Deploying AI in a regulated environment requires more than model validation. It requires a governance framework that covers model versioning, performance drift thresholds, escalation protocols, and audit trail requirements. This post describes the framework we use internally, the decisions that shaped it, and the open questions we are still working through.