The regulatory constraint
Every ranked workflow was scored against model risk management expectations and fair-lending explainability before it entered the roadmap.
How we work in this industrySituation
Eleven AI pilots were running across five business units, none tied to a P&L owner. Leadership could not answer which one to fund next, and two prior requests for a central AI budget had been sent back by the board for lack of a comparable business case.
Roughly $2.1M in annualized tooling and pilot spend was already committed, with no mechanism to retire the pilots that were not working.
Approach
A four-week readiness assessment covering workflow volume and variance, data and content trustworthiness, decision ownership, and regulatory constraints across all eleven pilots.
Every candidate workflow scored on the same rubric, then ranked by the return achievable inside ninety days rather than by theoretical ceiling.
A governance structure defined alongside the roadmap: named owners per system, quarterly review cadence, and a documented escalation path for model decisions in credit and servicing.
What we built
- A ranked opportunity register covering 27 workflows with sizing, dependencies and named owners.
- A board-facing investment case for commercial loan document intake, the first funded initiative.
- An AI governance charter adopted by the risk committee, with named accountability per high-impact system.
“We had been arguing about which pilot was most exciting for a year. Four weeks later the board approved the first one on the numbers, and nobody in the room questioned the ranking.”
SVP, Enterprise Transformation — regional commercial bank