Notes / Workers
An AI Worker Should Know When Not to Act
Useful AI workers know when to handle work, draft for review, ask for context, or escalate.
By Rich Hill III. Published Jul 16, 2026. 8 min read.
AI agents are getting better at moving work across tools.
They can read an intake form, summarize a thread, classify a request, draft a reply, update a record, route an issue, and prepare the next step in a workflow.
That capability matters. But the business question is not only what the system can do.
It is also this:
Real work is rarely clean from beginning to end. A lead arrives without a service area. A customer request looks routine until it includes a cancellation. A CRM record disagrees with billing. A teammate asks for something outside the process the worker was assigned.
In those moments, a useful AI worker should not fill in the blanks, silently choose a version of the truth, or stretch beyond its authority.
It should pause.
That is not a failure of automation. It is part of doing the recurring job responsibly.
Key takeaways
- Workers should pause when context is missing or records conflict.
- Approval belongs at deliberate decision boundaries.
- Handle, draft, ask, escalate is a practical operating model.
Frequently asked questions
When should an AI agent ask for approval?
Before actions that cross a boundary the business has defined, such as external commitments, exceptions, important record changes, access, or money-related decisions.
Is escalation failure?
No. A good escalation preserves context and gives the right person a clear decision.
Explore Workers