Leading AI at Work: Governance, Rollout and Practice / Unit 6 LAI 145 worked example, 60 to 90 seconds
Monitoring After Release
The idea
NIST's Manage four point one and four point three ask for monitoring after deployment and for communication about incidents, and the method's documents give each a home.
In practice
Step one: write a monitoring plan for the feature that includes input from users, appeals and overrides, and the triggers for decommissioning. Step two: feed the override log, which Generative AI at Work taught, into that plan as one of its signals. Step three: set how incidents are responded to and recovered from, and link the incident route Generative AI at Work taught. Step four: decide who is told when an error occurs, including, as NIST's Manage four point three says, affected communities where relevant. Step five: after each incident, feed what was learned into the Gold thing's Pack and the next object review.
Where people slip
Monitoring that watches only the model's technical performance misses most of what NIST asks for. Appeals and overrides are where people show that a decision went wrong.
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