OpenAI dots: what they are and how to hand one your first workflow

What a dot actually does

• It has its own computer. Every dot works on a cloud computer with its own browser. You can open that computer at any point to see what it is doing.
• It reaches 4,000+ apps through ChatGPT's plugin ecosystem, and you choose which ones it may use.
• It keeps going after you leave. A dot works toward your goals 24/7, including once the chat window is closed, and it picks up your preferences from the feedback you give it.
• You can reach it where you work: ChatGPT on desktop, web and mobile (by message or voice call), Slack and Microsoft Teams. Texting is listed as coming soon. Context follows it across these channels, and it messages you with progress, questions or decisions.
• It can act, with your OK. In OpenAI's own example, an early tester's dot spotted an invoice he had forgotten to send to a publication, prepared it, and sent it once he approved.

Set up your first dot

1. 01Open ChatGPT on a computer

   Create your first dot in the ChatGPT desktop app or in a desktop browser. The phone comes in after setup.

2. 02Create your dot and name it

   Your primary dot gets a name you choose.

3. 03Connect only the apps this workflow needs

   Use ChatGPT's app controls to pick what it can reach. Start with the two or three apps your first workflow touches, not everything you have. Then let it introduce itself.

4. 04Set your Custom Rules

   Built-in rules already decide when it acts alone and when it asks. Custom Rules let you go further: allow specific actions, require approval for others, or block them outright.

5. 05Give it one outcome-shaped goal

   Describe the finished result you want, not a list of clicks. One clear outcome beats five vague chores.

6. 06Watch Activity View, then go mobile

   Activity View shows its progress, background work included, and lets you redirect it. Once setup is done, you can message your dot from the ChatGPT mobile app.

How to pick the first workflow to hand off

Pick one repeatable outcome with clear inputs: something you do every week, where the sources are known and a good result is easy to recognise. A weekly client update is a strong first candidate:

1. 01Read this week's client threads in your inbox
2. 02Pull the numbers the update needs
3. 03Draft the update
4. 04Add notes to the CRM
5. 05Prepare the follow-up email and wait for your OK before it goes out

An example goal you can adapt

One outcome, clear inputs, and the send held for your approval. Swap the [brackets] for your own client, sources and CRM.

Prepare this week's client update for [client name]. Read this week's emails from [client] in my inbox, pull the numbers from [report or sheet], draft the update, add a short note to [client] in [CRM], and draft the follow-up email. Do not send anything: ask me to approve the email first.

Keep the risky parts with you

• Keep customer-facing sends and anything involving money behind require-approval in your Custom Rules.
• Start with read-only or reversible work: reading, summarising, drafting, preparing. Let it send or change things only after you have watched a few runs.
• Add apps and permissions one at a time, as the workflow proves itself.

Safeguards and limits

• Proactive research is read-only. When a dot is idle it looks through your connected apps for ways to help, but those tools cannot send messages, change app content or control your browser or computer.
• Auto-review checks actions that could affect your accounts or share information.
• Some tasks always stay with you, such as changing a password.
• Your computer stays separate unless you choose to connect it. You can give a dot permission to use your laptop, but it is off until you do.
• It can sign in to supported sites with saved passwords without the model seeing them.
• Dots can still make mistakes. Review anything consequential before it counts.

Who can get it and what it costs

• Rolling out in ChatGPT to Pro and Business Premium users in eligible markets.
• Enterprise workspaces (including Edu and Healthcare) can try the beta once the admin enables it.
• Your first dot is included in Pro and Business Premium at no extra cost, with an allowance for deeper work and extended limits in the first month.
• Talking to your dot does not count toward your ChatGPT usage limits. Tasks it starts in Codex or ChatGPT Work do count.
• Business, Enterprise and Edu content is not used for training by default; personal plans can control this setting.

Official links

• OpenAI: Introducing dots (announcement)openai.com/index/introducing-dots ↗

Sources checked

The facts on this page were checked against OpenAI's announcement on 27 October 2026 and restated in our own words. Plans, limits and availability can change, so treat OpenAI's page as the final word.

• Announced 29 Sep 2026; GPT-6 Astra; own cloud computer and browser; 4,000+ apps; ChatGPT, Slack and Teams; Custom Rules, Activity View and auto-review; plans, pricing and usage limits; setup steps. Source ↗
• Independent coverage agreeing on the date, GPT-6 Astra, the cloud computer, 4,000+ apps, Slack/Teams and Pro + Business Premium. Source ↗
• Reworked: OpenAI introduces always-on dots agents. Source ↗
• The Hill: OpenAI's always-on AI agent. Source ↗