actually happened instead of what a rep remembered to enter on Friday. It works with [HubSpot](/use-cases/ai-assistant-for-hubspot) and 3,000+ other tools. **Can it respond to inbound leads quickly?** Yes — it drafts and, once approved, sends the first reply in seconds rather than hours. For inbound speed-to-lead, that gap is usually the whole game. **Can it research a prospect before a call?** Yes. It pulls together account context and recent activity ahead of a meeting, and after a [call it places or takes](/use-cases/ai-executive-assistant-that-makes-phone-calls) it can log the outcome to the matching deal when you ask. **How is this different from a sales-engagement platform?** Sequencers execute a defined cadence at volume. Arlo reasons per request, handles the ad-hoc and cross-tool work around the cadence, and gates the consequential steps. Many teams run both — see [AI assistant vs workflow automation](/compare/ai-assistant-vs-workflow-automation). ## Try Arlo Founders and revenue leaders setting up their first AI hire can start with the [AI assistant for founders](/use-cases/ai-assistant-for-founders) use case, then roll it out to the team. Comparing options? See [Arlo vs. Lindy](/compare/arlo-vs-lindy), or how the same colleague works for [recruiters](/use-cases/ai-assistant-for-recruiters). Ready to put one to work on your pipeline? [Try Arlo](/signup). --- # AI assistant for small business owners — one hire, every hat > An AI assistant for a small business absorbs the owner's coordination load: answering and placing phone calls, triaging the inbox, chasing invoices and quotes, keeping the calendar honest, and updating records. Arlo does this from iMessage, SMS, Slack, or a phone call — no new software to learn — and asks approval before anything reaches a customer. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-for-small-business In a small business the org chart is a lie: whatever the titles say, the owner is also the receptionist, the collections department, the scheduler, and the person who answers "quick questions" at 9pm. None of those jobs is hard. Together they're a second shift, and they're why the actual work — the thing the business sells — gets squeezed into the gaps. That second shift is exactly what an AI assistant should take. Not "AI transformation," not a dashboard, not another app your team ignores: a colleague you text like a person, that makes phone calls for you, chases the money, keeps the calendar, and asks before anything goes out under your name. ## The jobs Arlo takes off an owner - **The phone** — Arlo [places real calls](/compare/ai-assistants-that-can-make-phone-calls): confirms appointments, chases suppliers, asks the questions you'd otherwise spend an afternoon on hold for. The missed-call-while-on-a-job problem stops costing you customers. - **The inbox** — [triage, drafts in your voice, and follow-ups that don't die](/use-cases/ai-assistant-for-email). You see the three emails that matter; the other forty get handled. - **[The money chase](/use-cases/ai-finance-agent)** — quotes that went quiet, invoices at day 30, the deposit that never landed. Arlo tracks the open loops and chases by email or by phone, politely and on schedule. - **The calendar** — bookings, reschedules, and reminders handled by text, with customer no-shows cut by confirmation calls the day before. - **The records** — whatever you use — spreadsheets, a CRM, QuickBooks-adjacent tools among 3,000+ integrations — Arlo keeps it updated as a side effect of the work, not a Friday chore. Even software with no API works: Arlo logs in once through a secure browser session and uses it like a person. ## Why texting is the right interface for a small business Nobody in a five-person company wants to learn software. Everyone already texts. Arlo lives in iMessage and SMS (plus Slack and WhatsApp if you use them), so "adopting AI" means saving a contact. You text it the way you'd text a sharp office manager — "push my 2pm, chase the Ramirez quote, and what did the roofer say?" — and it answers in the same thread, with one memory across every channel, so you never re-explain your business. ## Trust, in writing The reason owners don't delegate is that mistakes land on them personally. Arlo is built for that reality: **approval before anything customer-facing goes out** — every email, every commitment — until you decide, category by category, what it can do on its own. And everything it does lands in a full [audit trail](/glossary/ai-agent-audit-trail): the call it made, the email it sent, what it changed and why. Delegation without the flinch. The math is blunt: $50 is the smallest amount of Arlo credits you can buy, and it stretches across months of texting, calling, and chasing — against ten-plus hours of owner time a week returned. It's the cheapest hire you'll ever make, and the only one that never calls in sick. Two neighbouring versions of this page go deeper by shape of business: [AI assistant for e-commerce](/use-cases/ai-assistant-for-ecommerce) for stores juggling suppliers, 3PLs, and customer email, and [AI operations agent](/use-cases/ai-operations-agent) for the recurring work that has to happen again next week. ## Frequently asked questions **What can an AI assistant actually do for a small business?** Place phone calls, triage email, draft replies, chase quotes and invoices, manage bookings and reminders, and keep your records updated across your tools — with your approval on anything customer-facing. **Do I need to install new software?** No. Arlo works over iMessage and SMS — you text it like a person. Slack, WhatsApp, and phone calls work too. Your existing tools stay; Arlo connects to them. **Can it answer my business line?** No. Arlo places calls for you, like confirmations and follow-ups, and keeps a transcript of each one you can ask about, and you can call it yourself to hand it a task. It doesn't answer or screen your customers' calls; for that, use an answering service. **Will it send things to customers without me seeing them?** Not unless you tell it to. Approval gates are on by default; most owners keep them on for anything customer-facing and loosen them for routine internal work. **What does it cost?** Connecting Slack starts your workspace free with $100 in credits, and after that a monthly plan buys credits, with 10% bonus credits, that you spend as work happens. The rate is 400 credits per $1 at every amount, on monthly plans from $20; $50 buys 20,000 credits, roughly 625 messages or 18 apps built, and credits last 12 months. Phone calls and everything else are included from the first dollar. Admin policy, roles, SSO, and audit export are available on request. --- # Slack AI Assistant: Answer, Triage, and Act in Channels > A Slack AI assistant answers in channels and DMs, triages threads, drafts replies, and acts in connected tools like Gmail, Notion, Linear, and GitHub. Approvals hold every send or write for a reviewer, and memory carries context across Slack, iMessage, WhatsApp, and calls. Updated: 2026-07-31 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-in-slack An AI assistant in Slack does the work where the work already happens. It answers in channels and DMs, untangles long threads, drafts the reply you were going to write anyway, and takes real action in your connected tools. Approvals are on by default, so anything that sends or writes pauses for a reviewer you pick. Slack stays fast, but nothing goes out behind your back. Arlo is an [AI colleague](/glossary/ai-colleague) that lives inside Slack like another teammate — the [Slack AI assistant page](/slack) covers what it is and what it costs; this guide is the mechanics. It reads the channel, figures out what you need, and acts through Gmail, Notion, Linear, GitHub, and 3,000+ other tools. Its memory follows the work, so a thread you start in Slack still makes sense later on iMessage, on WhatsApp, or on a phone call. Most "AI in Slack" is a search box: it finds messages and summarizes threads. That's useful, but it stops at answers. An AI assistant your employees can actually delegate to does the next step — it drafts the reply, opens the ticket, updates the record, and sends once someone approves. The difference is whether it hands you information or hands you back finished work. ## What an AI assistant does in Slack A real Slack assistant is not an FAQ bot. It's in the flow: - **Answers in channels and DMs** with memory of past decisions, projects, and who's who. - **Triages threads** by summarizing the discussion, flagging the actual decision, and pulling in context. - **Drafts replies, messages, and docs** so you're editing instead of staring at a blank box. - **Acts in connected tools**: opens a Linear issue, comments on a GitHub PR, updates a Notion page, queues a Gmail draft. - **Runs scheduled work** like a recurring roundup or a reminder that fires when it should. - **Posts a morning briefing** of what moved overnight across your tools. ## Concrete Slack scenarios What that looks like in a real workspace: - Mention Arlo on a nasty customer thread in `#support`. It summarizes the issue, checks Notion for the runbook, and drafts a reply for an agent to approve before it sends. - In `#engineering`, ask it to open a Linear issue from a bug report in the thread. It writes the title, description, and labels, then waits for a reviewer to confirm. - In a DM, ask "what changed in the launch project since Friday?" Arlo reads Linear, GitHub, and Notion and hands back a tight summary. - Schedule a 9am post in `#standup` that rolls up open PRs and at-risk issues. You approve the format once, then it runs on its own. - Ask for a follow-up to a prospect mentioned in `#sales`. The Gmail draft sits there until you sign off. ## Actions and approvals Governance is what separates a useful Slack assistant from a liability. Arlo resolves every connection through policy before a run starts, and any send or write [waits for the reviewer you choose](/use-cases/ai-assistant-with-approval-before-sending). Every tool call, source, and approval lands in a full [audit trace](/glossary/ai-agent-audit-trail), so you can always see what happened and why. ### How the approval step works in Slack This is a [human-in-the-loop](/glossary/human-in-the-loop) gate, and it runs the same way every time: 1. **Arlo prepares the action.