Copilot Studio Analytics for Inaccurate Answers

Analyze, monitor, and tune AI-powered business solutions

Answer Correct answer: C, E — Use topic usage with low resolution and quality of generated answers to identify and fix inaccurate agent responses.

A company has a Microsoft Copilot Studio agent that provides answers based on a knowledge base for customer support. Users report that, occasionally, the agent provides inaccurate answers. You need to use metrics from the Analytics tab in Copilot Studio to identify the cause of the inaccuracies. Which two options should you use? Each correct answer presents part of the solution. NOTE: Each correct selection is worth one point.

survey results
session information and session outcomes
topic usage and topics with low resolution Correct Answer
engagement, resolution, and escalation rates
quality of generated answers Correct Answer

Community Insight

The exam tests the ability to distinguish between high-level engagement metrics and diagnostic quality metrics, with the common trap being the selection of general session data instead of specific answer quality indicators.

This question addresses how to diagnose inaccurate responses in a Microsoft Copilot Studio agent using specific analytics metrics. The solution involves leveraging topic usage data and quality assessment tools to pinpoint knowledge gaps and response failures.

Learners often select 'Session information' (B) because it provides conversation context, but it is less direct than 'Topic usage' (C) for identifying structural knowledge issues or 'Quality of generated answers' (E) for evaluating the AI's output accuracy.

Community Discussion (9 comments)

Joselito79 👍 1 Selected: CE

✅ Correct answers: C and E Option Why C. Topic usage and topics with low resolution Helps identify which topics are associated with poor outcomes, revealing areas where the agent may have knowledge/coverage problems. E. Quality of generated answers Directly measures whether generative AI answers are Good or Poor, allowing you to investigate inaccurate responses and the knowledge sources involved.

Ninashnash 👍 3 Selected: BE

keyword in question is "identify the cause of the inaccuracies." Only B and E can help to identify the root cause: B (session information and session outcomes) lets you inspect the actual query, the knowledge source used, and the conversation flow — which can reveal the cause (e.g., wrong knowledge source retrieved, user question misunderstood, missing content). E (quality of generated answers) directly gives a cause — Copilot Studio assigns a specific reason category to each poor-quality answer (e.g., irrelevant, incomplete, ungrounded), which is a textbook "cause."

Juan_Covili 👍 1 Selected: BE

Quality of generated answers es la métrica más directa para investigar respuestas inexactas generadas desde una knowledge base. Copilot Studio analiza la calidad de respuestas generativas considerando factores como completeness, relevance y groundedness, y marca respuestas como buenas o pobres. Además, para respuestas pobres muestra razones de baja calidad.

omath 👍 4 Selected: BE

B. Session information and session outcomes This metric shows: - How conversations unfolded - Where the agent misunderstood the user - Whether the session ended in success, escalation, or failure - Patterns of incorrect grounding or mis‑routed topics This is essential for diagnosing why the agent produced an incorrect answer. E. quality of generated answers This metric evaluates: - Relevance - Accuracy - Grounding quality - Whether the answer aligns with the knowledge base If the agent is hallucinating or misusing the knowledge base, this metric exposes it.

madimad 👍 3 Selected: CE

correct answer C & E

Newb007 👍 4 Selected: CE

IT's C and E !!!

Newb007 👍 2 Selected: BC

The two analytics areas in Copilot Studio that directly help diagnose why a knowledge‑base–driven agent is giving inaccurate answers are: B. Session information and session outcomes C. Topic usage and topics with low resolution

Newb007 👍 5

I was wrong it's this... The correct analytics areas to diagnose why a Copilot Studio agent sometimes gives inaccurate answers from a knowledge base are: ✔ C. topic usage and topics with low resolution ✔ E. quality of generated answers These two directly map to the root‑cause signals described in the Microsoft Learn page you have open.

rveney 👍 1 Selected: CE

C. topic usage and topics with low resolution E. quality of generated answers

Comments & Corrections

No comments yet — spotted an error or have a note? Share it below.

Expert Analysis

Why the Answer Is Correct

The correct options are C and E. Topic usage and topics with low resolution (C) helps identify which specific intents or topics are failing, indicating where the knowledge base lacks coverage or the agent is misrouting queries. Quality of generated answers (E) directly measures the relevance and groundedness of the AI's responses, allowing you to filter for poor-quality answers and investigate the underlying sources.

Why the Other Options Are Wrong

Survey results (A) provide user sentiment but not technical root causes. Session information and session outcomes (B) show flow and success/failure states but do not directly evaluate the content accuracy of the generated text. Engagement, resolution, and escalation rates (D) are high-level performance KPIs that indicate volume and efficiency rather than specific answer inaccuracies.

Community Comment Notes

Multiple users, including Newb007 and Joselito79, correctly identified C and E as the answer. One commenter noted that CE maps to "root-cause signals." Another user initially voted for BE but later corrected to CE, emphasizing that while session info is useful, C and E are the primary diagnostic tools for this specific issue.

Official Reference

https://learn.microsoft.com/en-us/copilot-studio/analytics-overview

Exam Strategy

Focus on the specific goal: 'identify the cause of inaccuracies.' Prioritize metrics that evaluate content quality (E) and topic-specific failure rates (C) over general session tracking or user feedback.

Frequently Asked Questions

Why is survey results (A) not the best choice?

Surveys measure user satisfaction, not technical accuracy. They don't help you debug the agent's logic or knowledge base.

Can I use session information (B) to find bad answers?

While B shows failed sessions, it doesn't analyze the text content. E directly flags poor quality answers based on grounding and relevance.

More AB-100 FAQ →

Related Analysis

← Back to AB-100 Study Guide