Microsoft AB-100 Practice Test 59

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.

Question 581

A municipal app receives complaints in multiple languages and must classify them into approved departments. Incorrect automatic translation has changed street names. What processing sequence is safer?

Classify solely from the alphabet used without interpreting the complaint content.
Use an English-only classifier and treat all unsupported languages as public relations requests.
Detect language, preserve the original text and location entities, then test classification with localized examples.
Replace every original complaint with its translated version and delete the source wording.

Explanation

Correct Answer

A multi-language pipeline must maintain evidence from the original submission and verify that translation does not corrupt identifiers.

Incorrect Answers

Language identification alone cannot tell which department owns the underlying service issue.

A forced default category can misroute time-sensitive safety or infrastructure complaints.

Discarding the original removes a way to detect altered street names and challenge routing errors.

Question 582

A code-first generative page proposes a dynamic view of purchase orders. The model response includes executable script and a direct database write command. What should the host application permit?

Execute the generated script in the user’s browser because the model described it as necessary.
Let the page bypass Dataverse permissions when the user requests a faster response.
Apply the generated database command immediately before the user sees the proposed UI.
Only validated allowed UI schema/components, with any write handled by separately authorized typed actions.

Explanation

Correct Answer

Generated presentation is untrusted output and must not become arbitrary executable code with access to business data.

Incorrect Answers

Natural-language rationale does not authorize code execution or prove it respects the app’s security model.

Performance convenience does not grant access to restricted business records.

A page preview is not consent to modify purchase-order records or a valid transaction boundary.

Question 583

A Power Apps reporting page generates a filter for “my overdue tasks.” The model inserts a condition returning every employee’s tasks. What must the app enforce?

Intersect generated filter expressions with server-enforced user permissions and approved task-status semantics.
Use the generated filter as-is if it is syntactically valid JSON.
Retrieve all tasks and hide names after the page is rendered.
Ask the model to include the words “for current user” while leaving the backend query unrestricted.

Explanation

Correct Answer

A model-suggested query cannot widen data visibility beyond what the current user is authorized to view.

Incorrect Answers

Valid syntax says nothing about whether other employees’ tasks may be returned.

Late masking cannot prevent unauthorized records being sent into client state or model context.

Textual labels do not enforce actual record filtering on the data service.

Question 584

An AI-generated reconciliation page highlights apparent customer address errors and suggests corrections. The user has read-only access to the customer master. What is the correct action behavior?

Store corrections only in the page’s local display and claim the master data was repaired.
Show proposed changes and route them to an authorized steward workflow without writing from the read-only session.
Commit the suggestions directly because displaying them in the page indicates user interest.
Change the session’s user ID to the steward’s identity before the write.

Explanation

Correct Answer

A generated edit recommendation cannot override the source master’s role-based modification permissions.

Incorrect Answers

Local UI changes do not update the authoritative data source and create a false completion claim.

A visualization event is neither explicit record authorization nor reliable confirmation of correct data.

Identity impersonation bypasses the permission design and invalidates audit attribution.

Question 585

A generative Power Apps page builds a service-case dashboard with unlabeled icon-only approval buttons. Some agents rely on keyboard navigation and screen readers. What acceptance gate is appropriate?

Increase icon size but leave the button purpose unstated to assistive technologies.
Let a model infer a control’s meaning after the user clicks an unidentified button.
Validate accessible names, focus order and keyboard operation against the same business-action controls.
Exclude keyboard-only users from the acceptance set because most staff use a mouse.

Explanation

Correct Answer

Dynamic UI generation still needs accessible interaction semantics and safe approvals for all intended users.

Incorrect Answers

Larger visuals alone do not give a screen reader the action’s accessible name.

Users must know which business action a control performs before invoking it.

Access needs are part of the operational requirement, not a reason to remove affected employees from testing.

Question 586

A generative business page displays an approval table on a narrow phone screen. The model proposes an unrestricted inline HTML fragment that overflows and hides the reject control. What should the frontend do?

Allow overflow as long as the approve button remains visible.
Disable the reject action on mobile to reduce horizontal scrolling.
Trust the generated HTML whenever it contains a viewport meta tag.
Use validated responsive components and test that both approve and reject actions remain visible and operable.

Explanation

Correct Answer

Layout generation must preserve action availability across viewport sizes without executing arbitrary HTML or code.