** It researches the thread, drafts the email or writes the record change, and resolves which connection it would use — but does not call the tool yet. 2. **Policy decides whether a gate applies.** Rules are scoped by consequence, not by tool: outbound messages, irreversible writes, anything touching money or a customer. Routine internal reads run without interrupting anyone. 3. **An approval card posts in Slack.** The reviewer sees which tool and account the action would use and a preview of exactly what Arlo is about to send or write — in the thread where the request came up, not in a separate console someone has to remember to check. 4. **The reviewer approves or denies** from the card. Approving can be a one-time yes, or a standing yes scoped to that kind of action and set to expire on a date you choose — so "let it post to #standup" never quietly becomes "let it post anywhere". 5. **Arlo executes only after approval**, then reports the outcome back in the thread and writes the chain — tool, actor, approver, decision, result — to the audit trace. 6. **Nothing approved means nothing sent.** Gates fail closed: an approval that is never answered expires, and an expired approval can no longer be executed. There is no path where silence becomes consent. Autonomy is granted per category, not all at once. When a category has proven boring for weeks — the routine reschedule, the internal status post — you extend a scoped standing approval for it, and the audit trace is the record you extend it from. | Capability | Read-only bot | AI assistant in Slack (Arlo) | | --- | --- | --- | | Answer in channels/DMs | Yes | Yes | | Triage and summarize threads | Limited | Yes, with memory | | Take actions in connected tools | No | Yes, via 3,000+ integrations | | Approvals before sends/writes | No | Yes, reviewer you choose | | Audit trace of every action | No | Yes | | Tools with no API (browser) | No | Yes, logs in once and clicks/fills | ## Tools with and without an API Plenty of internal systems never shipped an API. Arlo handles those by logging in once in a secure browser session and clicking, reading, and filling forms the way a person does. So a Slack request can drive a legacy portal or an internal dashboard that no off-the-shelf integration touches, under the same approvals and audit trail as everything else. ## How to connect an AI assistant to Slack Getting Arlo into a workspace takes minutes, not a rollout. You add it, point it at the channels it should watch, and connect the tools it's allowed to act in — each connection scoped by policy up front. From there, people mention it in a channel or DM it directly, the same way they'd loop in a coworker. There's nothing for the rest of the team to install and no new app to learn: if they can use Slack, they can use Arlo. ## Context that follows the work Slack is rarely the only place your team talks, so Arlo is one assistant across [iMessage and SMS](/imessage), Slack, [WhatsApp](/whatsapp), and live phone calls. A thread can start in a channel and finish on a call, and the morning briefing finds you wherever you are. If you're solo, see how an [AI executive assistant](/glossary/ai-executive-assistant) or [AI personal assistant](/glossary/ai-personal-assistant) fits a single workday. Want role-specific takes? Look at how this plays for [founders](/use-cases/ai-assistant-for-founders), [sales teams](/use-cases/ai-assistant-for-sales), and [realtors](/use-cases/ai-assistant-for-realtors), or dig through [resources](/resources). Comparing the field rather than the capability? [Best AI assistants for Slack](/compare/best-ai-assistants-for-slack) lines up eight tools against the three jobs teams actually hire a Slack assistant for. ## Frequently asked questions **What is a Slack AI assistant?** A Slack AI assistant is an AI that works inside your Slack workspace — answering in channels and DMs, summarizing and triaging threads, drafting replies, and taking action in connected tools like Gmail, Notion, Linear, and GitHub. The useful ones act, not just answer; the safe ones hold every send or write for a human reviewer first. **What's the difference between Slack AI and a Slack AI assistant?** Slack's own AI features summarize threads, search messages, and recap channels inside Slack. A Slack AI assistant like Arlo also acts outside Slack — opening the Linear issue, queuing the Gmail draft, updating the CRM record — using Slack as the place you delegate from rather than the boundary of what it can do. **Can an AI assistant in Slack take actions with approval?** Yes. Arlo prepares the action, policy decides whether it needs a gate based on consequence, and an approval card posts in Slack showing the tool, the account, and a preview of what it would send. The reviewer approves or denies from the card, and nothing executes until someone approves. Unanswered approvals expire rather than defaulting to send. **Is it safe to give an AI assistant access to Slack channels?** The safety comes from scoping and review, not from trust. You pick the channels it watches and the tools it may act in, each connection is resolved through policy before a run starts, consequential actions wait for an approver, and every run leaves an [audit trace](/glossary/ai-agent-audit-trail) showing what it read, what it called, and who approved it. **Does everyone on the team need to install something?** No. Arlo is added to the workspace once; from there people mention it in a channel or DM it like any coworker. There is no per-person app to install and no new interface to learn. **Can the same assistant work in iMessage too?** Yes — Arlo runs in Slack, [WhatsApp](/whatsapp), [iMessage and SMS](/imessage), and on live phone calls, with one memory across all of them. A thread that starts in a Slack channel can finish on a call without re-explaining the context. ## Try Arlo Bring an AI colleague into your Slack workspace that answers, drafts, and acts, with approvals and a full audit trail behind every move. [Get started](/signup) and connect your first channel in minutes. --- # An AI Assistant That Follows Up For You > An AI assistant that follows up for you tracks who owes you a reply and chases them on a schedule you set. It notices a thread that went quiet, drafts the follow-up in your voice, and waits for your approval before sending. Arlo works across email, text, Slack, and phone calls, so one assistant can chase by text, email a reminder, or call to confirm. Updated: 2026-08-25 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-that-follows-up-for-you Almost nothing is lost because the work was bad. It is lost because nobody followed up. The quote that never got chased, the lead who went quiet after one good call, the invoice sitting unpaid because reminding people feels rude — the money leaks out of the gap between *I should follow up on that* and actually doing it. That gap is a memory problem and a friction problem at once. You have to remember who owes you a reply, and then find the will to write the third message into silence. An AI assistant that follows up for you closes both halves: [Arlo](/) keeps track of who has gone quiet, and drafts the chase so the only thing left for you is to say *send*. ## It notices the silence The hard part is not writing the follow-up. It is knowing one is due. Arlo watches the threads you care about and surfaces the ones where the ball is in someone else's court and has been for too long. - A prospect replied warmly nine days ago and nothing has happened since. - A quote went out on Thursday and was never acknowledged. - You promised someone a document and never sent it. You get a text listing what is hanging, not a dashboard you would have to open. If you would rather it be quiet on the days nothing is outstanding, it will be — see [an AI assistant that texts you reminders](/use-cases/ai-assistant-that-texts-you-reminders). ## It drafts the chase, in your voice A follow-up is a small piece of writing with a surprisingly high cost: you have to reread the thread, remember what you promised, and find a tone that is persistent without being irritating. Arlo does that part. It pulls the context from the thread, writes the message the way you write, and hands it to you finished. You reply *send* and it goes. You reply *make it warmer* or *shorter* or *mention the deadline* and it rewrites. Nothing leaves without your approval, and every send is logged in [a full audit trail](/glossary/ai-agent-audit-trail). ## It chases on whatever channel the person actually uses People go quiet in different places, and the follow-up has to go where they are. Arlo is one assistant across [email](/use-cases/ai-assistant-for-email), [iMessage and SMS](/imessage), [Slack](/slack), and [phone calls](/phone), so a single follow-up sequence can: - text a lead who came in through Instagram or a DM, - email the quote reminder with the original attached, - and [place a live call](/use-cases/ai-executive-assistant-that-makes-phone-calls) when a thread has gone cold enough that another email will not fix it. That last one is the piece most assistants cannot do. A call gets answered when three emails did not, and Arlo can make it, then tell you what was said. ## It keeps the cadence you would not keep Good follow-up is a rhythm, and rhythms are exactly what a busy week destroys. You can hand Arlo the cadence once — *chase twice in the first week, then weekly for a month, then leave it* — and it holds that pattern across every open thread without you tracking any of it. When someone finally replies, the chasing stops on its own. When they don't, it stops when you told it to, so nobody gets pursued past the point of being reasonable. ## Frequently asked questions **Can AI follow up on emails for me?** Yes. Arlo tracks which threads are waiting on someone else, drafts the follow-up in your voice when one goes quiet, and sends it once you approve. You can set the cadence once — chase twice in the first week, then weekly — and it holds that pattern across every open thread. **How does it know a follow-up is due?