Microsoft AB-100 Practice Test 4

Question 30

A support assistant analyzes scanned forms with some unreadable signatures. The capture pipeline emits a blank signature field. What data-quality action is appropriate?

Mark the field as missing with source confidence and require verification before using it for an authorized action.
Treat blank as affirmative consent because the scan was processed successfully.
Use the agent’s fluent summary as proof the physical signature exists.
Populate an unreadable signature field with the value from a previous document by the same customer without verifying consent.

Explanation

Correct Answer

Missing evidence must not be converted into a legal assertion; verification and explicit missing-value handling are needed.

Incorrect Answers

Successful extraction does not establish positive consent.

Generated prose is not direct evidence of a signature.

Historical signatures cannot be transplanted into a new legally meaningful approval event; missing evidence requires verification.

Question 31

Two Dynamics 365 applications consume the same account identifiers but differ in their date formats and currency fields. What data product control matters most?

Store the same entity under separate arbitrary IDs to avoid dealing with integration.
Convert all currencies to the first encountered code without recording conversion rates.
Publish documented canonical identifier, currency and date semantics with interface-level contract tests.
Let each agent infer formats from sample prompts without schema guarantees.

Explanation

Correct Answer

Consistent semantics enable interoperable calculations and reliable joins across agents and systems.

Incorrect Answers

Disconnected identifiers prevent verifiable reconciliation across systems.

Currency conversion without rate and date introduces accounting errors.

Inferred format conventions can change and break downstream decisions.

Question 32

A subsidiary uses a shared multilingual HR index. A line manager requests answers about another country’s employees. Which retrieval rule must be enforced?

Filter content by the requester’s authorized geography and role before grounding, then test denial paths.
Show a geography filter in the browser while continuing privileged backend retrieval.
Grant the manager a global index key because translation needs broad access.
Allow every document through search and ask the model to translate only permitted answers.

Explanation

Correct Answer

Language translation does not override the source authorization and regional access restrictions for HR data.

Incorrect Answers

Front-end filtering is not authorization.

Shared global keys defeat scope controls and actor attribution.

Post-retrieval translation cannot remove forbidden content from model context.

Question 33

A procurement agent produces a supplier compliance finding from three policies with conflicting effective dates. What should its answer include?

Hide conflicts from the auditor and choose the most affirmative answer.
Cite the controlling approved policy version and surface conflict where authority cannot be resolved.
Use the most recently uploaded policy file as the controlling source without checking whether Legal approved or superseded it.
Quote every retrieved policy as simultaneously binding without checking supersession.

Explanation

Correct Answer

A defensible finding depends on version precedence and traceable source evidence, not retrieval frequency.

Incorrect Answers

Suppressing conflict creates unsupported compliance assertions.

Upload recency does not establish policy authority. A draft can be newer than the binding approved revision.

Superseded and active policies cannot both be assumed controlling.

Question 34

After a data steward rejects a vendor record correction, an agent keeps reapplying its inferred value nightly. What governance change should stop the loop?

Allow the model to overwrite a rejected correction until the output stabilizes.
Use fuzzy supplier-name matching to reclassify a rejected vendor correction, then replay the change as a new request.
Treat steward decisions as authoritative records, prevent unauthorized reupdates and escalate unresolved discrepancies.
Disable change logging to avoid drawing attention to repetitive updates.

Explanation

Correct Answer

Authoritative stewardship and update permissions must govern the data lifecycle independently of agent confidence.

Incorrect Answers

Repeated writes do not create valid source authority.

A semantic match does not overcome a steward rejection. Replaying an unauthorized change bypasses the data-approval workflow.

Removing logs hides the control failure.

Question 35

An HR operations agent can suggest candidate onboarding tasks, but a statutory check must occur before activating accounts. Which behavior fits?

Disable the compliance step for requests with confident agent summaries.
Use a generative response as the source of record for legal clearance.
Let the agent prepare and prioritize tasks while a deterministic account workflow blocks activation pending the verified statutory check.
Activate accounts first and have the agent explain later why it assumed the check passed.

Explanation

Correct Answer

Task assistance can speed preparation, but the authoritative gate must remain outside model-controlled prose.

Incorrect Answers

Confidence is not documented legal clearance.

Generated text is not the statutory authority record.

Retrospective explanation does not satisfy a prerequisite compliance check.

Question 36

An AI service desk pilot was evaluated only for average response time. It sometimes reveals restricted attachments in citations. What criterion must be added before rollout?

A larger throughput target that overlooks information disclosure.
A higher satisfaction threshold without testing source access control.
Explicit negative authorization tests that attempt retrieval of documents outside the requesting user’s permissions.
An instruction that the chatbot should “be careful” about citations.

Explanation

Correct Answer

Release evidence must prove confidentiality as well as performance through enforceable retrieval-denial tests.

