Microsoft AB-100 Practice Test 15

Question 140

A canvas app sends a customer complaint to an AI summarization service but must not display the draft outside the authorized case team. Which design is appropriate?

Place the entire CRM database in the browser so users can pick records.
Set a hidden UI flag as the only access-control mechanism.
Use scoped app/Dataverse permissions and pass only the minimum case text to the summary service.
Share summaries through a public anonymous link for faster review.

Explanation

Correct Answer

App and service access must enforce caller rights before sending or showing sensitive case data.

Incorrect Answers

Broad browser data increases exposure.

A hidden element does not enforce server-side permissions.

Anonymous sharing bypasses case roles.

Question 141

A Power Platform intelligent app experiences high response latency during peak demand. Which Well-Architected pillar guides capacity and caching tradeoffs?

Performance Efficiency, balanced with data correctness, cost and operational monitoring.
Reliability solely by declaring service-level targets without load tests.
Improve the canvas page rendering time and animations first, without measuring how model and connector stages affect peak end-to-end latency.
Security alone, ignoring concurrency and throughput.

Explanation

Correct Answer

Performance planning requires measured workload volume and architectural tuning.

Incorrect Answers

Targets without empirical testing do not establish performance.

Client-side UI changes cannot resolve a server-side model, connector or concurrency bottleneck without stage-specific latency evidence.

Security is essential but does not by itself size capacity.

Question 142

A Copilot Studio agent routes a fixed five-category incident queue. Historical labels are stable and testers need confusion-matrix metrics. What design is suitable?

Use a bounded intent classifier or CLU capability validated against labeled utterances.
Use the nearest embedding match as the final queue label with no calibrated confidence threshold or controlled unknown-intent fallback.
Use a generic model with no stable category schema or evaluation.
Generate a free-form category name at each request with no allowed-value check.

Explanation

Correct Answer

A bounded classification task benefits from explicit labels and measured precision/recall.

Incorrect Answers

Semantic proximity is not proof of a correct fixed label. Evaluate routing errors on labeled incidents and handle uncertain classifications.

Unconstrained categories make accuracy difficult to measure.

Unvalidated labels can break routing automation.

Question 143

A Copilot Studio agent flow updates an order and posts to a payment service. The second call fails after the order was committed. What recovery should be included?

Mark payment successful because the order write succeeded.
Retry the entire order-and-payment sequence with a new transaction identifier whenever the payment call fails.
Retry the order creation with a new ID every time payment fails.
Record durable intermediate state and apply a compensating refund/cancellation or manual intervention.

Explanation

Correct Answer

Multi-service workflows need explicit partial-failure handling and safe idempotency.

Incorrect Answers

One committed step does not imply end-to-end success.

Changing identifiers may create a second committed order or charge; a durable workflow must reconcile which stages completed before compensating or retrying.

New IDs risk duplicate orders.

Question 144

A prompt action produces purchase-category codes for downstream routing. Which response design improves reliability?

Accept any adjective and expect the buyer tool to infer the code.
Use the first word of the model response as the category ID.
Constrain the code to an allowed enumeration and validate its business meaning before execution.
Treat a valid JSON shape as sufficient proof the category is correct.

Explanation

Correct Answer

Structured output validation reduces parsing failures; substantive mapping still needs business checks.

Incorrect Answers

Free-form adjectives can be ambiguous.

Word order is not a stable identifier contract.

Schema validity is not semantic correctness.

Question 145

A Dynamics 365 Sales record labels an opportunity “Closed - Won,” but a Copilot report calls all closed records “won,” including lost deals. What semantic mapping is required?

Distinguish close-state and outcome codes so only genuinely won opportunities are aggregated.
Let the model choose the most positive outcome for every closed opportunity.
Add closed-lost deals into won revenue to make totals stable.
Treat any opportunity whose general state begins with “Closed” as won revenue, without evaluating its status reason.

Explanation

Correct Answer

Business terminology must map to exact CRM status and outcome fields before reporting.

Incorrect Answers

Optimistic guesses cannot replace source-state semantics.

Misclassifying lost business overstates performance.

Closed-won and closed-lost can share the closed state while contributing very different accounting outcomes.

Question 146

A customer service Copilot can draft case closure notes, but a customer’s support contract forbids automated closure of safety incidents. How should the customization work?

Assume every confident closure summary means policy has been met.
Keep a prompt warning only and permit the agent to call the closure tool anyway.
Allow drafting but enforce safety-case type checks in the close action, routing restricted cases to an authorized reviewer.
Hide the safety field from the prompt and allow default closure.

