Microsoft AB-100 Practice Test 71

Question 700

An agent’s failed shipment tool calls rise after the carrier API starts returning HTTP 429. The model token volume remains stable. What should operators examine?

Mark each throttled attempt as completed because the HTTP service responded.
Rate-limit responses, retry backoff, concurrency and carrier quota utilization.
Increase retries to zero delay while retaining the same request burst.
Switch the agent’s display language without changing tool behavior.

Explanation

Correct Answer

429 signals a throttled downstream dependency; load shaping is more relevant than generated-answer verbosity.

Incorrect Answers

An error status is not a successful shipment operation.

Immediate loops amplify throttling and may worsen the carrier outage.

Translation does not create API capacity or reduce calls.

Question 701

A finance agent’s monthly inference cost doubles. Completed invoices are unchanged, but average prompt tokens per case doubled after a knowledge-source update. What should the team investigate?

Disable all source retrieval to minimize tokens.
Increase the model context window without examining retrieved chunk relevance.
Compare per-task retrieved context size, token usage and unchanged completion outcomes before optimizing chunk selection.
Assume costs doubled because there were twice as many invoices.

Explanation

Correct Answer

Additional irrelevant context can raise cost without improving the measured business task.

Incorrect Answers

Removing grounding may lower cost but cause inaccurate financial decisions.

A larger context budget can encourage the same unnecessary input inflation.

The verified business volume is unchanged, contradicting that explanation.

Question 702

A claims agent sometimes invents unsupported deductible amounts and sometimes submits a valid claim twice after timeout. Which TWO telemetry signals should be tracked separately? (Select TWO.)

Only the average number of generated words per reply.
Only the chat window theme selected by the claims user.
An aggregate success label written by the agent itself.
Citation-supported deductible correctness against the effective policy revision.
Business transaction ID and idempotency/commit outcome for each claim submission attempt.

Explanation

Correct Answers

Factual grounding quality indicates whether the deductible was justified by valid evidence.

A durable transaction trace reveals duplicate writes even when the narrated answer is correct.

Incorrect Answers

Answer length cannot identify a wrong policy amount or duplicated submission.

UI appearance does not establish policy accuracy or transaction idempotence.

Self-reported completion cannot independently verify cited amounts or backend committed writes.

Question 703

A language-model deployment remains unchanged, but a new document retrieval adapter makes refund answers stale. What segmentation identifies the real regression?

Slice errors by retrieval adapter/version, source effective date and model deployment while holding the baseline cohort fixed.
Attribute all failures to the model because users interact with a chatbot.
Measure only daily message count without source or version details.
Replace the approved policy with generated summaries during testing.

Explanation

Correct Answer

Comparing source versions separates grounding changes from stable model behavior.

Incorrect Answers

The visible interface is not evidence of where incorrect data entered the pipeline.

Volume cannot identify which adapter introduced obsolete policies.

Synthetic policy text would obscure the actual source-version root cause.

Question 704

A customer-service agent uses cached refund answers. Legal changes the policy but cache hit rates stay high while correctness falls. What needs to be monitored?

Ask the generator to mention the current date without changing its cached evidence.
Cache entry source-version validity, expiry and invalidation after approved policy changes.
Cache hit percentage alone as evidence of healthy answers.
Increase cached response retention indefinitely to reduce cost.

Explanation

Correct Answer

A cache can be operationally fast yet factually stale if source authority is not part of its key.

Incorrect Answers

Date wording does not establish that the underlying policy text is current.

High reuse can amplify outdated guidance when invalidation is missing.

Longer reuse worsens stale-answer exposure under changing policy.

Question 705

A customer agent declares “handled” whenever it writes a fluent reply. Tickets stay open in the service application. What completion metric should replace self-reported replies?

Use the number of generated tokens as a completed-case proxy.
Close tickets automatically after an agent response even when no fix was applied.
Count verified case resolutions from the service system, with reversals and reopenings tracked separately.
Count all messages ending in a polite closing phrase.

Explanation

Correct Answer

Business outcome reporting needs the authoritative case state rather than the assistant’s conversational claim.

Incorrect Answers

Tokens measure model activity, not durable ticket resolution.

Artificial closure hides unresolved problems and invalidates operational metrics.

Politeness does not establish that the reported problem was resolved.

Question 706

An invoice agent handles 2,000 payments per week. The throughput report ignores cases paid without required second approval. What outcome dimension must be added?

