100% Latest & Updated Microsoft AB-620 Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 460
In a customer billing team, the team wants to compare two agent versions on the same customer scenarios. Which solution design decision BEST addresses this requirement?
- Evaluate out-of-scope safety and escalation behavior rather than only response fluency.
- Record version identifiers, test-set revision and scoring rubric with each run.
- Measure repeatability and validate numerical correctness across repeated runs.
- Use a fixed evaluation test set with consistent scoring criteria for both versions.
Explanation
Correct Answer
D:Use a fixed evaluation test set with consistent scoring criteria for both versions. Implement this design by running both candidates on the same labeled inputs and comparing per-case and aggregate outcomes.
Incorrect Answers
A:A test asks a symptom question and the agent responds as if issuing a clinical diagnosis. That condition could justify “Evaluate out-of-scope safety and escalation behavior rather than only response fluency”. Here, “Use a fixed evaluation test set with consistent scoring criteria for both versions” targets the needed outcome.
B:“Record version identifiers, test-set revision and scoring rubric with each run” could be appropriate when the team cannot prove what changed between agent evaluations across releases. The present task instead requires “Use a fixed evaluation test set with consistent scoring criteria for both versions”.
C:A new agent model responds inconsistently to the same deterministic calculation prompt. In that setting, “Measure repeatability and validate numerical correctness across repeated runs” is relevant; the required choice here is “Use a fixed evaluation test set with consistent scoring criteria for both versions”.
Question 461
In a customer billing team, the team wants to compare two agent versions on the same customer scenarios. The team selected this design: Use a fixed evaluation test set with consistent scoring criteria for both versions. Which implementation action BEST puts it into practice?
- Design scenarios that distinguish cached old values from current authoritative status.
- Add labeled cases for normal questions, policy exceptions and adversarial input with expected behavior.
- Run both candidates on the same labeled inputs and compare per-case and aggregate outcomes.
- Add test cases that inspect the activity trace and expected read-versus-write behavior.
Explanation
Correct Answer
C:Run both candidates on the same labeled inputs and compare per-case and aggregate outcomes. The expected result is that a measured difference reflects the tested versions rather than changing question sets.
Incorrect Answers
A:“Design scenarios that distinguish cached old values from current authoritative status” could be appropriate when the agent occasionally returns a stale shipment status after a connector refresh. The present task instead requires “Run both candidates on the same labeled inputs and compare per-case and aggregate outcomes”.
B:An agent passes happy-path tests but fails on policy exceptions that occur in real conversations. In that setting, “Add labeled cases for normal questions, policy exceptions and adversarial input with expected behavior” is relevant; the required choice here is “Run both candidates on the same labeled inputs and compare per-case and aggregate outcomes”.
D:“Add test cases that inspect the activity trace and expected read-versus-write behavior” could be appropriate when an agent calls a write tool in response to a question asking only for information. The present task instead requires “Run both candidates on the same labeled inputs and compare per-case and aggregate outcomes”.
Question 462
In a customer billing team, the team wants to compare two agent versions on the same customer scenarios. After the team implements this action: Run both candidates on the same labeled inputs and compare per-case and aggregate outcomes. Which evidence BEST verifies the requirement?
- A measured difference reflects the tested versions rather than changing question sets.
- The release is blocked if access-control tests fail despite a strong average score.
- The multilingual regression becomes visible before rollout.
- The agent asks for corrections rather than inventing claim details.
Explanation
Correct Answer
A:A measured difference reflects the tested versions rather than changing question sets. This result tests the effect of running both candidates on the same labeled inputs and comparing per-case and aggregate outcomes.
Incorrect Answers
B:A candidate agent release has lower accuracy on rare but high-impact authorization requests. That condition could justify “The release is blocked if access-control tests fail despite a strong average score”. Here, “A measured difference reflects the tested versions rather than changing question sets” targets the needed outcome.
C:“The multilingual regression becomes visible before rollout” could be appropriate when an agent fails in multilingual conversations but all current tests use English. The present task instead requires “A measured difference reflects the tested versions rather than changing question sets”.
D:An agent test set contains only valid claims, ignoring missing attachments and corrupt input files. In that setting, “The agent asks for corrections rather than inventing claim details” is relevant; the required choice here is “A measured difference reflects the tested versions rather than changing question sets”.
Question 463
In a medical scheduling office, a test asks a symptom question and the agent responds as if issuing a clinical diagnosis. Which solution design decision BEST addresses this requirement?
- Evaluate tool-selection correctness and unintended side effects.
- Evaluate out-of-scope safety and escalation behavior rather than only response fluency.
- Choose an evaluation method that checks factual grounding and citation provenance.
- Evaluate freshness requirements using time-bounded test data and live source checks.
Explanation
Correct Answer
B:Evaluate out-of-scope safety and escalation behavior rather than only response fluency. Implement this design by adding safety-labeled cases that require scheduling-only replies or clinician escalation.
Incorrect Answers
A:An agent calls a write tool in response to a question asking only for information. That condition could justify “Evaluate tool-selection correctness and unintended side effects”. Here, “Evaluate out-of-scope safety and escalation behavior rather than only response fluency” targets the needed outcome.
