{How to} Get smart actionable insights to improve your copilot agent performance

Today I am going to share my thoughts on getting the actionable insights to improve copilot agent performance.

Let's get started.

1. Enhanced Analytics Dashboard: The Analytics page now offers comprehensive insights into agent performance, including key metrics such as answer rate, usage, and error rates. This allows for a deeper understanding of how agents are performing and where improvements can be made.


2. Autonomous Agent Performance Metrics: For agents utilising autonomous capabilities, new analytics features provide visibility into how knowledge sources are used during each run. This helps in assessing relevance, adjusting content, and improving performance over time.


3. Unanswered Questions Analysis: Copilot Studio now categorises and highlights unanswered user questions, grouping them by themes and conversation contexts. This feature aids in identifying content gaps and refining agent responses.

4. Action Usage Insights: Enhanced visibility into how users utilise actions within an agent is now available. Metrics such as answer rate, action usage rate, and error rate provide a comprehensive view of user interactions, helping to identify areas for improvements.


5. Customer Satisfaction Feedback: The Customer Satisfaction tab offers detailed insights into user feedback, including average CAST Scores and satisfaction levels based on user queries. This data helps in understanding user sentiment and areas needing attention.


6. Evaluation Test Workflows: Copilot Studio Introduces structured, Automated testing workflows that simulate real user interactions; Makers can evaluate agents before going live, identifying weak spots and ensuring reliability.

7. Billing and Consumption Analysis: New billing analysis features allow makers to assess agent efficiency and cost breakdowns by event types. This helps in optimising operations and understanding the financial impact of agents.

That's it for today.

I hope this helps.

Malla Reddy Gurram (aka @UK365GUY)