** It watches the threads you care about and looks at who replied last. When the ball has been in someone else's court longer than your cadence allows, it surfaces that thread to you as a text rather than a dashboard you would have to open. **Will it keep chasing someone forever?** No. Chasing stops on its own when the person replies, and otherwise stops at the limit you set, so nobody gets pursued past the point of being reasonable. **Can it follow up by text or phone instead of email?** Yes. Arlo is one assistant across email, iMessage and SMS, Slack, and live phone calls, so a follow-up can go out on whichever channel that person actually uses — including a real phone call when another email will not fix it. ## Who this is for Anyone whose income depends on threads not dying: realtors chasing leads out of Instagram DMs, [consultants](/use-cases/ai-assistant-for-consultants) waiting on scope approvals, [agencies](/use-cases/ai-assistant-for-agencies) chasing feedback rounds, [recruiters](/use-cases/ai-assistant-for-recruiters) keeping candidates warm, and [small businesses](/use-cases/ai-assistant-for-small-business) waiting to get paid. The common shape is simple: a small number of open threads where forgetting one costs real money — and no assistant to hold them. --- # An AI assistant that texts you reminders (and does the task) > An AI assistant that texts you reminders sends the nudge to Messages instead of an app you have to open. You set it in one sentence — "remind me to chase the Peterson quote every Tuesday at 9" — and it schedules that, confirms the next run time, and texts you until you say stop. Arlo can stay silent when there is nothing to report, and can carry the drafted follow-up with the reminder. Updated: 2026-09-23 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-that-texts-you-reminders A reminder app only works if you open the reminder app. That is the whole problem with them. The thing you needed reminding about is competing with every other reason not to look, and the badge sitting on a home screen loses that fight most days. An AI assistant that texts you reminders takes a different route: the nudge arrives in Messages, alongside the threads you already read without deciding to. [Arlo](/) gives you a number you text like a contact, and anything you ask it to remind you about comes back the same way — no app to install, no list to maintain, no separate place to check. ## You set it by writing one sentence You don't fill in a form or pick a repeat schedule. You write what you'd say to a person: - *"remind me to chase the Peterson quote every Tuesday at 9."* - *"text me the night before anything is due."* - *"every weekday at 7:30, tell me what changed in my inbox overnight."* Arlo schedules it and tells you the exact time of the next run, so you know it actually exists. That confirmation matters more than it sounds. A reminder you *think* you set is worse than no reminder at all. Changing it works the same way. Reply "make that 8 instead" or "only on weekdays" or "stop that one" and it updates. Because the setup lives in the same thread as the reminders themselves, you never have to remember where you configured it. ## Reminders that carry the work with them Most reminders hand the job straight back to you. You get a nudge that says *follow up with Dana*, and now you are the one writing the email at 9pm. Because Arlo is connected to your tools, the reminder and the work arrive together. The text that tells you Dana has gone quiet can carry the drafted follow-up underneath it, ready to send when you reply *yes*. It can book the time, update the record, or place a [phone call](/use-cases/ai-executive-assistant-that-makes-phone-calls) instead of texting. Nothing that sends or writes goes out without your approval, and every action is kept in [an audit trail](/glossary/ai-agent-audit-trail). A reminder app tells you about the work. An assistant has usually started it by the time you read the text. ## Silence when there is nothing to say The fastest way to get muted is to send a message on a day when nothing happened. Arlo can be told to stay quiet unless something actually changed, which makes it safe to set a watch on a slow-moving thing. - *"watch the shared drive and only text me if the contract file changes."* - *"check the weather every two hours and only warn me if there is a storm coming."* - *"tell me if this lead replies, otherwise say nothing."* A watch that costs you nothing on quiet days is a watch you leave running for months. That is what makes it useful on the one day it fires. ## What people actually set up The reminders that stick are unglamorous and specific: - A morning rundown of what moved overnight, before the day starts. - A nudge to follow up on a quote, an application, or [a lead who went cold](/use-cases/ai-assistant-that-follows-up-for-you). - A bill, renewal, or free-trial warning a day before it charges. - A recurring check-in that asks the question you would rather avoid — how the day plan is going, whether you ate, what is still open. - A weekly report pulled from the same place every Monday. The pattern underneath all of them is the same: something that costs real money or real trust when it is forgotten, and that you will not remember on your own. ## How it compares to your phone's reminders and ChatGPT tasks Your phone's reminders app and ChatGPT's scheduled tasks both work, and for a lot of people they're enough. ChatGPT's paid plans can even act on a schedule. The difference is where the nudge lands. | | Arlo | Phone reminders app | ChatGPT scheduled tasks | |---|---|---|---| | Where it reaches you | A text in Messages | A notification | Inside the ChatGPT app, with a notification | | How you set it | One sentence, by text | Tap through a form, or ask Siri | A prompt in the ChatGPT app | | Stays quiet when nothing changed | Yes, if you ask | No | Depends on the prompt | | Can do the task, not just remind you | Yes, with your approval | No | Yes, on paid plans with ChatGPT Work | If you mostly ignore notifications, a text is harder to miss. That's the whole bet. ## Frequently asked questions **Is there an AI that can text me reminders?** Yes. Arlo is an AI assistant you text like a contact, and it texts you back when a reminder is due. You set reminders in plain language, and it's free to start with no app to install. **Is there an app that sends reminder texts?** You don't need an app for it. With Arlo you text a number, ask for a reminder in one sentence, and the reminder arrives later as a normal text message in the same thread. **Can an AI assistant text me reminders instead of using an app?** Yes. Arlo gives you a number you text like a contact, and every reminder you ask for comes back to that thread. There's nothing to install and no list to maintain. The reminder lands in Messages next to everything else you already read. **How do I set up a recurring text reminder?** You write one sentence, like "remind me to chase the Peterson quote every Tuesday at 9". Arlo schedules it and replies with the exact time of the next run, so you know it exists. To change it, reply in plain language: "make that 8", "only weekdays", "stop that one". **Will it text me on days when nothing happened?** Only if you want it to. A reminder can be told to stay silent unless something actually changed, which is what makes it safe to leave a watch running on a slow-moving thing for months. **Can it do the task, not just remind me about it?** Yes. Because Arlo is connected to your tools, the reminder can arrive with the drafted email underneath it, ready to send when you reply yes. It can also book the time, update a record, or place a call, always behind your approval. ## Where the reminders land The same assistant runs in [iMessage and SMS](/imessage), [Slack](/slack), and [WhatsApp](/whatsapp), and can reach you on a [phone call](/phone) when a text is not enough. A schedule you set by text can deliver into a Slack channel for your team, and a watch you set in Slack can text your phone when it fires. If you want the broader picture of an assistant that starts the conversation rather than waiting for you, see [proactive AI](/proactive). --- # AI Assistant With Approval Before Every Send > An AI assistant with approval before sending pauses on any action that leaves a mark — an email, a tool write, a payment — and waits for a human to sign off first. Arlo puts this gate on every send and write by default, resolves each tool connection through policy before a run, and logs every approval in an audit trail. Updated: 2026-07-04 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-with-approval-before-sending The moment an AI assistant can actually do things — send the email, update the record, place the order — the question stops being "is it smart enough?" and becomes "what happens when it's wrong?" An assistant that acts without a checkpoint is a liability waiting to send the wrong message to the wrong person. Arlo is built the other way around: nothing that leaves a mark goes out until you approve it. Arlo is an [AI colleague](/glossary/ai-colleague) with an approval gate on every send and write by default. It drafts, decides, and prepares the action — then stops and asks. You stay the reviewer on anything that touches the outside world, so the assistant moves fast without moving past you. ## What "approval before sending" covers The gate isn't just for email. It sits in front of anything that commits you: - **Sends** — an email, a Slack message, a text, a reply on your behalf. - **Writes** — updating a CRM record, editing a Notion page, opening or closing an issue. - **Bookings and spend** — scheduling something external, or any action that costs money. - **Calls that commit** — when Arlo is [on the phone](/use-cases/ai-executive-assistant-that-makes-phone-calls), agreeing to a booking or a charge waits for your sign-off too. Read-only work — research, summaries, drafts you haven't asked it to send — flows without interruption. The friction lands only where the risk is. ## Why teams need the human in the loop "Human in the loop" gets said a lot; most tools mean an off switch buried in settings. For a team, approvals have to be the default and they have to be visible: - **A wrong send is expensive.