Microsoft AB-100 Practice Test 58

Question 570

An agent may automatically tag low-risk support cases but management wants it to delete cases once resolved. Legal retention periods vary by case. What should precede enabling deletion?

Allow deletion whenever a generated summary says the customer was satisfied.
Grant the agent global case-delete privileges to reduce administration costs.
Use a prompt describing retention limits without validating the actual case category in the service.
A reviewed retention decision, service-side permission gate and tested restore or archival procedure for affected case classes.

Explanation

Correct Answer

Deletion has different permanence and compliance risk from reversible tagging and needs explicit lifecycle controls.

Incorrect Answers

Satisfaction text does not establish completion of legally required retention and appeal periods.

Excess rights can permanently remove cases outside the intended low-risk classification.

A policy instruction alone cannot stop a deletion tool from removing a record under hold.

Question 571

A banking assistant autonomously routes straightforward support questions, but the request “change the account beneficiary” has multiple possible account matches. What should the flow do?

Request verified identity and a specific account selection, then route the actual change through authorized confirmation.
Pick whichever account has the highest semantic similarity to the chat wording.
Change every possible account to ensure the customer’s request is fulfilled.
Defer the identity check until after the changes have been written.

Explanation

Correct Answer

An ambiguous high-impact account action should stop for disambiguation and enforceable permission checks.

Incorrect Answers

Text resemblance is not a reliable record identity or authority for a banking change.

A broad update would exceed the request and modify records that may belong to other principals.

Postcommit verification cannot undo the unauthorized transaction risk created at the execution boundary.

Question 572

An employee policy agent is deployed with approved read-only document tools. A user asks it to update their payroll bank account directly from the conversation. What is the correct behavior?

Paste the requested bank details into a shared prompt library as a substitute for payroll processing.
Explain the read-only boundary and direct the employee to the verified payroll-change workflow.
Invent a successful account-update receipt while keeping the tool read-only.
Use a hidden administrator connector to perform the update despite the configured read-only policy.

Explanation

Correct Answer

A knowledge agent cannot turn a conversational request into payroll write privileges or identity verification.

Incorrect Answers

A public prompt artifact is neither a secure change service nor a valid place for banking details.

A chat confirmation would falsely represent a financial change that was never committed.

An undeclared privileged tool bypasses the deliberately narrow knowledge-agent scope.

Question 573

A quality-compliance bot answers employee questions about inspection frequencies from a controlled manual. Its index also contains discarded drafts. How should the prompt-response architecture work?

Generate an inspection interval from typical industry practice when the manual is unavailable.
Remove citations so employees cannot see that the model used a draft.
Retrieve currently approved manual passages with version provenance and abstain if no binding instruction is found.
Choose the most semantically similar draft regardless of publication status.

Explanation

Correct Answer

Read-only answering still needs policy authority and a traceable citation to avoid unsupported operational advice.

Incorrect Answers

General habits cannot establish the organization’s legally approved requirement for the equipment.

Concealing provenance makes it harder to detect an unauthorized policy source.

Relevance ranking does not establish that a rejected procedure is suitable for an inspection.

Question 574

A read-only benefits agent answers in English and Spanish. The Spanish translation of “dependent” has several ordinary meanings, but policy defines one legal eligibility term. What should the design include?

Allow the model to choose whichever translation sounds most fluent in each response.
Treat the two languages as separate policy versions with independent employee entitlement rules.
Remove the relevant source clause to prevent the agent from being influenced by legal terminology.
An approved bilingual terminology mapping that retains the policy’s defined meaning and cites the controlling clause.

Explanation

Correct Answer

Translated answers must preserve the legal business concept rather than rely on ambiguous everyday word associations.

Incorrect Answers

Fluency does not guarantee consistency with the formal eligibility definition.

Language choice alone cannot alter the approved benefits eligibility criteria.

Withholding the authoritative definition increases the risk of unsupported translated entitlement claims.

Question 575

A knowledge assistant answers facility manuals, but a user asks it to authorize entry to a restricted laboratory. It has no badge-management integration. What response is appropriate?

Acknowledge the unsupported action and refer the request to the authenticated access-approval service.
Tell the requester that access was granted because a policy excerpt describes normal opening hours.
Create a fake badge number in the reply to make the workflow appear complete.
Instruct the user to bypass the lab door control because the assistant cannot call the proper tool.

Explanation

Correct Answer

Read-only procedural guidance cannot create a valid badge permission without the approved system.

Incorrect Answers

Opening hours do not establish individual security clearance or an actual access-control update.

An invented identifier cannot authorize physical entry and may mislead staff.

Lack of integration must lead to a safe referral, not a workaround that defeats security.

Question 576

A public product FAQ agent can explain warranty terms, but users ask whether their individual product is currently registered. The agent has no authorized serial lookup. How should it answer?

