Generative AI at Work / Unit 9 GAI 211 worked example, 60 to 90 seconds
Lending and Fairness
The idea
The method's financial-services playbook treats credit decisions as the highest level of risk, and builds the documentation that lets an institution explain them, a need sharpened by NIST's warnings on bias.
In practice
Step one: map every input the model uses to a defined field of the decision record. Step two: make the model's output feed that defined decision record, with fields for the rationale and the policy reference. Step three: document how any change to the model changes the decision rules. Step four: be ready to answer the regulator's questions: how the decision is made, which factors influence approval or denial, and how fairness is ensured. Step five: test outcomes across groups, remembering NIST's point that mitigating bias does not by itself make a system fair.
Where people slip
If the institution cannot answer those questions from its own records, the model should not be making or shaping the decision.
Your place on the trail
Every lesson rests on earlier ones and opens later ones. Tap any to jump.
Unlocks
- Nothing yet depends on this.
Something wrong in this lesson? Flag it · Sources for Generative AI at Work