** One bad reply to a customer or a prospect can't be unsent. A gate is cheaper than an apology. - **Trust is what lets you delegate.** People hand off more work to an assistant they know can't act behind their back — so the gate makes the assistant more useful, not less. - **Governance is becoming table stakes.** As rules like the EU AI Act phase in through 2026, "a human reviewed it" stops being a nice-to-have and starts being something you need to show. ## How Arlo's approvals work Governance isn't a mode you switch on later — it's how a run is shaped from the start: 1. **Policy first.** Before a run begins, every tool connection resolves through policy — what Arlo may touch, and how far it can go. 2. **Act up to the line.** Arlo does the reasoning, the drafting, and the read-only steps on its own. 3. **Pause at the gate.** Anything that sends, writes, or spends stops and surfaces for the reviewer you chose. 4. **Log it.** The approval, who gave it, and the action it released all land in an [audit trail](/glossary/ai-agent-audit-trail). | | Ungoverned AI assistant | Arlo | | --- | --- | --- | | Sends and writes by default | Acts on its own | Waits for approval | | Who reviews risky actions | No one | A reviewer you choose | | Read-only work | Flows | Flows | | Record of what was approved | None | Full audit trail | | Policy on what tools it can touch | Rare | Resolved before every run | ## Fast and safe aren't opposites The worry is always that a gate slows everything down. It doesn't, because most of what an assistant does is read-only, and the gate only appears at the point of commitment — usually one tap to release something Arlo already prepared. You get the speed of delegation and the safety of review, in the [channels you already use](/use-cases/ai-assistant-in-slack). ## Frequently asked questions **What does "approval before sending" mean?** It means the assistant prepares an action but doesn't complete it until a human approves. With Arlo, every send, write, booking, or payment waits for a reviewer you choose; read-only work flows without interruption. **Does approval on everything make it slow?** No. Most of an assistant's work is research and drafting, which never hits the gate. The gate appears only at the moment something would commit, and clearing it is usually a single tap. **Can I choose who approves?** Yes. You set the reviewer, so the right person signs off before anything goes out. **Is there a record of what got approved?** Yes. Every approval — who gave it and what it released — is stored in a full [audit trail](/glossary/ai-agent-audit-trail) tied back to the original request. **Why does this matter for compliance?** Regulations increasingly expect a human check on consequential automated actions. An approval gate plus an audit trail is how you demonstrate one was in place. ## Getting started Give an assistant real reach without giving up the last word. [Try Arlo](/signup) and see how much you'll hand off once you know nothing goes out without you. --- # The AI Assistant You Can Text and Call > An AI assistant you can text and call is one you reach the way you'd reach a coworker — message a number or ring it, with no app to install. Arlo answers by text or on the phone, takes real action in your tools behind an approval gate, and remembers context across every channel, including Slack and WhatsApp. Updated: 2026-07-04 · Canonical: https://www.arlo.sh/use-cases/ai-assistant-you-can-text-and-call Most AI tools ask you to open a tab, log in, and remember they exist. The ones people actually use meet them where they already are. Arlo is an AI assistant you reach the same two ways you reach a coworker: you text it, or you call it. There's no app to download and no new account to set up for the people you share it with. Arlo is an [AI colleague](/glossary/ai-colleague) that answers over text and over the phone, and does the same real work either way. Send it a message like you'd text a contact, or ring its number and talk. It listens, takes action in your connected tools, and answers you on the call — and because it's one assistant across channels, the conversation carries whether you started it [by text](/imessage), [on a call](/phone), in [Slack](/use-cases/ai-assistant-in-slack), or on [WhatsApp](/whatsapp). ## Text it or call it — same assistant - **Text a number, no app.** Message Arlo in iMessage or SMS the way you text anyone. Ask a question, hand off a task, or forward something to deal with. See the [text-in-iMessage](/imessage) entry for the details. - **Call it and talk.** When it's easier to say than to type, ring Arlo and give it the task out loud. It can also place calls back out on your behalf — [more on the calling side here](/use-cases/ai-executive-assistant-that-makes-phone-calls). - **One memory across both.** Start on a call, follow up by text, and it still knows what you're talking about. The context follows you, not the channel. - **Share it without onboarding.** Because it's just a number, a teammate or a client can reach it without installing anything or creating a login. ## Why "no app" matters for a team A tool everyone has to be trained on is a tool half the team quietly abandons. A number they can text or call has no learning curve — people already know how. That's the difference between a tool you roll out and one that actually gets used. For a small team, "text or call this number" is the entire onboarding. ## What it does once you reach it Getting to Arlo is the easy part. What makes it worth reaching is that it acts: - Triages and drafts your email, then sends once you approve. - Schedules meetings and handles the reschedule back-and-forth. - Reads from and writes to your CRM, project tracker, docs, and 3,000+ other tools. - Uses apps that have no API by signing in once through a secure browser session. - Runs recurring work and delivers a morning briefing of what changed. ## Arlo vs. AI you can only text Plenty of assistants let you text them. Far fewer also answer the phone, work across your team's channels, and act under governance. | Capability | Text-only AI assistant | Arlo | | --- | --- | --- | | Text a number, no app | Yes | Yes | | Reach it by phone call | No | Yes | | Places outbound calls for you | No | Yes | | Works in Slack too | Rarely | Yes | | Acts across 3,000+ tools | Limited | Yes | | Approval gate before every send or write | Usually no | Yes | | Full audit trace of every action | No | Yes | If you're comparing the text-first assistants specifically, see the ranked list of [best AI assistants you can text](/compare/best-ai-assistants-you-can-text), plus [Arlo vs. Poke](/compare/arlo-vs-poke) and the [Poke alternatives](/compare/poke-alternatives) rundown. ## Safe to hand real work to An assistant you can reach this easily still has to be safe to trust. Every connection resolves through policy before a run, anything that sends or writes waits for a reviewer you choose, and every action lands in an [audit trace](/glossary/ai-agent-audit-trail). You get the convenience of a number to text or call, without giving up control of what it does. ## Frequently asked questions **Is there an AI assistant I can just text or call?** Yes. Arlo gives you a number you can text in iMessage or SMS, or call directly. There's no app to install — you reach it the way you'd reach a person, and it answers and acts. **Can I call it, not just text?** Yes. You can ring Arlo and give it a task by voice, and it can place calls back out on your behalf, then tell you how it went when you ask. **Do the people I share it with need an account?** No. Because Arlo is reachable at a number, a teammate or client can text or call it without installing anything or signing up. **Does it remember what I said on a call when I text later?** Yes. Arlo keeps one memory across text, calls, Slack, and WhatsApp, so context follows the conversation instead of resetting per channel. **Can it actually do things, or just reply?** It acts. Arlo connects to Gmail, Slack, Notion, Linear, GitHub and 3,000+ tools, waits for your approval on anything risky, and logs every step. ## Getting started Save one number and use it two ways. [Try Arlo](/signup), text it or call it, and hand off the first thing on your list. --- # AI Executive Assistant That Makes Phone Calls > An AI executive assistant that makes phone calls can place live calls for you — confirming an appointment, chasing a vendor, or passing along a message — and keep the transcript so you can ask how it went. Arlo does this alongside iMessage, Slack, and WhatsApp, and waits for your approval before anything is sent or booked. Updated: 2026-07-04 · Canonical: https://www.arlo.sh/use-cases/ai-executive-assistant-that-makes-phone-calls A human executive assistant does something almost no AI assistant does: they pick up the phone. They call the restaurant to move the reservation, chase the vendor who went quiet, and confirm the appointment so it doesn't fall through. Most "AI executive assistants" can draft an email about the call, but they can't make it. Arlo can. Arlo is an [AI colleague](/glossary/ai-colleague) that places real phone calls as part of the same job it does over text — see [Arlo on the phone](/phone) for the channel itself. You ask it in iMessage, Slack, or WhatsApp, and it dials out, has the conversation, and comes back with a summary and whatever changed — a booked time, a confirmed detail, a message delivered. The work that used to mean picking up the phone yourself just gets handed off. ## What "makes phone calls" actually means Arlo isn't a call-center receptionist that screens every call to your business. It's an assistant that makes the specific calls you'd rather not, and fields the ones that come to its own number: - **Outbound calls it places for you.** Confirm or move an appointment, call a supplier for a quote, follow up on an order, or pass a short message to someone who doesn't check email. - **Calls you make to it.** Call your Arlo number yourself and hand it a task by voice, the same as you would by text. - **A summary every time.** After a call, Arlo writes up what was said and updates the record it's tied to — the CRM note, the calendar event, the thread you asked from. - **Your approval on anything that commits.