Incorrect Answers

Speed cannot compensate for a data exposure defect.

Satisfaction cannot detect every unauthorized source leak.

Natural-language caution is not a tested security boundary.

Question 37

A retailer has 12 proposed AI initiatives but no skills inventory or risk owner. Which first Cloud Adoption Framework action creates a sustainable adoption foundation?

Choose the largest available model as the organization-wide strategy.
Start every pilot in production without approval because the cloud is elastic.
Assess skills, data readiness and prioritized use cases before buying production capacity.
Delay any review of data access until the first customer-facing release.

Explanation

Correct Answer

AI adoption is a program of business goals, people, governance and platform readiness; equipment selection follows measured demand.

Incorrect Answers

Model size cannot substitute for a ranked business case and operating model.

Production-first rollouts create security and operations debt before use cases are validated.

Data protection and policy are architecture inputs, not last-minute optional tests.

Question 38

A company can either automate a call-center claim triage with high data quality or use an agent to advise on a rarely repeated board decision. What is the strongest use-case prioritization criterion?

Prioritize the initiative with the highest estimated gross benefit before validating input data readiness or delivery risk.
Prioritize the board-advice agent because its decisions are higher-impact, without assessing review controls or repeated workload.
Choose the initiative with the highest chat-resolution sentiment score even though its financial and operational outcomes were not measured.
Expected measurable value, feasible data, implementation risk and repeatable workload volume.

Explanation

Correct Answer

Value-led ranking needs baselines, data readiness and delivery risk; frequent high-friction workflows often provide testable ROI.

Incorrect Answers

A benefit projection cannot support investment prioritization when its underlying data quality and execution risks remain unvalidated.

Potential consequence alone does not establish a feasible automation candidate; sparse decisions with uncertain approval boundaries need different risk and cost analysis.

Sentiment may indicate user acceptance, but does not establish business value or risk for the actual process being prioritized.

Question 39

An accountant requests a writing assistant inside an existing Microsoft 365 workflow, while another unit needs a service that commits ERP inventory updates. Which design distinction matters?

Assume every Microsoft 365 Copilot chat has unrestricted ERP write permissions.
Avoid any integration because all agents must work only from static documents.
Use Copilot assistance for bounded productivity tasks and a separately authorized agent/action architecture for ERP transactions.
Implement a standalone email-summary agent for inventory transfers without integrating the approved reservation transaction or its authorization checks.

Explanation

Correct Answer

A productivity assistant and a transactional agent have different permission, validation and human-approval requirements.

Incorrect Answers

Chat entitlement does not confer external database write access.

A governed tool can safely extend workflows when supported and authorized.

A summary cannot reserve stock or establish which employee is permitted to execute a warehouse transfer.

Question 40

A distributor wants one agent to recommend shipment plans and another to reserve warehouse capacity. Reservations require exact units and idempotent tool calls. How should the architect divide them?

Let both agents independently reserve the same capacity without a shared transaction identifier.
Exchange only free-text promises between agents with no schema for units.
Allow the planning agent to bypass warehouse availability during peak demand.
Use a planning agent that proposes a typed reservation and a specialized execution agent/tool that validates capacity and authorization.

Explanation

Correct Answer

Separating reasoning from controlled execution avoids duplicated writes and supports typed handoff and compensating actions.

Incorrect Answers

Concurrent independent reservations can double-book limited resources.

Unstructured instructions can confuse units and account IDs.

A proposal should not be treated as validated inventory state.

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

Microsoft AB-100 Practice Test 2 - 10 Questions (Last Reviewed 12-Nov-2026)
Microsoft AB-100 Practice Test 3 - 10 Questions (Last Reviewed 11-Nov-2026)
Microsoft AB-100 Practice Test 4 - 10 Questions (Last Reviewed 10-Nov-2026)
Microsoft AB-100 Practice Test 5 - 10 Questions (Last Reviewed 09-Nov-2026)
Microsoft AB-100 Practice Test 6 - 10 Questions (Last Reviewed 08-Nov-2026)

Pass Your Microsoft AB-100 Exam Easy!

Question 30

Explanation

Correct Answer

Incorrect Answers

Question 31

Explanation

Correct Answer

Incorrect Answers

Question 32

Explanation

Correct Answer

Incorrect Answers

Question 33

Explanation

Correct Answer

Incorrect Answers

Question 34

Explanation

Correct Answer

Incorrect Answers

Question 35

Explanation

Correct Answer

Incorrect Answers

Question 36

Explanation

Correct Answer

Incorrect Answers

Question 37

Explanation

Correct Answer

Incorrect Answers

Question 38

Explanation

Correct Answer

Incorrect Answers

Question 39

Explanation

Correct Answer

Incorrect Answers

Question 40

Explanation

Correct Answer

Incorrect Answers

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