Catalog, lineage, and access patterns that satisfy risk while enabling self-service analytics and controlled AI experiments.
Risk needed lineage and policy evidence; analysts needed speed. We rolled out a catalog-first program with stewardship, classification, and access patterns that scaled to self-service and controlled AI pilots.
How we approached it
Stewardship model
Named owners for critical datasets with RACI clarity between business, data office, and platform engineering.
Policy & access
Classification aligned to regulatory expectations; ABAC/RBAC patterns that auditors can follow in documentation.
Lineage & AI readiness
Coverage for reporting chains and model inputs; hooks for approving training data and monitoring production use.
What we delivered
- Stewardship model with domain ownership on critical datasets
- Policy and classification aligned to regulatory expectations
- Lineage coverage for key reporting and model inputs
- Continuous improvement cadence with data office and engineering
Outcomes
-
Faster onboarding for analysts with trusted dataset entries
-
Evidence pack for reviews without blocking every ticket
-
Controlled path for GenAI experiments on approved corpora
Representative stack: