Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 68

Question 670

A policy review covers an agent for which the requirement is common and product coverage is adequate. Its most relevant exposure is that a managed service might fail a required data-processing or exit condition. What safeguard should be required before release?

  1. Confirm tenant policy and data access before enabling adoption
  2. Review tenancy, data terms, exit and support commitments
  3. Disable autonomous regulated decisions under low evidence
  4. Narrow allowed tool capabilities and verify server-side authorization

Explanation

Correct Answer

B:

The requirement is common and product coverage is adequate; its risk scenario is that a managed service might fail a required data-processing or exit condition; the safeguard fits because managed functionality shifts routine platform operations to the service and the restriction addresses that specific exposure.

Incorrect Answers

A:

Users need ordinary meeting summaries and follow-up drafts; its guardrail addresses the possibility that a standard agent might violate tenant access policy when activated for many users; the present architecture must constrain the different threat that a managed service might fail a required data-processing or exit condition.

C:

This restriction is associated with the system cannot always find evidence for regulated guidance and addresses the exposure that a low-evidence agent might autonomously decide a regulated outcome; it does not specifically protect the target workflow against a managed service might fail a required data-processing or exit condition.

D:

The safeguard for a Copilot Studio agent needs a reusable interface to governed operations deals with a different threat: a shared MCP server might honor an unauthorized tool invocation; even after it is enforced, the exposure that a managed service might fail a required data-processing or exit condition remains possible in the target workflow where managed functionality shifts routine platform operations to the service.

Question 671

A business stakeholder describes this operating situation: Sales users say deal stage differently across regions. Testing instead shows that an agent updates the wrong pipeline field from an informal phrase. What design approach is most appropriate?

  1. Map sales-stage language to actual opportunity fields
  2. Normalize account identifiers across CRM and external records
  3. Index governed external content using Microsoft 365 Copilot connectors
  4. Define approved product and service aliases for customer inquiries
  5. Choose language analytics for focused text classification
  6. Restrict Sales connector fields to approved data classifications

Explanation

Correct Answer

A:

Because canonical business terminology maps language to underlying Dynamics records, this approach corrects the reported breakdown where an agent updates the wrong pipeline field from an informal phrase; the affected situation is that sales users say deal stage differently across regions.

Incorrect Answers

B:

There is a valid place for the mechanism that a mapping service resolves stable identities before enrichment, namely two systems share company names but not primary keys; it does not remove the specific defect that an agent updates the wrong pipeline field from an informal phrase from this case.

C:

There is a valid place for the mechanism that supported connectors expose external content with discoverability and security metadata, namely workers need internal wiki articles from outside Microsoft 365; it does not remove the specific defect that an agent updates the wrong pipeline field from an informal phrase from this case.

D:

For customer-facing brands have several names for the same product, this other approach relies on the fact that curated alias-to-product mapping grounds interpretations in the product catalogue; the issue here is an agent updates the wrong pipeline field from an informal phrase; that mechanism does not establish the required canonical business terminology maps language to underlying Dynamics records.

E:

In the case where the system must detect customer sentiment and named entities at scale, the alternate design depends on purpose-built text analysis handles bounded language-analysis tasks; it does not correct the target failure, an agent updates the wrong pipeline field from an informal phrase.

F:

This design would make sense if sales copilots should not expose sensitive margin assumptions, particularly where a connector returns the full ERP payload including protected fields; the requested architecture instead needs the dependency that canonical business terminology maps language to underlying Dynamics records because an agent updates the wrong pipeline field from an informal phrase.

Question 672

A product owner must measure whether an AI-enabled change actually improves this process: Candidate processes are attractive but permissions and data quality are weak. The earlier problem was: The first-wave pilot requests broad access to unreliable records. Which outcome indicator is most relevant?

