OpenAI brings workspace administration into ChatGPT Work
Reviewed by Helen Jones
OpenAI has introduced an Admin plugin for ChatGPT Work and Codex that lets authorised administrators inspect workspace activity, manage access and usage, and take supported administrative actions in chat. OpenAI positions it for routine onboarding, offboarding and team changes while retaining existing permissions and controls.
This is a practical example of a useful AI workflow changing from 'answer a question' to 'complete a bounded administrative job'. The value comes from bringing the relevant workspace information and approved action into the same conversation, not from letting the model act without limits. For managers building their own AI workflows, the useful pattern is the boundary: define which requests are safe to complete, which records the assistant may use, and which exceptions must still go to a human owner. That is more reusable than a clever one-off prompt.
Analysis
Take one repetitive team-admin request and write a three-column handoff: information the assistant may read, actions it may complete, and exceptions that require a named human. Use that as the review standard for any proposed AI workflow.
Source note
Pulse published by Collab365 Spaces, reviewed by Helen Jones on . Cite as "OpenAI brings routine workspace administration into ChatGPT Work", Collab365 Spaces.