Explanation

Correct Answer

Business-system constraints must prevent prohibited mutations independent of generated language.

Incorrect Answers

Confidence is not contractual closure authority.

An instruction cannot replace server-side authorization and policy.

Removing safety context increases risk rather than reducing it.

Question 147

A Copilot Studio connector shows Dynamics 365 Sales opportunity details but uses a cache that has not been refreshed since yesterday. An agent must give current account status. What should be designed?

Ignore server record versions whenever search relevance is high.
Assume the opportunity cache is current because its connector has low network latency, without comparing record versions or timestamps.
An explicit freshness policy with authoritative refresh or a timestamped unavailable-current-status response.
Ask the model to predict today’s actual state from older opportunity text.

Explanation

Correct Answer

Connector availability and data currency are independent, so the response must honor effective time.

Incorrect Answers

Relevance does not prove a field has not changed.

Fast retrieval can serve stale data; transaction freshness depends on source state, not response speed.

An estimate is not the current CRM record.

Question 148

A Contact Center agent transfers a distressed caller from voice to a live support specialist. What should the handoff include?

Transfer only a sentiment score without the authenticated case ID, supporting evidence or actions already completed.
Authenticated case context, a concise evidence-grounded summary and the reason human intervention is required.
The complete unrestricted transcript of every unrelated customer call.
A claim that the human already approved the resolution without a logged handoff.

Explanation

Correct Answer

A safe channel transfer preserves relevant authorized context and avoids forcing customers to repeat verified details.

Incorrect Answers

Sentiment can inform agent empathy but cannot give a human specialist the verified customer context or pending escalation steps.

Cross-customer data disclosure violates privacy.

The receiving specialist needs a real assigned case and action state.

Question 149

A task agent checks a delivery and then opens a ticket when tracking is delayed. The ticket API returns 409 duplicate. What is the correct workflow response?

Look up the existing ticket by stable correlation key and reuse or update it rather than creating another.
Assign a fresh random key and retry until two tickets are created.
Mark the duplicate response as proof the customer was refunded.
Delete the shipment record and retry the ticket action from scratch.

Explanation

Correct Answer

Idempotent case handling prevents duplicate workflow side effects and preserves a single issue history.

Incorrect Answers

Repeated creation can multiply work and confuse agents.

A duplicate error has no relationship to a refund.

Deleting source shipment data damages traceability.

Question 150

An autonomous service agent may issue low-value credits when an outage is independently verified. A customer asks for a credit when telemetry is missing. What should happen?

Treat the customer’s request itself as verified outage telemetry.
Withhold autonomous issuance and escalate for source verification or authorized manual review.
Treat the customer-service agent’s high confidence in its outage classification as enough evidence to issue a credit.
Grant the agent new billing permissions to force a successful transaction.

Explanation

Correct Answer

Delegated actions must satisfy their evidence gates; missing data is not positive proof.

Incorrect Answers

A request is an allegation, not independent system status.

A model confidence score does not replace independent outage telemetry or an authorized eligibility review.

Privilege escalation cannot repair evidence gaps.

Agentic AI Business Solutions Architect (AB-100) Practice Tests

Microsoft AB-100 Practice Test 13 - 10 Questions (Last Reviewed 01-Nov-2026)
Microsoft AB-100 Practice Test 14 - 10 Questions (Last Reviewed 31-Oct-2026)
Microsoft AB-100 Practice Test 15 - 10 Questions (Last Reviewed 30-Oct-2026)
Microsoft AB-100 Practice Test 16 - 10 Questions (Last Reviewed 29-Oct-2026)
Microsoft AB-100 Practice Test 17 - 10 Questions (Last Reviewed 28-Oct-2026)

Pass Your Microsoft AB-100 Exam Easy!

Question 140

Explanation

Correct Answer

Incorrect Answers

Question 141

Explanation

Correct Answer

Incorrect Answers

Question 142

Explanation

Correct Answer

Incorrect Answers

Question 143

Explanation

Correct Answer

Incorrect Answers

Question 144

Explanation

Correct Answer

Incorrect Answers

Question 145

Explanation

Correct Answer

Incorrect Answers

Question 146

Explanation

Correct Answer

Incorrect Answers

Question 147

Explanation

Correct Answer

Incorrect Answers

Question 148

Explanation

Correct Answer

Incorrect Answers

Question 149

Explanation

Correct Answer

Incorrect Answers

Question 150

Explanation

Correct Answer

Incorrect Answers

Agentic AI Business Solutions Architect (AB-100) Practice Tests