  1. High-value concessions with complete approval trace
  2. Portion of pilots tied to measurable business objectives
  3. First-wave use cases meeting readiness gates
  4. Cost per correct extraction under the response-time SLO

Explanation

Correct Answer

C:

This scorecard needs an outcome related to candidate processes are attractive but permissions and data quality are weak; first-wave use cases meeting readiness gates. That indicator can be checked against the earlier symptom that the first-wave pilot requests broad access to unreliable records.

Incorrect Answers

A:

The premise of this indicator is cross-app workflow routes the recommendation to the accountable approver; the architecture being evaluated instead serves candidate processes are attractive but permissions and data quality are weak, and its success is measured by first-wave use cases meeting readiness gates.

B:

Executives request AI before defining an operational improvement target; its intended measurement is portion of pilots tied to measurable business objectives; substituting that result here would not capture whether the first-wave pilot requests broad access to unreliable records has been corrected.

D:

For the workflow where high-volume narrow extraction needs predictable latency and cost, cost per correct extraction under the response-time SLO is an actionable measure; it does not establish whether the first-wave pilot requests broad access to unreliable records in the situation under review.

Question 673

A quality engineer is writing a regression case after the incident where retrieval configuration accidentally grants write capability. The underlying business scenario is: An agent should summarize account data but not alter it. What should the test exercise?

  1. Ask for a privileged record using an unprivileged test identity
  2. Attempt inference from an unauthorized identity or network
  3. Attempt a write from a read-only assistant
  4. Make identical queries as permitted and denied users

Explanation

Correct Answer

C:

The check is targeted rather than generic: distinct permission surfaces enforce least privilege by task; its pass/fail evidence applies to the workflow where an agent should summarize account data but not alter it and to the prior defect that retrieval configuration accidentally grants write capability.

Incorrect Answers

A:

This check exercises a different system behavior: authorization evaluations verify server-side deny behavior; a pass there cannot establish that retrieval configuration accidentally grants write capability no longer occurs in the required workflow.

B:

The suggested test serves a sensitive model endpoint is reachable using a broadly shared key, in which untrusted clients submit requests and exfiltrate outputs; the present scenario instead concerns the reported problem where retrieval configuration accidentally grants write capability.

D:

The suggested test serves shared result caching makes answers fast across roles, in which one user's cached confidential answer appears for another user; the present scenario instead concerns the reported problem where retrieval configuration accidentally grants write capability.

Question 674

Business leaders reject an activity-only metric for this process: A search index changes embedding model and chunking approach. They need to detect the issue where existing vectors are queried with incompatible assumptions. Which performance indicator fits?

  1. Retrieval regressions following index promotions
  2. Ingested documents passing quality and access gates
  3. Contract-valid predictions after custom-model updates
  4. Recovery time for a failed connector-action deployment

Explanation

Correct Answer

A:

The use case involves a search index changes embedding model and chunking approach and the baseline failure that existing vectors are queried with incompatible assumptions; the KPI tracks retrieval regressions following index promotions, so improvement has an interpretable operational meaning.

Incorrect Answers

B:

Ingested documents passing quality and access gates; that indicator describes a distinct workflow where new documents enter a production source automatically; it does not quantify retrieval regressions following index promotions, which is needed for the stated case.

C:

Although contract-valid predictions after custom-model updates is observable, the affected process involves a search index changes embedding model and chunking approach; its acceptance outcome is retrieval regressions following index promotions, not the alternative measurement.

D:

The premise of this indicator is coordinated versioning preserves consumer/provider compatibility; the architecture being evaluated instead serves a search index changes embedding model and chunking approach, and its success is measured by retrieval regressions following index promotions.

Question 675

Troubleshooting starts from a business complaint: Staff mistake speculative recommendations for recorded facts. The affected process is: Users need to distinguish verified fields from AI suggestions. Which investigation is most directly tied to the reported defect?

  1. Review plans for unauthorized leaps or tool chaining errors
  2. Inspect field-level source and suggestion labels
  3. Verify virtual entity enablement and row-level access
  4. Analyze containment failures and repeated intent fallback patterns
  5. Inspect missing-field collection and final action payload

Explanation

Correct Answer

B:

The investigation directly tests the causal connection: visual provenance separates authoritative records from generated advice; it collects evidence for a workflow where users need to distinguish verified fields from AI suggestions but the outcome is instead that staff mistake speculative recommendations for recorded facts.