Incorrect Answers

Hiding reject distorts the approval choice and violates basic functional parity on mobile.

A presentation constraint must not silently remove a required business decision.

A viewport declaration does not validate responsive layout, permissions or control behavior.

Question 587

A purchasing agent has a topic for invoice-status requests requiring an invoice ID. A user asks “Where is invoice INV-642?” but the generative fallback is answering from unrelated policy text. Which routing change addresses the error?

Prioritize the invoice-status topic and capture INV-642 as the validated invoice identifier before calling its lookup action.
Add the invoice status to a global prompt and allow the fallback to guess payment state.
Disable the invoice topic so the fallback sees every conversation.
Route any message containing a number directly to the refund-approval topic.

Explanation

Correct Answer

The known intent and supplied entity should select the specialized workflow rather than an unconstrained knowledge response.

Incorrect Answers

A prompt cannot discover the actual transaction status without the invoice-specific system-of-record lookup.

Removing the targeted handler increases ambiguity and loses typed invoice identifiers and action controls.

Numeric tokens alone do not distinguish invoice status from approval or other transactional intents.

Question 588

A Copilot Studio assistant hears “change my delivery,” which can mean delivery address or scheduled delivery date. Both actions modify orders. What should its conversation do before invoking either action?

Send the entire conversation to a general knowledge topic as though it were a read-only FAQ.
Ask a clarifying question and route to the matching action only after the requested field and order authorization are verified.
Always choose address update because it is the most frequently used topic.
Run both update actions and roll back whichever the user rejects.

Explanation

Correct Answer

A single utterance fits two different write intents; disambiguation prevents modifying the wrong order attribute.

Incorrect Answers

A retrieval answer cannot commit an authorized update to the correct order field.

Historic intent frequency cannot establish which field this individual customer wants to change.

Executing two unintended writes can trigger shipping changes and does not satisfy prior user authorization.

Question 589

An employee asks whether a local benefit plan covers an uncommon procedure. No active topic matches and the approved benefits source returns no clause. What fallback should be designed?

Route the request to payroll adjustment because both topics mention employee benefits.
Use an archived benefits handbook as definitive evidence without checking its effective period.
Explain that the answer lacks verified plan evidence and offer a documented HR escalation route.
Generate an eligibility percentage from the model confidence score.

Explanation

Correct Answer

An explicit unsupported-answer state preserves trust when neither a structured topic nor approved source can justify eligibility.

Incorrect Answers

A payroll action is unrelated to determining plan coverage and may expose unauthorized changes.

Superseded terms may differ from the employee’s active plan and cannot support an authoritative answer.

Language-model confidence does not establish a contractual coverage decision for this employee.

Question 590

An airline customer moves from a flight-change topic to a refund topic. The previous topic already submitted a rebooking request, but its response was delayed. How should the handoff prevent duplicate changes?

Start a fresh rebooking action each time the dialog enters a new topic.
Discard the booking identifier as soon as the refund topic begins.
Return “rebooking complete” because the previous topic generated a confident acknowledgement.
Carry a stable booking correlation ID and query the rebooking operation state before attempting another mutation.

Explanation

Correct Answer

Topic transitions do not erase outstanding transaction state; reconciliation avoids two rebookings for one intent.

Incorrect Answers

New action IDs prevent deduplication and can create multiple purchases or reservations.

Losing the original transaction reference makes safe compensation and status checks harder.

Conversational success text does not prove the external booking service committed the operation.

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

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)
Microsoft AB-100 Practice Test 61 - 10 Questions (Last Reviewed 10-Nov-2026)

Pass Your Microsoft AB-100 Exam Easy!

Question 580

Explanation

Correct Answer

Incorrect Answers

Question 581

Explanation

Correct Answer

Incorrect Answers

Question 582

Explanation

Correct Answer

Incorrect Answers

Question 583

Explanation

Correct Answer

Incorrect Answers

Question 584

Explanation

Correct Answer

Incorrect Answers

Question 585

Explanation

Correct Answer

Incorrect Answers

Question 586

Explanation

Correct Answer

Incorrect Answers

Question 587

Explanation

Correct Answer

Incorrect Answers

Question 588

Explanation

Correct Answer

Incorrect Answers

Question 589

Explanation

Correct Answer

Incorrect Answers

Question 590

Explanation

Correct Answer

Incorrect Answers

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