Summary

  • 2 minutes

You now have a repeatable way to monitor an agent after you publish it, improve it from production evidence, and explain what it delivers. The Monitor tab shows where the agent needs attention. Session transcripts and the activity trace show why. Evaluations show whether a change improves response quality before you publish it, and production data confirms the change afterward. Environment-level telemetry adds detail you can query, and a value story connects the agent's signals to outcomes the business tracks.

What you learned

In this module, you learned how to:

  • Use overview metrics, reactions, credit usage, tool use, and sessions on the Monitor tab to decide which production signal to follow first.
  • Trace a signal through session transcripts and the activity trace to separate quality defects from runtime failures and choose a targeted fix.
  • Group recurring session findings into patterns, add them to an evaluation (preview), and confirm each change in evaluation results and production data.
  • Request, validate, and query environment-level telemetry (preview) in Application Insights for your own measures, alerts, and dashboards.
  • Build a value story that links adoption, operational KPIs, and cost to business outcomes measured against a baseline.

Learn more