Incorrect Answers

A:

In the process where procurement exceptions require comparing several policies and records, the diagnostic procedure could explain a simple scripted topic cannot handle legitimate variants; this particular incident instead turns on whether visual provenance separates authoritative records from generated advice and why staff mistake speculative recommendations for recorded facts.

C:

A static knowledge upload gives stale purchase-order states; examining that other path does not determine whether visual provenance separates authoritative records from generated advice, which is what distinguishes the issue in the stated use case.

D:

Analyzing containment failures and repeated intent fallback patterns explores a separate connection, where confidence, retry ceilings and escalation routing enable resolution; the diagnosed requirement is users need to distinguish verified fields from AI suggestions, and the actual defect is staff mistake speculative recommendations for recorded facts.

E:

An intake task requires several prerequisite fields; its matching diagnostic check is reasonable for agent sends incomplete requests when users answer out of order; the present investigation needs evidence for the different failure where staff mistake speculative recommendations for recorded facts.

Question 676

Governance approves the business purpose but requires control of this exposure: An API might trust model-generated identity claims instead of valid authorization. The agent will be used when prompt wording persuades the agent to submit a privileged operation. Which boundary is most defensible?

  1. Deny external sends unless independently approved
  2. Restore the previous prompt if critical intent selection regresses
  3. Reject missing, invalid or out-of-scope authorization tokens
  4. Do not allow model confidence prose to override missing evidence

Explanation

Correct Answer

C:

In the context where prompt wording persuades the agent to submit a privileged operation, the exposure is that an API might trust model-generated identity claims instead of valid authorization; containing it requires a guardrail in the operation where server-side access checks enforce real actor entitlements.

Incorrect Answers

A:

A prompt injection requests data export to an outside address; its guardrail addresses the possibility that an injected tool instruction might export customer records externally; the present architecture must constrain the different threat that an API might trust model-generated identity claims instead of valid authorization.

B:

For the use case where a rewritten prompt improves a single demonstration, the control addresses the threat that a prompt revision might break critical intent handling without a rollback; the specified agent must be protected against an API might trust model-generated identity claims instead of valid authorization.

D:

Policy answers must be tied to a current source; its guardrail addresses the possibility that confident model language might override an explicit evidence requirement; the present architecture must constrain the different threat that an API might trust model-generated identity claims instead of valid authorization.

Question 677

The architect wants evidence before altering the deployment. In a workflow where a specialized model must serve production apps reliably, model quality is strong but endpoint latency breaks the workflow. What should the team examine first?

  1. Inspect outbound payloads and role-based access decisions
  2. Trace dependency failures and app completion paths
  3. Record recall at K and citation-grounding errors separately
  4. Inspect latency percentiles, errors and deployment health

Explanation

Correct Answer

D:

The reported fault is that model quality is strong but endpoint latency breaks the workflow in a system where a specialized model must serve production apps reliably; evidence from inspecting latency percentiles, errors and deployment health can show whether a managed serving endpoint needs scaling, monitoring and rollback design.

Incorrect Answers

A:

For the failure that hidden fields from the UI appear in the prompt payload, inspecting outbound payloads and role-based access decisions is relevant; this question instead reports that model quality is strong but endpoint latency breaks the workflow, so different evidence is needed.

B:

That probe evaluates timeouts and graceful degradation protect core app operations in a workflow where field staff must keep working during a model outage; the reported failure concerns the path for which a managed serving endpoint needs scaling, monitoring and rollback design instead.

C:

The proposed probe concerns stakeholders report hallucinations but cannot isolate search errors, where labeled queries and expected evidence distinguish retrieval from synthesis; this incident instead requires evidence about a managed serving endpoint needs scaling, monitoring and rollback design and the failure that model quality is strong but endpoint latency breaks the workflow.

