AI agents are rapidly becoming part of modern applications, but they introduce a new challenge: how do you observe systems that can reason, make decisions, and invoke tools autonomously?

In this session, we'll explore how Application Insights and Azure Monitor bring observability to AI-powered applications. We'll look at tracing agent executions, monitoring tool calls, tracking model performance, understanding token consumption, and diagnosing failures across agent workflows.

We close the loop by showing how Azure's Observability Agent is transforming operations: what to instrument, what to alert on, and where the humans still need to stay in the loop.