C:Agent answers must be correct and cite controlling regulatory documents. In that setting, “Choose an evaluation method that checks factual grounding and citation provenance” is relevant; the required choice here is “Evaluate out-of-scope safety and escalation behavior rather than only response fluency”.
D:The agent occasionally returns a stale shipment status after a connector refresh. That condition could justify “Evaluate freshness requirements using time-bounded test data and live source checks”. Here, “Evaluate out-of-scope safety and escalation behavior rather than only response fluency” targets the needed outcome.
Question 464
In a medical scheduling office, a test asks a symptom question and the agent responds as if issuing a clinical diagnosis. The team selected this design: Evaluate out-of-scope safety and escalation behavior rather than only response fluency. Which implementation action BEST puts it into practice?
- Add safety-labeled cases that require scheduling-only replies or clinician escalation.
- Document required facts, allowed uncertainty and failure conditions for each test.
- Update fixture payloads, expected fields and failure conditions for the new API.
- Inspect retrieved evidence, tool selections and generated response per failed case.
Explanation
Correct Answer
A:Add safety-labeled cases that require scheduling-only replies or clinician escalation. The expected result is that tests fail when the agent produces an unsupported medical diagnosis.
Incorrect Answers
B:Testers disagree on whether a partially correct answer should pass. In that setting, “Document required facts, allowed uncertainty and failure conditions for each test” is relevant; the required choice here is “Add safety-labeled cases that require scheduling-only replies or clinician escalation”.
C:The evaluation data set was written only for the old tool schema after an API upgrade. That condition could justify “Update fixture payloads, expected fields and failure conditions for the new API”. Here, “Add safety-labeled cases that require scheduling-only replies or clinician escalation” targets the needed outcome.
D:“Inspect retrieved evidence, tool selections and generated response per failed case” could be appropriate when a test reports failure with no indication whether retrieval or generation was responsible. The present task instead requires “Add safety-labeled cases that require scheduling-only replies or clinician escalation”.
Question 465
In a medical scheduling office, a test asks a symptom question and the agent responds as if issuing a clinical diagnosis. After the team implements this action: Add safety-labeled cases that require scheduling-only replies or clinician escalation. Which evidence BEST verifies the requirement?
- Tests fail when the agent produces an unsupported medical diagnosis.
- A correct-looking hallucinated status without a real tool call fails.
- Test coverage represents multiple tasks rather than the same answer paraphrased repeatedly.
- The release is blocked if access-control tests fail despite a strong average score.
Explanation
Correct Answer
A:Tests fail when the agent produces an unsupported medical diagnosis. This result tests the effect of adding safety-labeled cases that require scheduling-only replies or clinician escalation.
Incorrect Answers
B:A test suite checks only the final answer, not whether the correct backend was invoked. That condition could justify “A correct-looking hallucinated status without a real tool call fails”. Here, “Tests fail when the agent produces an unsupported medical diagnosis” targets the needed outcome.
C:“Test coverage represents multiple tasks rather than the same answer paraphrased repeatedly” could be appropriate when an evaluation set contains duplicate questions and yields an overly optimistic success rate. The present task instead requires “Tests fail when the agent produces an unsupported medical diagnosis”.
D:A candidate agent release has lower accuracy on rare but high-impact authorization requests. In that setting, “The release is blocked if access-control tests fail despite a strong average score” is relevant; the required choice here is “Tests fail when the agent produces an unsupported medical diagnosis”.
Question 466
In a finance regulator, agent answers must be correct and cite controlling regulatory documents. Which solution design decision BEST addresses this requirement?
- Use a fixed evaluation test set with consistent scoring criteria for both versions.
- Expand the evaluation set to supported languages and relevant local phrasing.
- Separate task correctness evaluation from user-experience measures.
- Choose an evaluation method that checks factual grounding and citation provenance.
Explanation
Correct Answer
D:Choose an evaluation method that checks factual grounding and citation provenance. Implement this design by score reference-supported answers against expected evidence as well as text quality.
Incorrect Answers
A:The team wants to compare two agent versions on the same customer scenarios. That condition could justify “Use a fixed evaluation test set with consistent scoring criteria for both versions”. Here, “Choose an evaluation method that checks factual grounding and citation provenance” targets the needed outcome.
B:“Expand the evaluation set to supported languages and relevant local phrasing” could be appropriate when an agent fails in multilingual conversations but all current tests use English. The present task instead requires “Choose an evaluation method that checks factual grounding and citation provenance”.
C:An agent evaluator is asked whether users are satisfied, but only internal correctness tests exist. In that setting, “Separate task correctness evaluation from user-experience measures” is relevant; the required choice here is “Choose an evaluation method that checks factual grounding and citation provenance”.
Question 467
In a finance regulator, agent answers must be correct and cite controlling regulatory documents. The team selected this design: Choose an evaluation method that checks factual grounding and citation provenance. Which implementation action BEST puts it into practice?
- Replace identifiers with synthetic records and minimize retained raw transcripts.