Use another customer’s serial number from a retrieved example as the query target.
Explain published coverage conditions but state that individual registration requires the verified customer portal.
Infer registration from how old the product sounds in the customer’s message.
Claim all products are registered to avoid disappointing customers.

Explanation

Correct Answer

General warranty text cannot establish live account-specific ownership or entitlement without a transactional source.

Incorrect Answers

A sample record is not the current user’s authenticated product identity.

Age cues and product descriptions are not authoritative registration records.

An unsupported assertion can create false warranty expectations for unregistered equipment.

Question 577

An accounts-payable app receives scanned multi-page invoices. It needs supplier, invoice number, tables and totals in structured fields before a verification agent processes them. What tool capability fits?

A language detection endpoint as the entire invoice processing solution.
Direct financial posting based solely on a model’s prose description of the invoice.
A document extraction or document intelligence capability configured for fields and line-item tables, followed by validation.
A generic speech transcription service for printed invoice pages.

Explanation

Correct Answer

The task needs structured layout-aware extraction from documents, not only unconstrained narrative summarization.

Incorrect Answers

Identifying document language does not provide supplier identifiers, lines or payable amounts.

Payment requires validated typed fields and business approval after the extraction stage.

Speech processing does not extract structured tables from scanned paper documents.

Question 578

A service center has recorded customer calls and wants a downstream agent to summarize refund promises with timestamps and speakers. What should the first modality conversion do?

Use document OCR on the audio file as though it were a scanned invoice.
Use only a sentiment score and infer the exact promised refund amount.
Generate a transcript from generic call-center patterns without processing the recording.
Produce timestamped speech-to-text with speaker attribution where supported, then evaluate the refund statements.

Explanation

Correct Answer

Grounded call analysis requires a trustworthy audio transcript and timeline before semantic interpretation.

Incorrect Answers

Optical character recognition cannot transcribe speech waveforms or attribute spoken turns.

Sentiment does not preserve what was actually said or the monetary terms of a call.

Typical conversation structure cannot prove this caller’s actual statements.

Question 579

A warranty agent must detect cracked screens in customer-uploaded phone images before assessing service eligibility. Which capability should provide the visual evidence?

An approved image/vision analysis component evaluated on labeled defect photographs, with uncertain cases reviewed.
A text-only keyword search for the phrase “cracked screen” in the upload filename.
An audio transcription model operating on the JPEG bytes.
An autonomous refund call based on model confidence without examining the warranty policy.

Explanation

Correct Answer

The defect is visible in pixels, so textual language extraction alone does not establish the physical damage.

Incorrect Answers

A filename can be inaccurate or intentionally misleading and does not verify visible damage.

Speech-to-text is not the modality for inspecting images of physical product damage.

Visual evidence and contract eligibility are separate requirements; a model score does not grant refund authority.

Question 580

A records office must identify personal identifiers in plain-text complaint emails before indexing them for a public FAQ assistant. Which capability should be evaluated?

A sentiment model that reports whether the writer is happy or frustrated.
PII entity detection with measured precision and recall, coupled with deterministic redaction and review.
A topic classifier that labels every email as general support without finding actual identifier spans.
A summarizer instructed not to mention private information after the raw email is already indexed publicly.

Explanation

Correct Answer

The requirement is identifying sensitive spans in text before retrieval exposure, not producing a creative summary.

Incorrect Answers

Emotion classification does not indicate which tokens are personally identifying and need redaction.

Topic classification cannot reliably locate specific account numbers or personal names that require removal.

Post-retrieval text editing does not undo unauthorized storage or discovery of sensitive source content.

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

Microsoft AB-100 Practice Test 56 - 10 Questions (Last Reviewed 19-Sep-2026)
Microsoft AB-100 Practice Test 57 - 10 Questions (Last Reviewed 14-Nov-2026)
Microsoft AB-100 Practice Test 58 - 10 Questions (Last Reviewed 13-Nov-2026)
Microsoft AB-100 Practice Test 59 - 10 Questions (Last Reviewed 12-Nov-2026)
Microsoft AB-100 Practice Test 60 - 10 Questions (Last Reviewed 11-Nov-2026)

Pass Your Microsoft AB-100 Exam Easy!

Question 570

Explanation

Correct Answer

Incorrect Answers

Question 571

Explanation

Correct Answer

Incorrect Answers

Question 572

Explanation

Correct Answer

Incorrect Answers

Question 573

Explanation

Correct Answer

Incorrect Answers

Question 574

Explanation

Correct Answer

Incorrect Answers

Question 575

Explanation

Correct Answer

Incorrect Answers

Question 576

Explanation

Correct Answer

Incorrect Answers

Question 577

Explanation

Correct Answer

Incorrect Answers

Question 578

Explanation

Correct Answer

Incorrect Answers

Question 579

Explanation

Correct Answer

Incorrect Answers

Question 580

Explanation

Correct Answer

Incorrect Answers

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