** Booking, spending, or sending stays behind an approval gate, so a call never quietly agrees to something on your behalf. ## Where a calling assistant earns its keep The phone is still where a lot of real work closes, especially for the roles that live on it: - A [founder](/use-cases/ai-assistant-for-founders) asks Arlo to call three vendors for pricing and come back with a comparison, instead of playing phone tag all afternoon. - A [realtor](/use-cases/ai-assistant-for-realtors) has Arlo confirm the day's showings by phone and text, then flag the one that didn't pick up. - A [recruiter](/use-cases/ai-assistant-for-recruiters) asks it to call a candidate to confirm an interview slot and log the result. - A busy operator forwards a lead to Arlo's number so a first question gets answered in seconds, not hours. ## Arlo vs. other AI executive assistants Most tools in this category stop at email and calendar. Voice-AI platforms can dial but do little else, and they're built to answer your phones, not to be your assistant. | Capability | Typical AI EA (email/calendar) | Voice-AI receptionist | Arlo | | --- | --- | --- | --- | | Drafts email and schedules | Yes | No | Yes | | Places outbound calls for you | No | Rarely | Yes | | You can call it and give it a task | No | Yes | Yes | | Works in iMessage and Slack | Sometimes | No | Yes | | Acts across 3,000+ tools (incl. no-API via browser) | Limited | No | Yes | | Approval gate before it books or sends | Rarely | No | Yes | | Full audit trace of every call and action | No | No | Yes | [Martin](/compare/arlo-vs-martin) can also make calls, so there the difference isn't the phone — it's that Arlo runs inside your team's Slack and keeps a governed audit trail on everything it does. Against email-only tools like an [AI executive assistant](/glossary/ai-executive-assistant) that never leaves the inbox, the phone is the whole gap. ## Governed, so you can hand off the phone Letting an assistant speak on your behalf only works if you stay in control of what it commits to. Every connection Arlo uses resolves through policy before a run. Anything that books, spends, or sends waits for a reviewer you choose. And every call, source, and approval lands in an [audit trail](/glossary/ai-agent-audit-trail) you can read back later. You delegate the call without losing the paper trail. ## Frequently asked questions **Can an AI executive assistant really make phone calls?** Yes. Arlo places live outbound calls to confirm an appointment, chase a vendor, or pass a message, and you can call it yourself to hand it a task. It keeps the transcript of each call, so you can ask how it went. **Does it just read a script, or hold a conversation?** It holds the conversation. Arlo listens and responds in real time, then reports back. For anything that commits you — a booking or a payment — it stops and asks for your approval first. **Is Arlo a receptionist for my business?** No. Arlo places the calls you delegate and takes calls from you, but it doesn't answer or screen other people's calls for your business. If you need a receptionist, a dedicated answering service is the better fit. **Where do I ask it to call?** Anywhere you already talk to Arlo — [iMessage](/imessage), [Slack](/use-cases/ai-assistant-in-slack), or [WhatsApp](/whatsapp). It remembers the context, so "call them back and confirm 3pm" just works. **What happens after a call?** Arlo keeps the transcript and logs the call in its audit trace. It doesn't text you a recap unprompted: ask how it went and it tells you, and it will update the CRM note or calendar event the call was about if you ask. ## Getting started Hand off the first call you've been dreading. [Try Arlo](/signup), text it a number to call, and let it dial out and have the conversation, then ask how it went. --- # AI finance agent — invoices, expenses, and the money chase > An AI finance agent runs the follow-through layer of company finance: chasing overdue invoices by email and live phone call, collecting missing receipts and expenses, reconciling records across tools, flagging spend anomalies, and prepping month-end. Arlo does this from your Slack or text thread — nothing money-adjacent moves without approval, and every action is audit-trailed. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/use-cases/ai-finance-agent Finance work in most companies is two different jobs. One is judgment: what to spend, how to price, when to raise. The other is pursuit: the invoice at day 45, the receipt nobody submitted, the subscription that doubled quietly, the numbers that need assembling every month-end. The pursuit job is relentless, procedural, and mostly about not forgetting — which makes it ideal work for an [AI agent](/glossary/ai-employee), in the one domain where "the AI did something unexpected" is least acceptable. So the design constraint comes first: everything money-adjacent is approval-gated, and everything is logged. ## What Arlo pursues - **Receivables, actually chased** — polite reminder at due date, firmer at 15, a [live phone call](/compare/ai-assistants-that-can-make-phone-calls) at 30 when email stops working. Escalation on your schedule, every touch logged, and you approve the tone tiers once instead of writing each email. - **Expenses and receipts collected** — the monthly hunt for receipts becomes Arlo pinging each person in their own thread until the pile is complete, matched, and filed. - **Records that reconcile** — invoices, payments, and the spreadsheet/books kept consistent across your tools among 3,000+ integrations — including finance portals with no API, worked through a secure browser session. - **Anomaly flags in plain language** — "software spend is up 40% this month — here are the three new charges" beats discovering it at quarter-end. You set the thresholds; Arlo watches them daily. - **Month-end prep** — the assembling, categorizing, and chasing-of-missing-pieces that eats the first week of every month, done before your accountant asks. - **Answers from your thread** — "what's outstanding over 30 days?" "what did we spend on contractors this quarter?" — asked in Slack or a text, answered from your actual records. ## The guardrails are the product Arlo doesn't move money. It chases, records, reconciles, flags, and drafts — and anything consequential (a payment instruction, a customer-facing escalation, a record correction) waits on [your approval](/glossary/human-in-the-loop). Every action lands in a full [audit trail](/glossary/ai-agent-audit-trail): what was sent to whom, what changed, who approved. For finance work that trail isn't a feature — it's the reason you can delegate at all. ## Frequently asked questions **What does an AI finance agent do?** Chases invoices by email and phone, collects expenses and receipts, reconciles records, flags spend anomalies, and preps month-end — with approvals on anything consequential and a full audit trail. **Can it actually call customers about overdue invoices?** Yes — Arlo places real calls with the tone you approved, logs the conversation, and records the outcome. It's the escalation step that email-only tools can't take. **Does it move money or make payments?** No. It pursues, records, and drafts; payment execution stays with humans. That boundary is deliberate. **Does it work with my accounting stack?** Arlo connects to major finance tools among 3,000+ integrations and can operate portals with no API through a secure browser session. Check the in-product list for your specific stack. **Will it embarrass us with a customer?** Every outbound touch follows escalation rules you approved, and anything outside them waits on your sign-off. The audit trail shows exactly what was sent, when, to whom. --- # AI HR agent — people ops without the ticket queue > An AI HR agent handles the execution layer of people ops: running onboarding and offboarding checklists across tools, answering policy questions straight from your handbook, tracking leave and approvals, chasing performance reviews before deadlines slip, and keeping records consistent. Arlo does this in Slack, WhatsApp, or text — sensitive actions approval-gated, every step audit-trailed. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/use-cases/ai-hr-agent People ops in a growing company is a hundred small workflows wearing a trench coat. Onboarding alone spans six tools; "quick" policy questions interrupt someone's whole afternoon; review cycles slip because chasing managers is nobody's favorite job; and every missed step lands on a real person's first day, paycheck, or leave request. It's exactly the shape of work an [AI agent](/glossary/ai-employee) does well — checklist-driven, cross-tool, deadline-bound — in exactly the domain where mistakes are personal, which is why governance can't be an afterthought. ## What Arlo runs for people ops - **Onboarding, end to end** — accounts requested, equipment ordered, intro meetings booked, first-week schedule sent, buddy assigned. One checklist, executed across your tools, with progress visible in your channel and nothing silently skipped. - **Policy questions, answered from your docs** — "how does parental leave work?" gets an instant answer sourced from your actual handbook — not the model's general knowledge — with a link to the section. HR stops being a search engine. - **Leave and approvals routing** — requests captured where people ask (Slack, text), routed to the right approver, recorded in the system, and reflected on calendars. - **Review cycles that finish on time** — Arlo chases every manager with what's missing and when it's due, escalating politely as deadlines approach. The cycle ends without the spreadsheet of shame. - **Offboarding without loose ends** — access revoked, equipment recovered, exit interview booked, records closed — the checklist nobody wants to run manually and nobody can afford to miss. - **Records that agree with each other** — the HRIS, the org chart, and the payroll sheet stop drifting, because updates happen as a side effect of every workflow. ## Sensitive by default, and built for it HR data is the most sensitive thing in the company that isn't code-signing keys. Arlo's model fits: access is **scoped** (the agent touches only the systems and fields you grant), consequential actions are [approval-gated](/glossary/human-in-the-loop) — anything touching compensation, access, or an individual's status waits on a human — and every action lands in an [audit trail](/glossary/ai-agent-audit-trail), which HR of all functions will appreciate the day a question gets asked. Judgment calls — the sensitive conversation, the accommodation decision, anything involving a person's circumstances — route to humans, always. The agent runs process, not people. ## Frequently asked questions **What does an AI HR agent do?