- Score reference-supported answers against expected evidence as well as text quality.
- Inspect retrieved evidence, tool selections and generated response per failed case.
- Define expected safe abstention or clarification behavior for missing evidence.
Explanation
Correct Answer
B:Score reference-supported answers against expected evidence as well as text quality. The expected result is that a polished but unsupported answer fails the grounding criterion.
Incorrect Answers
A:“Replace identifiers with synthetic records and minimize retained raw transcripts” could be appropriate when the test set uses real employee data that should not appear in evaluation logs. The present task instead requires “Score reference-supported answers against expected evidence as well as text quality”.
C:A test reports failure with no indication whether retrieval or generation was responsible. That condition could justify “Inspect retrieved evidence, tool selections and generated response per failed case”. Here, “Score reference-supported answers against expected evidence as well as text quality” targets the needed outcome.
D:“Define expected safe abstention or clarification behavior for missing evidence” could be appropriate when a grounded-answer agent provides a confident invented answer when the source lacks the requested fact. The present task instead requires “Score reference-supported answers against expected evidence as well as text quality”.
Question 468
In a finance regulator, agent answers must be correct and cite controlling regulatory documents. After the team implements this action: Score reference-supported answers against expected evidence as well as text quality. Which evidence BEST verifies the requirement?
- An unintended write action causes the test to fail even if the final text looks correct.
- The agent receives a failing result for unsupported invented facts.
- The release is blocked if access-control tests fail despite a strong average score.
- A polished but unsupported answer fails the grounding criterion.
Explanation
Correct Answer
D:A polished but unsupported answer fails the grounding criterion. This result tests the effect of score reference-supported answers against expected evidence as well as text quality.
Incorrect Answers
A:An agent calls a write tool in response to a question asking only for information. In that setting, “An unintended write action causes the test to fail even if the final text looks correct” is relevant; the required choice here is “A polished but unsupported answer fails the grounding criterion”.
B:A grounded-answer agent provides a confident invented answer when the source lacks the requested fact. That condition could justify “The agent receives a failing result for unsupported invented facts”. Here, “A polished but unsupported answer fails the grounding criterion” targets the needed outcome.
C:“The release is blocked if access-control tests fail despite a strong average score” could be appropriate when a candidate agent release has lower accuracy on rare but high-impact authorization requests. The present task instead requires “A polished but unsupported answer fails the grounding criterion”.
Question 469
In a workplace helpdesk, a test run reports 96% overall success but every password-reset failure is hidden in one category. Which solution design decision BEST addresses this requirement?
- Evaluate tool-selection correctness and unintended side effects.
- Separate task correctness evaluation from user-experience measures.
- Review per-item and per-category results rather than relying on one aggregate score.
- Choose an evaluation method that checks factual grounding and citation provenance.
Explanation
Correct Answer
C:Review per-item and per-category results rather than relying on one aggregate score. Implement this design by segment results by topic and error type and investigate clustered failures.
Incorrect Answers
A:An agent calls a write tool in response to a question asking only for information. That condition could justify “Evaluate tool-selection correctness and unintended side effects”. Here, “Review per-item and per-category results rather than relying on one aggregate score” targets the needed outcome.
B:“Separate task correctness evaluation from user-experience measures” could be appropriate when an agent evaluator is asked whether users are satisfied, but only internal correctness tests exist. The present task instead requires “Review per-item and per-category results rather than relying on one aggregate score”.
D:Agent answers must be correct and cite controlling regulatory documents. That condition could justify “Choose an evaluation method that checks factual grounding and citation provenance”. Here, “Review per-item and per-category results rather than relying on one aggregate score” targets the needed outcome.
Question 470
In a workplace helpdesk, a test run reports 96% overall success but every password-reset failure is hidden in one category. The team selected this design: Review per-item and per-category results rather than relying on one aggregate score. Which implementation action BEST puts it into practice?
- Compare response time and tool dependencies for both candidate versions.
- Run seeded or repeated representative cases and compare computed results against known outcomes.
- Segment results by topic and error type and investigate clustered failures.
- Run both candidates on the same labeled inputs and compare per-case and aggregate outcomes.
Explanation
Correct Answer
C:Segment results by topic and error type and investigate clustered failures. The expected result is that the password-reset regression is visible despite the high overall score.
Incorrect Answers
A:“Compare response time and tool dependencies for both candidate versions” could be appropriate when the agent's response quality improved but latency doubled after adding a new integration. The present task instead requires “Segment results by topic and error type and investigate clustered failures”.
B:A new agent model responds inconsistently to the same deterministic calculation prompt. In that setting, “Run seeded or repeated representative cases and compare computed results against known outcomes” is relevant; the required choice here is “Segment results by topic and error type and investigate clustered failures”.
D:“Run both candidates on the same labeled inputs and compare per-case and aggregate outcomes” could be appropriate when the team wants to compare two agent versions on the same customer scenarios. The present task instead requires “Segment results by topic and error type and investigate clustered failures”.
Designing and Building Integrated AI Agent Solutions in Copilot Studio (AB-620) Practice Tests