Context
Business units were adopting AI tools independently with no shared review process, creating inconsistent risk exposure that compliance had flagged as a growing liability across 15+ units.
My role
Defined the intake process and review criteria, then led the rollout and adoption tracking across the portfolio.
Approach
Rather than write a single top-down policy, I interviewed teams already running AI in production to learn what a review actually needed to catch versus what would just slow people down. That became a tiered intake - light-touch for low-risk use, full review for anything touching PHI or automated decisions.
Outcome
Cut compliance escalations 45%, with no reported slowdown in delivery.
What I'd do differently
Invest earlier in training the intake reviewers themselves - the framework was only as fast as the humans running it, and that gap cost us an extra quarter.