Agentic AI broke your cost model. A single autonomous agent can burn more tokens in an afternoon than an entire department did last quarter — and by the time it shows up on the invoice, the money is gone. Traditional API governance was never designed for consumption that's non-deterministic, streaming, and priced per token.
This session shows how to turn Azure API Management into a real cost-control plane for AI workloads: enforcing token budgets, attributing spend to actual consumers, and building the telemetry that makes chargeback possible. Everything demonstrated comes from production deployments, including the parts that broke.