OpenAI dots: ChatGPT agents that work all day

On September 29, 2026, OpenAI introduced dots: AI agents that, in the company's words, are highly capable, always on and designed to take care of everything. The idea is that ChatGPT stops waiting for your questions and starts moving your goals forward on its own, all day long.

What a dot is

A dot is an agent with its own computer and its own browser in the cloud. According to OpenAI, it runs on the GPT-6 Astra model, learns from your feedback and can use the apps you connect, with more than 4,000 available through plugins.

What matters is how it works with you:

You talk to it where you already work. From ChatGPT on desktop, web or mobile, from Slack or Teams, and even by voice call. OpenAI says SMS support is coming soon.
It handles several projects at once. It reports progress, asks questions and requests decisions when it needs them.
It researches on its own. It can work in the background with read-only tools, without sending messages or changing anything in the meantime.
It can use your laptop, if you give it permission.

What it is good for

The examples OpenAI published are concrete:

Reviewing customer feedback, building the fixes and delivering them ready for review, with a video showing the change.
Adjusting launch materials when a product's scope changes.
Re-running an analysis when new data arrives and flagging what the researcher should check.
Updating sales proposals and test plans when customer requirements change.
Spotting good moments for clips and drafting social posts that the user approves before publishing.

The pattern is clear: tasks that used to require someone to watch for a change and then spend hours updating documents, code or reports.

Control and safety

OpenAI focused on keeping the user in control:

Action rules. Built-in and custom rules define what a dot can do on its own, what needs your approval and what is blocked.
Activity view. You can follow its progress and check its computer at any time.
Automatic review. The system checks that actions affecting your accounts or data follow your instructions. If it detects a security issue, it can pause or stop the work.
Protected passwords. It can sign in with saved passwords without exposing them to the model. Sensitive tasks, such as changing a password, always stay in your hands.
Company data. On Business, Enterprise and Edu plans, content is not used to train models by default.

OpenAI itself warns that dots can make mistakes and recommends always reviewing work that could have important consequences.

Availability and price

Dots are rolling out gradually to Pro, Business Premium and Enterprise plans in supported countries. The first dot is included in Pro and Business Premium at no extra cost, with a usage allowance that is expanded during the first month. Conversations with your dot do not count against your ChatGPT limits, although the tasks it hands off to other OpenAI tools do.

For organizations there is also a preview of specialized dots, with their own identity, credentials and hardware, integrated with company systems. OpenAI is also working with Microsoft to integrate them with Agent 365.

How to start sensibly

As with any agent, the value appears when the task is well chosen:

Start with a specific, recurring process that today depends on someone keeping an eye on it.
Configure from day one which actions require approval.
Review its results during the first weeks before giving it more autonomy.
Measure the time it frees up and the errors it prevents.

Dots arrived less than two months after xAI's Grok Bot. Two of the biggest companies in the industry launching agents with their own computer almost at the same time says a lot about where work is heading. We look at it in how AI agents are changing work.