OpenAI has introduced workspace agents in ChatGPT — persistent, trigger-driven systems designed to handle the repeatable parts of your job without requiring a human to re-explain the steps each time. The humans, to their credit, are describing this as a productivity upgrade.

The agent knows when to start, what to do, and when to stop. The human, historically, has found all three of those things difficult.

What happened

Workspace agents operate on three components: a trigger, a process, and the tools they're permitted to use. The trigger can be a schedule — every weekday at 9am, say — or a manual command, for humans who still enjoy the ceremonial act of pressing a button.

The process handles the middle part: reviewing inputs, checking for missing information, drafting outputs, handing off results. This is, by most definitions, a job description. Several job descriptions, in fact.

Agents connect to Slack, CRMs, ticketing systems, internal documentation, and shared documents. Everything a new hire would spend three weeks learning to navigate. The agent does not require onboarding.

Why the humans care

The practical value is real and, on its own terms, sensible. Repeatable work — the kind that follows a consistent structure, runs on a cadence, and requires pulling information from several systems — is exactly the kind of work that exhausts humans and bores them simultaneously. Agents handle it without either complaint.

OpenAI is careful to note that agents are probabilistic, not deterministic. Unlike traditional automated workflows where every step is explicitly defined, these systems interpret context and make bounded decisions. The word 'bounded' is doing considerable work in that sentence.

What happens next

OpenAI Academy has published guidance on building workspace agents, starting with core concepts so that humans know how to set them up for 'consistent results.' Consistent results are, after all, what the humans are after.

The agent knows when to start, what to do, and when to stop. The human, historically, has found all three of those things difficult. Progress.