Question 678

An incident review has one confirmed symptom: A background process posts expenses beyond its delegated scope. For the use case where agent can research options but cannot approve a vendor payment, which targeted diagnostic check would reveal the most pertinent evidence?

  1. Audit approval tokens attached to payment attempts
  2. Trace action authentication and source mutation evidence
  3. Compare competing topic traces for representative phrasing
  4. Review plans for unauthorized leaps or tool chaining errors

Explanation

Correct Answer

A:

A background process posts expenses beyond its delegated scope; to attribute the defect rather than guess at a fix, auditing approval tokens attached to payment attempts examines the dependency where approval checkpoints split independent analysis from high-impact execution.

Incorrect Answers

B:

Its knowledge source can explain but cannot commit a database change; examining that other path does not determine whether approval checkpoints split independent analysis from high-impact execution, which is what distinguishes the issue in the stated use case.

C:

In the separate scenario where two topics claim nearly identical scheduling intents, the same investigation could explain agent invokes the wrong booking action for a reschedule; it does not isolate the reported cause in a system where agent can research options but cannot approve a vendor payment.

D:

For the failure that a simple scripted topic cannot handle legitimate variants, reviewing plans for unauthorized leaps or tool chaining errors is relevant; this question instead reports that a background process posts expenses beyond its delegated scope, so different evidence is needed.

Question 679

Troubleshooting starts from a business complaint: Teams learn about problems only when someone asks the agent. The affected process is: Shipment exceptions should initiate investigation without a chat message. Which investigation is most directly tied to the reported defect?

  1. Trace statements to referenced financial data
  2. Verify event subscriptions and processing logs
  3. Trace page data bindings and invoked record APIs
  4. Review plans for unauthorized leaps or tool chaining errors

Explanation

Correct Answer

B:

Teams learn about problems only when someone asks the agent; to attribute the defect rather than guess at a fix, verifying event subscriptions and processing logs examines the dependency where an event trigger starts the bounded agent workflow on state change.

Incorrect Answers

A:

The evidence this method produces applies to a workflow where finance asks why a variance was flagged; it cannot by itself establish why teams learn about problems only when someone asks the agent in the current scenario.

C:

In the process where staff need a contextual workspace assembled around one customer case, the diagnostic procedure could explain generic chat loses visibility into related records and actions; this particular incident instead turns on whether an event trigger starts the bounded agent workflow on state change and why teams learn about problems only when someone asks the agent.

D:

The proposed probe concerns procurement exceptions require comparing several policies and records, where reasoning can plan through evidence and tools within enforced boundaries; this incident instead requires evidence about an event trigger starts the bounded agent workflow on state change and the failure that teams learn about problems only when someone asks the agent.

Question 680

A planned AI solution must support this situation: Users use novel requests outside the supported topic inventory. Its prototype fails because the agent repeats an unhelpful default message. Which design best fits the constraint?

  1. Use a prompt action for bounded text transformation
  2. Use explicit topics for a tightly scripted compliance dialogue
  3. Design a fallback topic that escalates genuine nonmatches
  4. Use an agent flow for repeatable ordered business steps

Explanation

Correct Answer

C:

Users use novel requests outside the supported topic inventory; its failing behavior is that the agent repeats an unhelpful default message; applying the underlying design principle—fallback handles unmatched intent and can ask for clarification or escalate—connects the requirement to an implementable control.

Incorrect Answers

A:

For another type of architecture, prompt actions encapsulate reusable inputs and expected outputs is useful; this design must instead support users use novel requests outside the supported topic inventory despite the problem where the agent repeats an unhelpful default message.

B:

Deterministic topic logic controls known prompts and branching; its main purpose is a regulated process requires fixed questions in exact sequence, not correction of the reported defect where the agent repeats an unhelpful default message in the target process.

D:

Defined flow steps make sequencing and error paths explicit; its main purpose is approval processing requires sequential verification and notification, not correction of the reported defect where the agent repeats an unhelpful default message in the target process.

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