The update
The task can continue after the first answer
OpenAI launched ChatGPT Work on 9 July for longer tasks with several steps. It can research and analyse information, work with connected apps and files, and create documents, spreadsheets, presentations, reports and Sites.
People can follow the task, answer questions, change direction and approve important actions while it runs. Scheduled Tasks can also repeat work or monitor for changes, which moves the product beyond a single conversation.
What changes
A brief now has to survive the whole journey
A short prompt can be enough for a short answer. A task that lasts an hour and touches several files needs more structure. The agent needs to know the outcome, the approved sources, the constraints and the moment when it should stop for review.
For businesses and creatives, this makes the brief more operational. It needs to carry the direction through research, production and delivery, rather than relying on somebody to correct the course in every message.
- Describe the finished deliverable before listing the steps.
- Name the files and sources the agent is allowed to use.
- Add approval points where the work becomes public or expensive to reverse.
The practical check
A finished file still needs somebody who understands the work
The ability to produce a polished file can create false confidence. A deck may be editable and a report may be neatly structured while the reasoning underneath is still weak.
Use Work to give the agent a stronger route, while keeping a person close to the decisions that shape the result. A finished file is safe to use only after someone who understands the work reviews the reasoning beneath it.
Checked sources