Analyze, Monitor, and Tune AI-powered Business Solutions AB-100 Practice Questions (Page 7)

Questions 31–35

1. 31application · medium

   An e-commerce company's AI chatbot suddenly starts giving incorrect product recommendations. The team has access to Azure Application Insights telemetry and the chatbot's conversation logs. They suspect a recent change to the product catalog is causing the issue. What should they do to identify the root cause?

   • AIncrease the chatbot's confidence threshold to reduce incorrect recommendations.
   • BAnalyze telemetry for error rates and latency, and compare conversation logs before and after the catalog change.
   • CRoll back the product catalog change immediately and assume that fixes the issue.
   • DAsk the development team to review the code changes from the last deployment.
   Reveal Solution Discussion Select an answer first
2. 32expert · hard

   A company is deploying a new AI agent for internal HR queries. The compliance team requires that all monitoring data be retained for at least 3 years and that alerts be sent to a specific security team. The operations team wants to design a monitoring process that meets these requirements. What should they do?

   • AUse Azure Log Analytics with default retention and set up email alerts to the security team.
   • BUse Azure Service Health to monitor the agent and send alerts to the security team.
   • CConfigure Application Insights with a 3-year retention policy and route alerts to the security team's action group.
   • DUse Azure Dashboards to display monitoring data and manually export logs for retention.
   Reveal Solution Discussion Select an answer first
4. 33application · medium

   An AI agent that handles employee benefits inquiries has a high escalation rate. The backlog shows that many escalated tickets contain the phrase 'what about my dependents'. User feedback also mentions that the agent does not understand dependent-related questions. What should the team do?

   • ARetrain the model on all historical tickets to improve overall understanding.
   • BIncrease the confidence threshold to avoid incorrect responses.
   • CAdd a new intent for dependent-related queries and update the knowledge base with dependent coverage details.
   • DRemove dependent-related queries from the agent's scope.
   Reveal Solution Discussion Select an answer first
5. 34application · medium

   A company has deployed an AI agent for customer onboarding. The agent's performance is measured by task completion rate, average handling time, and user satisfaction score. The team wants to monitor these metrics in real time. Which tool should they use?

   • AAzure Monitor with Application Insights and custom metrics.
   • BAzure Service Health for real-time alerts on agent performance.
   • CAzure Dashboards with static snapshots of performance data.
   • DAzure Log Analytics with scheduled queries that run daily.
   Reveal Solution Discussion Select an answer first
6. 35application · medium

   A travel agency's AI booking assistant has a high rate of 'false declines' — it tells users a flight is unavailable when it is actually available. Telemetry shows the agent's confidence threshold is set very high (0.95). User feedback indicates frustration with the assistant. What tuning technique should they apply?

   • ADisable the assistant and use a rule-based booking system.
   • BLower the confidence threshold to 0.85 to reduce false declines.
   • CRetrain the model with more flight availability data.
   • DIncrease the confidence threshold to 0.99 to be more certain.
   Reveal Solution Discussion Select an answer first