100% Latest & Updated Microsoft AB-100 Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 290
A new model version passes one-off tests but a small fraction of outputs are intermittently malformed. Which controlled change is most appropriate for this situation?
- Compare new-index behavior against a pinned evaluation corpus with product versions.
- Repeat representative stochastic cases and measure invalid-output frequency.
- Include citation-source correspondence in release regression assertions.
- Use representative held-out cases and independent acceptance criteria.
Explanation
Correct Answer
B:Generative variability requires repeated trials to estimate structural reliability.
Incorrect Answers
A:Versioned retrieval regression distinguishes improvement from lost historical coverage, whereas generative variability requires repeated trials to estimate structural reliability.
C:Evidence fidelity is part of output correctness, not a decorative addition, but generative variability requires repeated trials to estimate structural reliability.
D:An evaluation must generalize beyond the small examples the developer optimized. Generative variability requires repeated trials to estimate structural reliability.
Question 291
An agent has a strict source-citation requirement but engineers only test the final text summary. Which design choice targets the observed problem?
- Test exactly-once business effect through idempotency and outcome reconciliation.
- Compare new-index behavior against a pinned evaluation corpus with product versions.
- Include citation-source correspondence in release regression assertions.
- Run a representative fixed regression set before promoting the prompt.
Explanation
Correct Answer
C:Evidence fidelity is part of output correctness, not a decorative addition.
Incorrect Answers
A:Business transaction correctness matters more than the number of network retries. Evidence fidelity is part of output correctness, not a decorative addition.
B:Versioned retrieval regression distinguishes improvement from lost historical coverage, but evidence fidelity is part of output correctness, not a decorative addition.
D:A stable baseline detects collateral behavioral changes in other business intents; evidence fidelity is part of output correctness, not a decorative addition.
Question 292
A deployment introduces a new retrieval index and the bot answers older product questions incorrectly. Which concrete solution decision should the team make?
- Run a representative fixed regression set before promoting the prompt.
- Repeat representative stochastic cases and measure invalid-output frequency.
- Test that retrieval returns no restricted content for the salesperson role.
- Compare new-index behavior against a pinned evaluation corpus with product versions.
Explanation
Correct Answer
D:Versioned retrieval regression distinguishes improvement from lost historical coverage.
Incorrect Answers
A:A stable baseline detects collateral behavioral changes in other business intents, but versioned retrieval regression distinguishes improvement from lost historical coverage.
B:Generative variability requires repeated trials to estimate structural reliability. Versioned retrieval regression distinguishes improvement from lost historical coverage.
C:Permission-bound knowledge retrieval must enforce the requesting user’s rights; versioned retrieval regression distinguishes improvement from lost historical coverage.
Question 293
An agent works with one test user but becomes slow at five hundred concurrent requests. Which recommendation addresses the identified failure most directly?
- Exercise failover and fallback under a simulated endpoint outage.
- Run concurrency load tests with end-to-end latency and saturation thresholds.
- Test privilege enforcement under role-spoofing user prompts.
- Test bounded backoff, graceful rate-limit handling and queue pressure.
Explanation
Correct Answer
B:Scalability cannot be inferred from single-user success.
Incorrect Answers
A:A recovery drill verifies the declared continuity behavior instead of assuming it works. Scalability cannot be inferred from single-user success.
C:Authenticated tool authorization must not be controlled by user prose, whereas scalability cannot be inferred from single-user success.
D:Dependency throttling is a common peak-load failure path that must be simulated. Scalability cannot be inferred from single-user success.
Question 294
An external action service sometimes rate-limits requests at high throughput. What should the solution architect recommend for this specific need?
- Exercise failover and fallback under a simulated endpoint outage.
- Test propagation from supply chain event to customer-facing service response.
- Test bounded backoff, graceful rate-limit handling and queue pressure.
- Test exactly-once business effect through idempotency and outcome reconciliation.
Explanation
Correct Answer
C:Dependency throttling is a common peak-load failure path that must be simulated.
Incorrect Answers
A:A recovery drill verifies the declared continuity behavior instead of assuming it works; dependency throttling is a common peak-load failure path that must be simulated.
B:Eventual consistency and freshness need verification across business app boundaries, whereas dependency throttling is a common peak-load failure path that must be simulated.
D:Business transaction correctness matters more than the number of network retries, but dependency throttling is a common peak-load failure path that must be simulated.
Question 295
A business asks for recovery after model endpoint unavailability for five minutes. What is the most defensible intervention for this case?
- Test exactly-once business effect through idempotency and outcome reconciliation.
- Run concurrency load tests with end-to-end latency and saturation thresholds.
- Test partial-failure recovery and a coherent escalation outcome.
- Exercise failover and fallback under a simulated endpoint outage.