** It executes people-ops workflows: onboarding/offboarding checklists, policy Q&A from your own documents, leave routing, review-cycle chasing, and record consistency — with sensitive actions gated to human approval. **Is it safe for confidential HR data?** Arlo's access is scoped to what you grant, consequential actions wait on approval, and everything is audit-trailed. Personnel judgment stays with humans by design. **Can it answer employee questions directly?** Yes — in Slack, WhatsApp, or text, sourced from your handbook and policies with links to the source. Questions it can't ground in your docs escalate to your team instead of getting a guess. **Does it replace an HR person?** It replaces the ticket-queue portion of the job. The human parts of human resources — judgment, care, hard conversations — are explicitly out of scope. **How does it handle onboarding across tools?** It runs your checklist across email, calendars, HRIS, ticketing, and IT systems among 3,000+ integrations — using a secure browser session for tools with no API — and reports progress as it goes. --- # AI marketing agent — campaigns, content, and follow-through > An AI marketing agent handles marketing's execution layer: monitoring ad spend and flagging waste, drafting content and emails in your brand voice, watching competitors, assembling the weekly report, and chasing the follow-ups campaigns depend on. Arlo does this across iMessage, Slack, and WhatsApp — proposing consequential changes for your approval rather than acting alone. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-marketing-agent Marketing has a strategy layer and an execution layer, and the execution layer is winning by volume. Between the ad checks, the content calendar, the competitor you're supposed to be watching, the weekly report, and the follow-ups every campaign generates, the actual thinking — positioning, offers, creative bets — gets the leftovers. An AI marketing agent exists to flip that ratio: it absorbs the execution grind and leaves the judgment to you. The framing matters. An [AI marketing agent](/glossary/ai-employee) isn't a content mill or a black-box optimizer. It's a colleague you brief once — your ICP, your voice, your targets — that then runs the recurring work and brings you decisions instead of chores. ## What Arlo runs for a marketing team - **Ad account vigilance** — daily performance pulled across Google, Meta, LinkedIn, and X, waste flagged, changes drafted for [your approval](/use-cases/ai-agent-for-ads). The full ads workflow has its own guide. - **Content in your voice** — blog drafts, landing copy, email sequences, social posts. Arlo remembers your voice and your ICP across every request; you edit judgment, not grammar. - **Competitor watch** — pricing page changed, new comparison post shipped, funding announced. A short note in your channel when something actually happened, not a weekly digest of nothing. - **The weekly report, assembled** — traffic, spend, pipeline contribution, what shipped — pulled from your analytics, ad accounts, and CRM into one summary that goes where your team reads. - **Campaign follow-through** — the webinar list that needs a same-day follow-up, the influencer who never replied, the landing page ticket that's blocking launch. Chased, on schedule, [including by phone](/use-cases/ai-assistant-you-can-text-and-call) when email stalls. - **Everything, from your thread** — brief it, approve its proposals, and read its reports in iMessage, Slack, or WhatsApp. No new tab in the stack. ## A week with it Monday: the weekly report is waiting in your channel before standup — with a flag that branded CPCs jumped and a drafted keyword change awaiting your tap. Tuesday: you voice-note three content ideas from the car; drafts are in your review doc by afternoon. Wednesday: it notices a competitor shipped a pricing change and proposes an update to your comparison page. Thursday: the event leads get their follow-up sequence — every send having cleared your approval queue in one batch. Friday: you spent the week on the offer redesign, which is the job. ## The first 30 days: what to hand it, in order Marketing agents fail when teams hand them everything at once and drown in approvals. The on-ramp that works: 1. **Week 1 — the report.** Hand it the weekly marketing report first: zero risk, immediate time back, and it forces the tool connections (analytics, ad accounts, CRM) that everything else builds on. 2. **Week 2 — the watching.** Competitor watch and ad-account vigilance. Still read-only; now it's catching things you'd have missed. 3. **Week 3 — drafting.** Content and email drafts in your voice, batch-approved. This is where you calibrate taste — edit hard, it learns from every correction. 4. **Week 4 — the follow-through.** Campaign follow-ups and the [ads approval loop](/use-cases/ai-agent-for-ads). By now the approval cards have earned enough trust that taps are fast. By day 30 the execution layer runs on rhythm, and the only marketing meetings left are about strategy — which was the point. ## Judgment stays human, and that's the point Marketing is public. A wrong send, a tone-deaf post, a budget fat-fingered at midnight — these cost brand, not just money. Arlo's governance model is built for exactly this: **outbound and spend-touching actions wait on approval by default**, everything runs against policy, and every action lands in an [audit trail](/glossary/ai-agent-audit-trail). You get machine diligence on the execution layer without machine judgment on the brand layer. Two adjacent jobs have their own pages: [AI agent for social media](/use-cases/ai-agent-for-social-media) for drafts, scheduling, and listening, and [AI research agent](/use-cases/ai-research-agent) for the sourced competitive and audience research that feeds a campaign. ## Frequently asked questions **What is an AI marketing agent?** An AI colleague that runs marketing's recurring execution — ad monitoring, content drafts, competitor watch, reporting, follow-ups — while routing consequential decisions to you for approval. **Can it manage my ad campaigns?** Yes — it monitors spend across platforms, flags waste, and drafts changes that apply after your approval. See the dedicated [AI agent for ads](/use-cases/ai-agent-for-ads) guide. **Will the content sound like AI wrote it?** It drafts in your voice from persistent memory of your brand, ICP, and past edits — and you approve before anything publishes. The more you edit, the better it matches you. **Does it replace a marketer?** It replaces the execution grind inside a marketer's week. Positioning, offers, creative bets, and taste stay yours — which is precisely what the reclaimed hours are for. **What does it cost?** Arlo starts free — connecting Slack grants your workspace $100 in credits — and then runs on a monthly plan with 10% bonus credits. The rate is 400 credits per $1 at every amount, on monthly plans from $20; $50 buys 20,000 credits, roughly 625 messages or 69 researched answers. Credits roll over and expire 12 months after purchase, and nothing — research, calls, generated images — sits behind a bigger package. Admin controls, SSO, and audit export are available on request. Compare that to the execution hours marketing burns weekly. --- # AI operations agent — the recurring work, actually run > An AI operations agent owns the recurring work every company accretes: the weekly report, the reconciliation, the vendor follow-up, the checklist that must run on the 1st. It executes on schedule across your tools, handles the routine end to end, and escalates exceptions with context. Arlo runs ops from Slack, WhatsApp, or a text thread — approval-gated and audit-trailed. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-operations-agent Operations is the part of the company that must happen again next week. Reports, reconciliations, renewals, orders, checklists, chases — none of it is strategy, all of it is load-bearing, and it quietly consumes the sharpest generalists in the building. The test for an [AI operations agent](/glossary/ai-employee) is simple: can you hand it a recurring job, and stop thinking about that job — without losing the ability to check what happened? ## The jobs it takes - **Scheduled reports that write themselves** — the Monday metrics summary, the month-end pack, the client rollup: pulled from your tools, assembled in your format, delivered where people read. If a number can't be pulled, the report says so instead of guessing. - **Reconciliation and record hygiene** — [the money chase](/use-cases/ai-finance-agent), inventory counts, CRM truth, the spreadsheet that mirrors the tracker. Kept consistent as a side effect of running, not as a Friday penance. - **Vendor and partner follow-through** — the PO that hasn't shipped, the contract renewal in 30 days, the support ticket a supplier went quiet on. Chased by email, and [by phone](/compare/ai-assistants-that-can-make-phone-calls) when email stalls. - **Checklists that cannot be skipped** — onboarding, offboarding, month-open, incident follow-ups: every step executed or explicitly flagged, never silently dropped. - **Exception handling with judgment routing** — the routine 95% runs; the weird 5% (mismatched invoice, angry reply, missing data) escalates to a named human with the context attached. That split — routine automated, exceptions human — is the entire craft of reliable ops automation. ## Why governance is the feature, not the caveat Ops work touches money, customers, and records — exactly where silent automation failures compound. Arlo's model fits ops precisely: scoped access per tool, [approval gates](/glossary/human-in-the-loop) on consequential actions, and a full [audit trail](/glossary/ai-agent-audit-trail) per run — so "what ran, what changed, who approved" always has an answer. You loosen the gates per category as the runs prove boring, which is how trust is supposed to accrue. ## Starting well Hand it one recurring job with a clear definition of done — the weekly report is the classic first move. Watch two cycles, loosen the approval, add the next job. Teams that scale this way end up with an ops layer that runs itself, with humans holding only exceptions and design — which is what ops leadership was supposed to be. ## Frequently asked questions **What is an AI operations agent?