Module Assessment

1.

On the Monitor tab, an agent's total estimated credits used rose sharply this period, while total sessions stayed about the same. What should you check before you treat the increase as a production signal?

The credit breakdown by experience, to see whether production usage caused the increase

The Reactions card, to see whether negative reactions rose in the same period

The Tool use view, to see whether the agent invoked any new tools

2.

Users rate answers from an order status tool poorly. The session completes, and the activity trace for the reproduced request shows no Error node. The Tool node shows that the agent passed a customer name where the tool expects an order number. Where should you make the fix?

In the tool's connection, by updating its authentication credentials

In the agent's moderation level setting

In the tool's inputs, so the agent passes the value the tool expects

3.

Several failed sessions share the same tool and the same cause. You plan one change to the tool's description and want to confirm that the change works without breaking anything else. Which approach fits?

Publish the change, and then compare the success rate on the Monitor tab with the previous period

Add the pattern's requests and a request that already works to an evaluation, run it before and after the change, and then confirm in production data

Make the description change along with updates to the agent's instructions and knowledge sources, and then rerun the evaluation

4.

Your team wants an alert when calls to one tool start failing across agents in an environment. The agents use the GitHub Copilot harness, and your administrator has set up environment-level telemetry. Which approach fits?

Write a KQL query over the ExecuteTool spans in the dependencies table that counts failed calls by tool name, and base a log search alert rule on it

Configure agent-level telemetry on each agent and query the requests table

Filter the sessions list on the Monitor tab by failed status each day

5.

An agent that answers employee policy questions has high session counts and a steady success rate. Leadership asks whether the agent is worth its cost. What should your value story add?

A larger set of usage metrics, such as total users and autonomous runs

An estimate of time saved for each session, multiplied by total sessions, presented on its own as the agent's value

A business outcome, such as time per task, measured against a pre-launch baseline from your organization's own data, alongside the credits consumed

Feedback