What are OpenAI dots and what are they good for at work?

A dot is a persistent agent in ChatGPT that takes ongoing responsibility for a project rather than answering one prompt at a time. OpenAI says each dot runs on GPT-6 Astra, has its own cloud computer and browser, learns from feedback and can connect to more than 4,000 apps through its plugin ecosystem.

OpenAI's help centre suggests starting with a project such as preparing a presentation as new data arrives, turning an interview transcript into show notes and posts, or scoping and testing a change to an app. The common thread is work with a clear goal that unfolds over days. You can reach your dot in ChatGPT on desktop, web and mobile, and in Slack and Microsoft Teams.

Who can use ChatGPT dots?

According to OpenAI's help centre, dots are rolling out gradually to Pro users outside the European Economic Area, Switzerland and the UK, and to Business Premium users in all supported ChatGPT regions. Enterprise, Edu and Healthcare workspaces get a beta that an admin must enable. Dots are not available to anyone under 18.

If you cannot find it, open chatgpt.com/dots from the intended account. On a work account, your administrator also controls whether you may use dots, connect your local computer, add the dot to Slack or Teams and write custom rules.

How do I get started with my dot?

Create your dot in the ChatGPT desktop app, on desktop web or in the latest ChatGPT mobile app for iOS or Android, then follow the onboarding prompts. Dots are not supported on mobile web. You can give it a name during setup.

Your dot starts with context from ChatGPT memory. Connected apps are managed under Customize, then Plugins, with permissions shared across ChatGPT, ChatGPT Work and Codex. Access to your own laptop is optional and starts turned off.

What is a good 10-minute starter exercise for a dot?

Use a fictional 12-person events agency, Larkspur Events, that sends one client a weekly status note. Write a short set of mock meeting notes yourself so nothing real is involved.

Minutes 1 to 3: open Customize, then Custom rules, and add a rule for sending email set to Ask before taking action. Minutes 4 to 6: select + in the conversation, attach the mock notes and ask your dot to draft the weekly status note with sections for done, next and decisions needed, quoting the note each point came from. Minutes 7 to 8: ask it to schedule a check every Friday morning that reminds you to review the draft, then confirm the schedule appears under Scheduled in its profile. Minutes 9 to 10: open the task, compare three points with your notes and correct anything wrong inside that task.

How do I get better results from a ChatGPT dot?

These techniques come from OpenAI's own guidance.

Approve precisely

OpenAI advises stating who a message goes to, what it says and when it should be sent. Approving one message does not give ongoing permission to contact people.

Correct inside the task

Each task has its own conversation and does not automatically see every chat you have had with your dot, so put corrections, files and links in the affected task.

Ask for drafts first

OpenAI notes that asking for a draft does not authorise sending it, so a draft request is a safe default for anything external.

Write narrow Custom Rules

Choose take action without asking, take action if pre-approved, ask before taking action or hand off to you, and describe the exact action each rule covers.

What should I check before letting a dot act for me?

Follow your employer's AI policy first, and connect only the apps it allows. Then check which plugins the dot can use and whether each allows only reading or also changing and sending, your Custom Rules, whether local computer access is on and what is scheduled.

OpenAI says background proactive research uses read-only tools that cannot send messages, change app content or control a computer. Auto-review checks actions that could affect your accounts or share information, and the most sensitive steps, such as changing a password or transferring money, always come back to you.

On data, OpenAI states that Business, Enterprise and Edu content is not used to train its models by default. On personal plans, the Improve the model for everyone setting decides, and OpenAI says it does not train directly on proactive research or the dot's private notes. Limited human review, including for safety, can still occur.

Where do OpenAI dots fall short?

Memory control is coarse. OpenAI says you cannot yet view or delete individual dot memories, disconnecting an app does not delete what the dot learned from it, and only Reset clears its context, deleting its conversations, memories and scheduled tasks.

Protection against prompt injection reduces risk but does not remove it, some completed actions cannot be undone and OpenAI says dots can make mistakes, even when following your rules.

How does Tosheo.ai teach OpenAI dots?

Tosheo.ai covers agents like dots in its AI Agents Development category: what an agent is, how permissions and approvals work and how to review agent output, with no-code practice first and optional Code lessons in the AI agents development course for building agents yourself. It is not a bootcamp or a job guarantee.

Each 10 to 15 minute lesson ends with a task, such as writing a delegation brief and rule set for a fictional team, marked against a visible rubric. Start with the 7-day free trial; no credit card is needed.

Sources and scope

OpenAI, Introducing dots, checked 19 October 2026 ↗

OpenAI Help Center, Getting started with your dot, checked 19 October 2026 ↗

OpenAI Help Center, Dots privacy, security, and safety FAQs, checked 19 October 2026 ↗

OpenAI Help Center, Manage dots in ChatGPT workspaces, checked 19 October 2026 ↗