OpenAI dots and the move from chat to ongoing work

Always-on agents change the design problem from answering a request to maintaining context, permissions and trust across days of work.

From a chat response to an ongoing outcome

OpenAI describes dots as always-on agents that can keep working across connected applications, maintain context and bring completed work back for review. That is a meaningful change in the unit of interaction: the user delegates an outcome instead of composing every next prompt.

Persistent work needs a visible control surface

Persistent work creates a new product surface. People need to understand what the agent is doing, what it has learned, which systems it can reach and when it will stop for a decision. Activity views, scoped permissions and approval rules become as important as model quality because the work continues after the conversation closes.

Begin with narrow permissions

For organisations, an always-on agent should not begin with universal access. It should begin with one responsibility and a deliberately small permission envelope. Read-only research can run proactively; state-changing actions should be restricted by clear rules and named approval owners. Context should be retained only where it improves the job and where governance permits it.

Keep progress continuous and control visible

The opportunity is substantial: release coordination, account preparation, document review and research monitoring are all work that unfolds over time. The design challenge is to make progress continuous without making control invisible.

Inspection is part of delegation

Persistent agents will earn adoption when teams can inspect their work as easily as they can assign it.