Summary

Completed

In this module, you tested, published, and managed a support policy agent. You can now:

  • Choose the Preview tab for quick checks and the Evaluate tab for tests that you need to run again.
  • Create test conversations for common requests, unclear requests, out-of-scope questions, and escalations.
  • Test whether a main agent delegates in-scope requests to a connected agent and retains other requests.
  • Run an evaluation and review its General quality results.
  • Use the Monitor tab to find quality, tool, permission, and credit-capacity problems.
  • Publish an agent to Teams and Microsoft Copilot and test it with typical user permissions.
  • Explain how solutions, environments, pipelines, data policies, and access controls support ALM.
  • Identify activities that use Copilot Credits and where administrators track usage.

In the exercise, you created the Contoso Policy Agent, a support policy agent, and connected a specialized escalation agent. You tested the main agent with four conversations, including a delegation request. You fixed a conflicting instruction, ran the tests again, and tested the published agent in Microsoft Teams.

What you learned

  • Plan and run agent evaluations: Use the Preview tab for quick checks. Use the Evaluate tab to run the same tests before release and after important changes. Evaluations use the signed-in user's profile. General quality uses AI to assess responses and reports Pass or Fail per test case; it doesn't compare responses with optional expected answers. Manually verify boundaries and connected-agent routing.
  • Monitor and improve agents: You can open Monitor before publication; production usage and feedback appear after publication and user activity. Review sessions, reactions, tool use, and credit consumption, and identify the type of problem before making a change.
  • Publish and validate channels: Configure intended channels in the Publish agent dialog, publish, and follow the sharing or catalog approval path available in your organization. Always validate as a published user with typical permissions, not only as the agent's author or an administrator.
  • Govern agent lifecycle and costs: Solutions, separate environments, and pipelines help teams control releases. Building, testing, evaluating, and running agents can consume Copilot Credits. Administrators monitor consumption and configure limits in Power Platform admin center, and search activity audit logs in Microsoft Purview.

Learn more