Episode 13: Model governance at scale: Evaluating AI without breaking production

This AI Horizons episode focuses on model governance as an engineering discipline, not a policy exercise. 

Using Microsoft Foundry Evaluations, teams can treat model transitions as a repeatable, testable workflow:

• Compare candidate models against live workload baselines

• Detect regressions in quality, safety, latency, and cost before rollout

• Produce evidence for go / nogo decisions that stand up to risk and compliance review