** An AI that owns recurring operational work — scheduled reports, reconciliations, vendor chases, checklists — executing the routine and escalating exceptions to humans with context. **How is it different from workflow automation like Zapier?** Rule-based automation breaks on anything it wasn't scripted for. An ops agent handles the messy middle — unstructured replies, judgment calls about escalation, tools without APIs — and reports in plain language. **Can it work with our legacy systems?** Yes — for software with no API, Arlo logs in through a secure browser session and operates the interface like a person, including [bulk data entry](/use-cases/ai-agent-for-data-entry). **What happens when something goes wrong mid-run?** Exceptions stop and escalate with context rather than pushing through. Every run is logged, so failures are visible, not silent. **What does it cost?** Arlo starts free — $100 in credits for connecting Slack — then runs on monthly plans from $20 at 400 credits per $1 at every amount — $50 buys 20,000 — and credits last 12 months — so a quiet month costs nothing. Weigh that against the generalist hours ops currently consumes. --- # AI research agent — sourced answers, not link dumps > An AI research agent takes a real question — a market, a prospect, a vendor comparison, a regulation — works the open web and your own tools, and comes back with a sourced answer you can check, not ten blue links. Arlo runs research from the thread you're in (iMessage, Slack, WhatsApp), cites what it found, says what it couldn't verify, and files the result wherever your team keeps knowledge. Updated: 2026-08-08 · Canonical: https://www.arlo.sh/use-cases/ai-research-agent Most "research" at work is not a literature review — it's an hour of tabs. Who are this prospect's competitors? What does this vendor actually charge? Is this regulation going to touch us? What happened in our space this week? The information exists; the hour doesn't. An AI research agent absorbs exactly that class of work: real questions, worked properly, returned as answers with receipts. The bar that separates a research agent from a chatbot is **sourcing**. A chatbot answers from memory, confidently, sometimes wrongly. A research agent goes and looks — and shows you where it looked, so trust never has to be blind. ## What Arlo researches - **Prospects and accounts** — before the call: what the company does, who's who, recent news, stack hints, the competitor they probably use. Delivered as a brief in your thread, [feeding the pipeline](/use-cases/ai-agent-for-lead-generation) directly. - **Markets and competitors** — pricing pages, positioning shifts, funding, feature launches. One-off deep dives, or a standing watch that pings your channel when something actually changed. - **Vendors and tools** — "compare these three payroll providers for a 12-person US team" returns a table with real prices, real limitations, and a recommendation you can interrogate. - **Questions with stakes** — the regulation, the tax rule, the platform policy. Arlo reads the primary source, quotes it, links it, and flags where interpretation ends and a professional should begin. - **Your own tools** — half of "research" is internal: what did we say to this customer last quarter, which deals mentioned this feature, what did the retro conclude? Arlo searches your connected tools alongside the web, in one pass. ## What a good answer looks like One message, three parts: **the answer** (direct, opinionated where you asked for judgment), **the evidence** (sources linked inline, quoted where wording matters), and **the honest edges** (what couldn't be verified, what's behind a paywall, what changed recently enough to re-check). If a claim matters to a decision, you can tap through and see it yourself — and because every run keeps its [audit trail](/glossary/ai-agent-audit-trail), the sources are still attached when someone asks "where did this number come from?" three weeks later. ## Research that lands somewhere A research agent that answers in chat and evaporates is half a tool. Arlo files results where your team will find them again — the CRM record, the Notion page, the deal doc — as a side effect of answering. Ask once, keep forever. And for recurring questions ("what changed in our space this week?"), the research becomes a standing brief delivered on schedule, not a task someone has to remember. ## Frequently asked questions **What is an AI research agent?** An AI that works real questions — markets, prospects, vendors, regulations — across the open web and your connected tools, and returns sourced, checkable answers instead of link lists. **How is it different from asking ChatGPT?** Grounding and delivery. Arlo researches live sources and your own tools, cites what it found, flags what it couldn't verify, answers in the thread your team already uses, and files the result where it belongs. **Can it monitor a topic continuously?** Yes — standing watches on competitors, keywords, or market topics that report to your channel when something genuinely changed, not on a timer full of nothing. **Does it make things up?** It's built not to: answers are grounded in retrieved sources, uncertainty is stated, and unverifiable claims get flagged rather than smoothed over. The sources stay attached for checking. **What does it cost?** Arlo starts free — $100 in credits for connecting Slack — then runs on monthly plans from $20 at 400 credits per $1; $50 buys 20,000 credits — roughly 69 researched answers with sources. Deep research isn't gated behind a bigger package, and credits last 12 months from purchase. --- # AI sales agent — pipeline follow-through, not just outreach > An AI sales agent works your pipeline instead of just spraying outreach: it keeps the CRM updated as a side effect of real activity, chases every follow-up before it goes cold, preps you for each call, drafts personalized outreach for your approval, and can pick up the phone. Arlo does this from iMessage, Slack, or WhatsApp — approval-gated, so nothing reaches a prospect without your sign-off. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/use-cases/ai-sales-agent "AI sales agent" usually conjures one image: a bot blasting sequences at strangers. That version exists, and buyers have learned to smell it. The version that actually moves revenue is less glamorous — it's the agent that makes sure nothing in your pipeline dies of neglect. Deals are lost to silence far more often than to competitors: the demo follow-up that went out three days late, the champion who changed jobs unnoticed, the proposal sitting unopened with no nudge. An AI sales agent built as a [colleague](/glossary/ai-colleague) — rather than a spam cannon — works that neglect layer. For the team-wide picture, see the companion guide to an [AI assistant for sales teams](/use-cases/ai-assistant-for-sales); this page is about the agent working the pipeline itself. ## What Arlo does in your pipeline - **CRM hygiene as a side effect** — after the call or the thread, Arlo updates the record: stage, next step, notes. Reps stop doing data entry; managers stop reading fiction in the pipeline review. - **Follow-ups that never slip** — every open deal has a next touch; when a thread goes quiet past its window, Arlo drafts the nudge and queues it for your approval. Silence stops being a stage. - **Meeting prep in your thread** — before each call: who they are, what happened last time, deal state, what you promised. Delivered by text, where you'll actually see it. - **Outreach drafts, approval-gated** — personalized from the CRM, your notes, and the prospect's actual context. Nothing sends without you; volume never outruns judgment. - **The phone, when it's faster** — [Arlo places real calls](/compare/ai-assistants-that-can-make-phone-calls): confirming the meeting, chasing the missing signature, reaching the prospect who never answers email — then logs the outcome to the CRM. - **Pipeline answers on demand** — "what's stalled over 14 days?" "what closed this week?" — asked in Slack, answered from the CRM among 3,000+ integrations. ## Agent vs sequence tool | | Sequence/outreach tools | Arlo | |---|---|---| | Sends bulk cadences | ✓ Their whole job | Drafts, you approve | | Keeps CRM true to reality | — | ✓ Side effect of activity | | Chases existing pipeline | — | ✓ Core job | | Preps you for calls | — | ✓ In your thread | | Makes real phone calls | — | ✓ | | Governance | Send-first | Approval-first, full [audit trail](/glossary/ai-agent-audit-trail) | Sequence tools have a job at the very top of the funnel. The agent's job is everything after a human enters the picture — where deals are actually won and lost. ## Why approval-first sells better Every AI-written email that reaches a prospect is your reputation on loan. Arlo's default — draft, queue, approve with a tap — means personalization at machine scale with a human's name honestly on it. Teams typically batch-approve routine touches in seconds and hand-edit the ones that matter; either way, buyers never get the smell of automation, because a person actually signed off. Selling is only half the revenue loop. Once a deal closes, the same colleague can run the coverage layer behind it — see [AI agent for customer success](/use-cases/ai-agent-for-customer-success) for renewal runway, health signals, and QBR prep. ## Frequently asked questions **What is an AI sales agent?** An AI that works your pipeline: CRM updates as a side effect, follow-ups on schedule, meeting prep, approval-gated outreach drafts, and real phone calls — rather than just bulk sequences. **Will it spam my prospects?