OpenAI has introduced dots, a new type of always-on AI agent designed to keep working on users’ goals after they leave the conversation. Powered by GPT-6 Astra, each dot gets its own cloud computer and browser, allowing it to work across connected apps, continue projects and return to users when something requires their attention.
OpenAI announced dots at DevDay on September 29, positioning them as a shift from asking an AI to complete one prompt at a time toward giving an AI an ongoing responsibility.
Dots are rolling out in ChatGPT across web, mobile and desktop for eligible users on Pro, Business Premium and Enterprise plans in supported markets. Enterprise customers need their workspace administrator to enable access.
What are OpenAI Dots?
A dot is designed to act more like an ongoing digital assistant than a conventional chatbot.
Users can give a dot a goal or responsibility and then step away. The agent can determine the next steps, work through a task, use connected tools and return with progress, questions or decisions that require human approval. OpenAI says dots can continue working between conversations rather than stopping when the user closes ChatGPT.
That changes the basic interaction model.
A conventional chatbot waits for the next prompt. A dot can be given an objective and continue pursuing it.
For example, OpenAI demonstrates a dot updating a leadership presentation as new information becomes available. It can examine sales figures, compare them with customer feedback, investigate changes and update the presentation for review.
Each Dot Gets Its Own Cloud Computer
One of the more significant technical details is that dots are not simply background conversations running inside ChatGPT.
Each dot operates on its own cloud computer and browser. That gives the agent an environment in which it can carry out multi-step work and interact with software rather than simply generate text.
OpenAI says dots can use connected tools and more than 4,000 apps through its plugin ecosystem. Users decide which apps a dot can access and what it is allowed to do.
This creates a practical distinction between an AI that tells someone how to perform a task and an AI that can actually carry out portions of the workflow.
A user could, for example, ask a dot to monitor a project, work through related information and prepare materials, rather than repeatedly returning to ChatGPT to provide instructions.
Dots Can Use Memory and Learn a User’s Working Style
OpenAI says dots begin with context from a user’s ChatGPT memory. They can also use Codex and connected tools, depending on the permissions given to them.
The intention is for a dot to become more useful as it receives feedback.
OpenAI’s examples include content creators giving a dot responsibility for finding interview moments, drafting show notes and preparing social posts. The agent can carry edits and preferences across subsequent work.
That makes persistent context one of the product’s central features.
Instead of starting every task from scratch, the agent can retain an understanding of how the user wants certain work handled.
Users Remain in Control of Important Actions
The always-on model also creates a new problem: an agent that can keep working needs clear limits.
OpenAI says dots have permission controls that allow users to select connected apps, establish custom rules and determine which actions require approval. Users can also decide whether a dot should connect to their own computer.
OpenAI also says actions that could affect accounts or share information go through an automatic review process that checks the action against the user’s instructions, custom rules and safety requirements.
This is an important difference between autonomous agents and ordinary AI assistants. The value of an agent depends partly on what it can do without intervention, while the risk increases when it has permission to make consequential changes.
Dots therefore combine autonomy with permission boundaries rather than giving an agent unrestricted access.
Dots Can Work Across ChatGPT, Slack and Teams
OpenAI says users can communicate with their dots through ChatGPT on the web, mobile and desktop. Dots can also be brought into Slack and Microsoft Teams, allowing work to continue inside collaboration environments. Text messaging support is planned for a later release.
For businesses, that could make dots relevant to workflows that already move between communication tools, documents, research and operational software.
Instead of creating a separate AI workflow outside an existing team environment, an agent can participate where employees already communicate.
What Makes Dots Different from Ordinary ChatGPT?
The biggest difference is persistence.
ChatGPT generally responds to a user’s request. A dot is designed to take ownership of an ongoing objective.
That difference can be illustrated simply:
| Traditional AI interaction | OpenAI dot |
| User asks a question | User gives the agent a responsibility |
| AI responds | Agent works toward an objective |
| User initiates the next step | Agent can continue between conversations |
| Work usually ends with the response | Work can continue over time |
| Tools are used when requested | Connected tools can become part of an ongoing workflow |
| User reviews the output | Agent can return with work, questions or decisions |
The change is therefore less about another chatbot interface and more about persistent AI workers operating in the cloud.
Who Can Use OpenAI Dots?
OpenAI says dots are rolling out for Pro, Business Premium and Enterprise users in eligible markets. Enterprise users can access them when their workspace administrator enables the feature.
OpenAI’s product page describes dots as available across web, mobile and desktop, while emphasizing that the feature is being rolled out rather than presented as universally available to every ChatGPT account.
For companies, administrator controls will therefore be an important part of deployment. Organizations need to determine which employees can use the agents, which applications they can access and what actions require approval.
Why Dots Matter for the Next Phase of AI Agents
The introduction of dots reflects a broader change in how AI products are being designed.
The first generation of consumer AI assistants focused heavily on generating answers. Newer agent systems are increasingly designed to take actions, use software and manage multi-step work.
OpenAI‘s dots push that model further by making the agent persistent. The goal is no longer simply to generate an answer when a person is present. It is to keep working toward an objective when the person is doing something else.
That could make AI useful for recurring research, content production, project monitoring, presentation updates, administrative work and other processes that require several steps.
The practical test, however, will be how reliably dots handle those workflows, how much human oversight they require and how effectively users can control their access to personal and business systems.
For OpenAI, dots represent a move toward AI that operates less like a tool people open when they need help and more like a digital coworker that remains available in the background.
