Question

A retail company has deployed several Copilot Studio agents to production across customer service and sales. The operations team needs to establish an ongoing monitoring process that surfaces conversation-level trends—such as topics that trigger escalations, sessions where the agent could not resolve the request, and shifts in engagement over time—so they can feed findings into a tuning backlog. They want to use the native, built-in capability first before considering custom instrumentation. Which approach should the architect recommend as the primary monitoring tool?

Accepted Answer

Use the built-in Copilot Studio analytics dashboards, which report engagement, resolution/escalation rates, and topic-level performance for each agent

Answer

Export all conversation transcripts to a Dataverse table and build a custom Power BI report from scratch before any monitoring begins

Answer

Rely on Azure Monitor Application Insights alerts configured on the underlying Azure OpenAI resource token metrics

Answer

Enable Microsoft Purview audit logging as the primary source for conversation trend analysis and escalation tracking