LimitWise at Corvana Bank
Governance analyst: Dr. Kyle David · AI Governance Practicum, Track 1 · the lifecycle of one AI system, Week 1 to Week 9
LimitWise is Corvana Bank's model for early warning of payment default in the next statement cycle. Over nine weeks I took it from the first question (should this system be built?) to the deployment decision and its monitoring plan. These are the six documents that record it. The organization, the system, and the people are fictional; the data is public.
The dossier
- Use Case Assessment (Week 1)
- EDA Observation Log (Week 2)
- Dataset Datasheet (Week 3)
- Model Selection Rationale (Week 4)
- Model Card (Week 6)
- Go/No-Go Decision Record (Week 7)
The repositories
- Model repository: the model card and the fixed weights of LimitWise v2, the version the Chief AI Officer signed.
- Dataset repository: the datasheet and the remediation log for the working file.
- The running model: the program's Space, where one account is scored, a file is ranked, and a monthly file is read against the training file and the tripwires.
- The collection: all of the above in one place.
A score says where a cardholder's record ranks on one list. It is not a decision about the cardholder. The decisions, and the conditions on them, are in the Go/No-Go Decision Record above.