Where drift gets in
When AI misreads what someone meant, the error rarely shows up right away. It slips in at a few predictable moments and grows from there. Ask these questions at each one, whether the AI is asking itself or your team is reviewing its work.
When a request arrives
What did they actually mean? Say it back before doing anything.
Before acting
If I’ve misread this, what would I be doing wrong right now?
Reaching for context
What was already decided that I don’t know about yet?
Turning intent into a plan
Does every step trace back to what was asked?
Deciding how careful to be
Who could be hurt if this is wrong, and how badly?
Calling it done
What proof would convince someone who wasn’t watching?
Handing off
What does the next person or session need so it doesn’t start from a guess?
Free to use and share. Credit appreciated.
Jonathan Haber · founder@ixcoach.com