When did you last say no to a change request?
The scenario:
Your change advisory board approved 18 changes last month. They rejected zero.
If you’re approving 100% of what comes through, you’re not running a change board. You’re running a rubber stamp. The point of CAB is to catch the changes that shouldn’t happen, push back on the ones that aren’t ready, and time the ones that need a better window. If nothing ever gets rejected, the board isn’t doing its job.
The prompt:
You’re redesigning your change advisory board to catch risky changes before production.
Works in any AI tool.
Context: [your last 20 change requests with status and outcome, your current CAB process]
Build:
Rejection criteria (what should result in a clear no? missing rollback, no testing, wrong window)
Deferral criteria (what should push to a later window?)
A named challenger role whose job is to push back on every change
A one-page scoring rubric (low/medium/high risk) with required evidence per tier
A target rejection rate (healthy CABs reject 10-15% outright and defer another 20%)
Run your last 20 changes through the rubric. If nothing would have been rejected, it’s still too soft.