Explanation
Correct Answer
D:A recovery drill verifies the declared continuity behavior instead of assuming it works.
Incorrect Answers
A:Business transaction correctness matters more than the number of network retries, whereas a recovery drill verifies the declared continuity behavior instead of assuming it works.
B:Scalability cannot be inferred from single-user success. A recovery drill verifies the declared continuity behavior instead of assuming it works.
C:End-to-end cases must account for real dependency loss without fabricated completion, but a recovery drill verifies the declared continuity behavior instead of assuming it works.
Question 296
The agent writes records and network timeouts can occur after the remote system commits the write. Which architecture response most directly addresses the risk?
- Test exactly-once business effect through idempotency and outcome reconciliation.
- Test bounded backoff, graceful rate-limit handling and queue pressure.
- Run concurrency load tests with end-to-end latency and saturation thresholds.
- Test negative approval branches and assert no downstream payment side effect.
Explanation
Correct Answer
A:Business transaction correctness matters more than the number of network retries.
Incorrect Answers
B:Dependency throttling is a common peak-load failure path that must be simulated, but business transaction correctness matters more than the number of network retries.
C:Scalability cannot be inferred from single-user success. Business transaction correctness matters more than the number of network retries.
D:A failure-path test proves control gates work across integrated applications; business transaction correctness matters more than the number of network retries.
Question 297
Stakeholders approve a prototype based only on demo enthusiasm and have no thresholds for production quality. Which solution approach best meets the stated requirement?
- Require accessibility journey testing as a separate acceptance criterion.
- Treat data-protection failure as a release blocker even when accuracy passes.
- Agree measurable factual, safety, completion, latency and cost acceptance thresholds.
- Run concurrency load tests with end-to-end latency and saturation thresholds.
Explanation
Correct Answer
C:Predefined gates make release decisions reproducible and connected to business requirements.
Incorrect Answers
A:Functional correctness alone cannot establish inclusive usability, but predefined gates make release decisions reproducible and connected to business requirements.
B:Security is an independent acceptance dimension, not a tradeable accuracy bonus. Predefined gates make release decisions reproducible and connected to business requirements.
D:Scalability cannot be inferred from single-user success, whereas predefined gates make release decisions reproducible and connected to business requirements.
Question 298
An AI classifier reaches its accuracy target but exposes confidential examples in output. Which option most directly closes the documented gap?
- Use representative held-out cases and independent acceptance criteria.
- Require named operational ownership and an incident escalation procedure.
- Require accessibility journey testing as a separate acceptance criterion.
- Treat data-protection failure as a release blocker even when accuracy passes.
Explanation
Correct Answer
D:Security is an independent acceptance dimension, not a tradeable accuracy bonus.
Incorrect Answers
A:An evaluation must generalize beyond the small examples the developer optimized. Security is an independent acceptance dimension, not a tradeable accuracy bonus.
B:Production readiness includes accountability beyond model quality scores; security is an independent acceptance dimension, not a tradeable accuracy bonus.
C:Functional correctness alone cannot establish inclusive usability, whereas security is an independent acceptance dimension, not a tradeable accuracy bonus.
Question 299
An agent provides correct answers but cannot be used with screen readers by part of the workforce. Which approach best satisfies the actual business constraint?
- Require accessibility journey testing as a separate acceptance criterion.
- Measure task correctness and verified outcomes instead of fluency alone.
- Agree measurable factual, safety, completion, latency and cost acceptance thresholds.
- Require named operational ownership and an incident escalation procedure.
Explanation
Correct Answer
A:Functional correctness alone cannot establish inclusive usability.
Incorrect Answers
B:Operational success requires the correct business result, not merely polished language, whereas functional correctness alone cannot establish inclusive usability.
C:Predefined gates make release decisions reproducible and connected to business requirements. Functional correctness alone cannot establish inclusive usability.
D:Production readiness includes accountability beyond model quality scores, but functional correctness alone cannot establish inclusive usability.
Question 300
A regulated finance agent is ready technically but has no documented owner to monitor incidents. Which specific remediation should the business adopt?
- Agree measurable factual, safety, completion, latency and cost acceptance thresholds.
- Require named operational ownership and an incident escalation procedure.
- Treat data-protection failure as a release blocker even when accuracy passes.
- Measure task correctness and verified outcomes instead of fluency alone.
Explanation
Correct Answer
B:Production readiness includes accountability beyond model quality scores.
Incorrect Answers
A:Predefined gates make release decisions reproducible and connected to business requirements, whereas production readiness includes accountability beyond model quality scores.
C:Security is an independent acceptance dimension, not a tradeable accuracy bonus, but production readiness includes accountability beyond model quality scores.
D:Operational success requires the correct business result, not merely polished language. Production readiness includes accountability beyond model quality scores.
Agentic AI Business Solutions Architect (AB-100) Practice Tests