OpenAI Introduces Dots, Always-On AI Agents That Can Work on Projects While You’re Away
OpenAI has unveiled Dots, a new generation of persistent AI agents built to do more than answer questions in a chat window. Instead of waiting for constant prompts, Dots are designed to take ownership of multi-step projects, work across connected apps, and continue making progress even when the user is offline.
Each Dot operates with its own cloud computer, browser, and access to approved applications. This means it can research, organize, draft, test, update, and prepare work independently while still asking for human approval when needed. OpenAI’s goal is to move AI assistants beyond simple conversations and turn them into reliable digital workers that can manage ongoing tasks from start to finish.
Dots are powered by GPT-6 Astra and can connect with more than 4,000 applications through OpenAI’s plugin ecosystem. Users can communicate with them through ChatGPT, Slack, Microsoft Teams, or voice commands. A Dot can also message the user when it needs clarification, additional information, or approval for an important action.
One of the biggest advantages of Dots is persistent memory across a project. Instead of losing context between chats or tools, a Dot can remember the project’s goals, user preferences, working style, and quality standards. Over time, OpenAI expects these agents to become more personalized and useful as they learn how each user likes work to be done.
OpenAI shared several examples of how Dots could be used in real-world workflows. For software developers, a Dot could monitor customer feedback, identify recurring bugs, create and test fixes, and prepare a pull request for review. It would not replace the developer’s judgment but could handle much of the repetitive work that slows teams down.
For product teams, Dots could keep launch documents, internal notes, and customer-facing materials up to date as requirements change. If a feature is delayed or updated, the Dot could revise related documents and flag anything that needs approval.
Researchers could use Dots to rerun data analysis when new information becomes available. The agent could update charts, summarize findings, and draft an email explaining the results. Content creators could use a Dot to turn interview transcripts into video clips, show notes, social media posts, and follow-up drafts, then refine them after feedback.
Sales teams may also benefit from the new AI agents. A Dot could review customer requirements, check account history, build a proof of concept for an integration, and update a proposal as the deal evolves. OpenAI emphasizes that users remain responsible for reviewing final work, approving commitments, and making important decisions.
Although Dots are built to act independently, OpenAI says user control is central to the system. By default, each Dot works inside its own cloud-based environment. A user’s personal computer, local files, and private data stay separate unless the user specifically grants access.
Dots can also use saved passwords to sign in to supported websites without revealing those passwords to the AI model. This allows the agent to work across services while reducing the risk of exposing sensitive login details.
When users are not actively interacting with a Dot, it can perform what OpenAI calls proactive research. In this mode, the agent looks for useful background tasks that may help move a project forward. However, proactive research is limited to read-only access. A Dot cannot send messages, edit content in connected apps, or control the user’s computer during this mode.
Users decide which applications each Dot can access. They can also create rules that automatically allow certain actions, require approval for others, or block specific actions entirely. OpenAI says actions that may affect accounts or expose information are reviewed by its system, while highly sensitive tasks, such as changing a password, must always be handled directly by the user.
OpenAI is also previewing specialist Dots for organizations. These enterprise-focused agents would have dedicated identities, credentials, and access to specific company systems. Businesses could assign them clearly defined roles such as procurement, invoice processing, email marketing, customer support, or commercial contracting.
The company plans to develop these workplace agents through focused pilot programs with businesses. Each specialist Dot would be configured around a specific role, toolset, access level, and approval process. OpenAI is also working on integration with Microsoft Agent 365 to support enterprise governance and security controls.
Dots are rolling out in ChatGPT for Pro and Business Premium users in eligible markets. Enterprise users, including Education and Healthcare workspaces, can try the beta once an administrator enables it. One Dot is included with Pro and Business Premium subscriptions at no additional cost.
OpenAI says those plans also include a usage allowance for more demanding work, with higher limits during the first month after launch. In the future, users will be able to add more Dots and increase either the speed at which they operate or the amount of work they can complete each month.
With Dots, OpenAI is pushing toward a future where AI assistants are not just reactive tools but ongoing collaborators. If the system works as promised, users could soon delegate complex digital work to AI agents that can research, build, update, and prepare tasks in the background while keeping humans in control of final decisions.
