Leading AI at Work: Governance, Rollout and Practice / Unit 5 LAI 119 worked example, 60 to 90 seconds
Acting on Overrides
The idea
The book's last metric is how often people overrule the AI, and the steps act on a log Generative AI at Work, lesson one hundred forty-four, teaches.
In practice
Step one: record each AI output as accepted, edited or rejected. Step two: give each edit or rejection a short reason: wrong fact, wrong rule, wrong tone, or missing information. Step three: review the reasons for patterns, not one case at a time. Step four: fix repeated wrong tone in the prompts. Step five: fix repeated wrong facts in the grounding, by improving what the tool is given. Step six: fix wrong rules that reached a reviewer through reviewer training, because the review missed them.
Where people slip
NIST's framework suggests that data on how often and why people overrule AI output may be worth collecting. The book adds that a team that never overrides its AI deserves a closer look, not a celebration: reviewers may not really be reviewing.
Your place on the trail
Every lesson rests on earlier ones and opens later ones. Tap any to jump.
Something wrong in this lesson? Flag it · Sources for Leading AI at Work: Governance, Rollout and Practice