It is easy to understand an assistant that handles a single request. An agent that keeps following a project after the conversation ends raises a harder question: who decides when it can move from watching to acting?
OpenAI introduced dots on September 29, describing persistent agents that can work across projects, use connected apps, and return with results or decisions for review. Each dot has its own cloud computer, runs on GPT-6 Astra, and can be reached through ChatGPT, Slack, or Teams. OpenAI says feedback helps it learn a user's preferences over time. That is the product's stated design, not independent evidence that it will maintain the right context throughout long-running work.
In OpenAI's development example, a dot follows customer feedback, finds recurring requests, implements and tests small fixes, and prepares pull requests for a person to review. This changes the workflow: a task need not fit in one prompt or rely on somebody remembering to restart each step. It also increases the consequences of a misunderstanding. A mistaken interpretation can cross applications and become an actual change.
Where research ends and action begins
OpenAI's technical explanation of its safeguards says a feature called proactive research can inspect connected sources in the background. At that stage, tools are read-only: they cannot send messages, change app content, or control a browser or computer. Any later action must pass through the usual authorization and review rules.
That distinction matters more than the promise of a 24/7 agent. Users choose the connected apps and can set rules that allow, block, or require approval for particular actions. A separate check evaluates steps that might affect accounts or share information. OpenAI itself acknowledges that dots can make mistakes and advises reviewing consequential work.
Persistent agents become useful at work when teams can see their access, follow their progress, and identify exactly where a human decision is required.
Our view at MnzAI Labs is to start with a narrow, observable workflow: defined information sources, a verifiable output, and explicit approval before changing systems or contacting other people. That provides a way to measure saved effort without treating continuous work as unlimited autonomy. Background research is not blanket permission to take action.
Who can use dots now
The rollout has started for Pro and Business Premium subscribers in eligible markets. Enterprise, Edu, and Healthcare users can try the beta if their workspace administrator enables it. The first dot is included with Pro and Business Premium, with an allowance for deeper work; tasks it launches in Codex or ChatGPT Work still count toward those products' usage limits. OpenAI has not supplied an independent measure of productivity or reliability for continuous use in its announcement.
OpenAI is also piloting specialist dots for enterprises, with their own identities and credentials for responsibilities defined by an organization. That is a separate stage from the personal dot rollout. For technical leaders, the immediate decision is less about how many agents to deploy than which permissions, activity records, and review points make their output auditable.
Sources: OpenAI's dots announcement and its safety and privacy explanation.