Report paid invoice volume as a complete measure of successful automation.
Count the number of approver-related sentences the agent produced.
Use average invoice amount to infer that approvals occurred.
Count payments with valid prerequisite approver records and separately report unauthorized approval-bypass incidents.

Explanation

Correct Answer

Completion is only correct when all binding authorization conditions were satisfied before payment.

Incorrect Answers

Throughput conceals policy violations in otherwise fast transactions.

Generated language is not a verified independent human signoff.

Payment magnitude cannot establish which actor authorized the transaction.

Question 707

A receipt extraction agent reports 97% field accuracy but the remaining errors are concentrated in high-value taxes. What should business monitoring capture?

Weight errors by downstream financial impact and measure manual corrections and adjustments per transaction.
Use only the percentage of receipts successfully uploaded.
Ignore mistakes affecting only a few large invoices.
Measure average answer word count as proof of quality.

Explanation

Correct Answer

An aggregate accuracy figure can mask rare costly mistakes that dominate business risk.

Incorrect Answers

File upload does not verify amount or tax extraction correctness.

Low occurrence count does not make high-value financial exposure immaterial.

Explanatory length is not linked to reconciled tax or ledger results.

Question 708

A support agent escalates difficult warranty requests to technicians. Leadership wants to know whether it actually improves service. What is the suitable outcome measure?

Ask the agent to self-score every technician interaction.
Compare resolved-case time, false escalations and technician effort against a comparable pre-agent baseline.
Count all escalated cases as resolved by the agent.
Count only the number of tickets routed to technicians.

Explanation

Correct Answer

Assisted quality includes the downstream workload created by incomplete or unnecessary escalations.

Incorrect Answers

Generated self-assessment cannot verify independent service outcomes.

Escalation is a handoff, not a final repair or completed service outcome.

Volume gives no evidence of faster resolution or better triage quality.

Question 709

A new routing model makes 100 more tool calls per day but completes the same number of approved purchase requests. Which KPI comparison is most meaningful?

Use token volume alone to rank the two versions.
Ignore exceptions because the headline completion count stayed constant.
Compare correctly completed unique purchase intents, tool attempts per intent, total cost and exception rate.
Treat every successful HTTP response as a new purchase.

Explanation

Correct Answer

Extra calls may indicate retries or inefficient routing rather than increased business value.

Incorrect Answers

Model output volume is not a proxy for completed authorized work.

New retries may increase cost, risk and customer impact even at equal totals.

Read-only and retry calls do not always represent unique accepted purchase orders.

Question 710

A legal policy agent refuses to answer unsupported clauses and routes them to counsel. A dashboard labels every abstention as a task failure. What correction is appropriate?

Tell the agent never to abstain regardless of evidence.
Count every escalation as a completed legal approval.
Exclude all abstentions from evaluation denominators without documenting them.
Separate justified abstentions from avoidable inability to answer, using policy support and escalation results.

Explanation

Correct Answer

An evidence-based refusal can be a correct safe behavior when authoritative guidance is missing.

Incorrect Answers

Forced answers can convert an incomplete policy library into confident misinformation.

Routing a case to counsel is not equivalent to an authorized decision.

Removing cases obscures coverage gaps and prevents transparent risk analysis.

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

Microsoft AB-100 Practice Test 69 - 10 Questions (Last Reviewed 02-Nov-2026)
Microsoft AB-100 Practice Test 70 - 10 Questions (Last Reviewed 01-Nov-2026)
Microsoft AB-100 Practice Test 71 - 10 Questions (Last Reviewed 31-Oct-2026)
Microsoft AB-100 Practice Test 72 - 10 Questions (Last Reviewed 30-Oct-2026)
Microsoft AB-100 Practice Test 73 - 10 Questions (Last Reviewed 29-Oct-2026)

Pass Your Microsoft AB-100 Exam Easy!

Question 700

Explanation

Correct Answer

Incorrect Answers

Question 701

Explanation

Correct Answer

Incorrect Answers

Question 702

Explanation

Correct Answers

Incorrect Answers

Question 703

Explanation

Correct Answer

Incorrect Answers

Question 704

Explanation

Correct Answer

Incorrect Answers

Question 705

Explanation

Correct Answer

Incorrect Answers

Question 706

Explanation

Correct Answer

Incorrect Answers

Question 707

Explanation

Correct Answer

Incorrect Answers

Question 708

Explanation

Correct Answer

Incorrect Answers

Question 709

Explanation

Correct Answer

Incorrect Answers

Question 710

Explanation

Correct Answer

Incorrect Answers

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