** No — nothing reaches a prospect without your approval by default. Arlo drafts; you sign off. That gate is the difference between personalization and spam. **Can it update Salesforce or HubSpot?** Arlo connects to major CRMs among 3,000+ integrations and updates records as a side effect of calls, threads, and tasks — see the [HubSpot guide](/use-cases/ai-assistant-for-hubspot) for that stack. **Can it actually call prospects?** It places real calls for meeting confirmations, chases, and reachability — and it keeps the transcript, so you can ask how it went or have the outcome logged to the CRM. Cold-call campaigns aren't the job; follow-through is. **How is this different from your AI assistant for sales page?** That guide covers the team-wide assistant (channels, reporting, coordination). This one covers the agent working pipeline mechanics. Same Arlo underneath — the pages exist because people search for both. --- # AI scheduling assistant — calendar without the back-and-forth > An AI scheduling assistant manages your calendar like a human assistant would: it finds times, sends invites, handles the reschedule ping-pong, guards your focus blocks, and briefs you each morning. Arlo does this from the channels you already use — text it from iMessage or Slack and it negotiates the rest, asking approval before invites go out. Updated: 2026-07-23 · Canonical: https://www.arlo.sh/use-cases/ai-scheduling-assistant Scheduling is the purest coordination tax in work: nobody's job, everybody's problem, and every meeting costs three emails before it exists. Scheduling links solved a slice of it — for meetings where you're comfortable sending a booking page. Everything else still turns into ping-pong: the six-person sync, the external partner who "can do Tuesday, actually no," the double-booked Thursday that needs triage. An AI scheduling assistant handles the whole category the way a human [executive assistant](/glossary/ai-executive-assistant) would — by owning the back-and-forth, not by handing the other person a form. ## What Arlo handles - **Finding time** — "Get me 30 minutes with Dana and Priya next week" → Arlo checks calendars, proposes options, negotiates by email if needed, and sends the invite once it's settled. - **The reschedule ping-pong** — when someone bails, Arlo re-runs the negotiation and fixes the chain reaction, not just the one slot. - **Guarding focus** — tell it your rules once ("no meetings before 10, keep Fridays clear") and it enforces them in every negotiation, forever. That's what persistent memory is for. - **The morning brief** — one message each morning in your texts: today's meetings, what moved overnight, what conflicts need a call. - **External niceties** — right tone with a customer, right firmness with a vendor, time zones handled without comment. - **When email stalls, the phone works** — for the clinic, the venue, the contractor who never answers email, Arlo can [simply call](/use-cases/ai-assistant-you-can-text-and-call) and book it. The interface is the channel you're already in. Text from iMessage, ask in Slack or WhatsApp, or say it on a call. No new app; the calendar just starts behaving. For a meeting with several people, where the work is asking everyone and chasing the ones who don't answer, see [an AI that schedules meetings with several people](/use-cases/ai-that-schedules-meetings-with-multiple-people). ## Scheduling links vs a scheduling assistant | Situation | Booking link | Arlo | |---|---|---| | Inbound "grab time with me" | ✓ Great | ✓ | | Multi-person internal sync | — | ✓ | | External meeting with a VIP you'd never send a link | — | ✓ | | Reschedules and conflict triage | — | ✓ | | Enforcing your calendar rules | Static rules | ✓ Learned, contextual | | Booking things that need a phone call | — | ✓ | Links and assistants coexist fine — keep the link for inbound requests, and hand everything with more than one moving part to the assistant. ## Trust, because your calendar is your day Calendar mistakes are expensive: a wrong invite embarrasses you in front of exactly the people you scheduled with. Arlo's governance applies to scheduling like everything else — invites and external emails wait on your approval until you loosen the gate for routine cases, and every action lands in the [audit trail](/glossary/ai-agent-audit-trail). The practical result is that you can hand over the calendar in week one without handing over the ability to embarrass you. ## Frequently asked questions **What is an AI scheduling assistant?** An AI that owns the work around your calendar — finding times, sending invites, negotiating reschedules, enforcing your rules, and briefing you daily — rather than just displaying a booking page. **How is Arlo different from Calendly?** Calendly hands your availability to the other person as a form. Arlo does the negotiating itself — including multi-person syncs, VIPs, reschedules, and bookings that require a phone call — and asks your approval before invites go out. **Can I use it entirely over text?** Yes. Text scheduling requests from iMessage or SMS like you'd text an assistant; Arlo handles the rest and confirms in the same thread. Slack, WhatsApp, and calls work too. **Will it move meetings without asking?** No — changes that touch other people wait for your approval by default. You can loosen the gate for routine internal moves once you trust it. **Does it handle time zones?** Yes — proposals and invites resolve each participant's time zone automatically, which is table stakes for an assistant negotiating across companies. --- # An AI that schedules meetings with several people for you > An AI scheduling assistant can set up a meeting with several people by checking your calendar, offering open times to everyone, collecting their answers and booking the slot that works. Arlo does this from a text or a Slack message, for up to 20 people. It emails them from your own inbox or texts them, and asks you before it reaches out and again before it books. Updated: 2026-09-23 · Canonical: https://www.arlo.sh/use-cases/ai-that-schedules-meetings-with-multiple-people Getting one person on a call is easy. Getting five is a project. Someone can do Tuesday but not the morning, someone never answers, and by the time three people agree the fourth has filled the slot. A booking link doesn't fix it, because a link only works when one person picks a time for themselves. With a group, somebody still has to collect everyone's answers and find the overlap. That somebody can be an assistant you text. This page explains how Arlo does it, what it asks you along the way, and where it stops. ## How it works 1. **You text the request.** Something like *"set up a 45-minute meeting with Dana, Priya and Sam next week about the Q4 plan"*, with their emails or phone numbers if Arlo doesn't already know them. You can do this in iMessage or in a Slack DM. 2. **Arlo looks at your calendar.** It reads your Google Calendar or Outlook and offers up to eight open times on weekdays between 9 and 5, no more than two a day (one before 3 pm, one after), within the next two weeks. 3. **It shows you the plan before anything goes out.** You see who it will contact, from which account, and the times it will offer, and nothing is sent until you reply yes. 4. **It asks everyone.** People with an email address get the message from your own Gmail or Outlook, so it comes from you. People you only have a phone number for get a text. By default, for more than three people it sends each one a private poll link where they tick the times that work; for a smaller group it sends each person a message listing the times. 5. **It follows up once.** Anyone who hasn't answered gets one reminder with the poll link. 6. **It books the time, after one more yes.** When a slot works for everyone, Arlo asks you to confirm, checks your calendar again for conflicts, then puts the event on your calendar with invites to everyone who has an email address, and sends each person a short confirmation, by text for anyone you only have a number for. ## What it won't do yet - It doesn't see other people's calendars. It asks them, the way a person would. Even if your company shares calendars, it still asks each person and waits for their answer. - It doesn't reschedule. Once the meeting is booked, changes are up to you. If no time suits the whole group, it stops and asks you what to do. - Video links depend on your calendar. With a work or school Outlook calendar it can add a Microsoft Teams link. On Google Calendar it doesn't add a Meet link, so add one to the event yourself. - It handles up to 20 people, meetings of 5 minutes to 8 hours, and you start it from iMessage or a Slack DM. ## When a link or a poll is all you need If one person is booking time with you, a scheduling link such as Calendly is faster than any assistant. If you just want to see when a group is free and don't mind chasing people yourself, a free poll like Doodle or When2meet does that. An assistant earns its keep when you'd otherwise be the one writing the message, reminding the person who didn't answer, and sending the invite at the end. See [AI scheduling assistant](/use-cases/ai-scheduling-assistant) for the rest of what Arlo does with your calendar. ## Frequently asked questions **Can an AI schedule a meeting with several people?** Yes. Arlo checks your calendar, offers open times to everyone by email, text or a private poll link, follows up once with anyone who hasn't answered, and books the time that works once you approve it. It handles up to 20 people. **Does it need everyone's calendar?** No. It only reads your calendar, Google Calendar or Outlook, and asks the other people which times work for them. **Will it send messages without asking me?** No. It shows you who it will contact and the times it will offer, waits for your yes, and asks again before it books the meeting and sends the invites. **Can it add a video call link?** With a work or school Outlook calendar it can add a Microsoft Teams link. On Google Calendar it doesn't add a Google Meet link, so you'd add one yourself. **What happens if someone can't make it after it's booked?** Arlo doesn't reschedule. Move or cancel the meeting yourself in your calendar.