Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 52

Question 510

The operating team must monitor whether an architectural correction works. Business context: Customer terminology changes over time. Historical symptom: Team retrains monthly despite unchanged model behavior. Which outcome measure is the best match?

  1. Deployed models with complete registry evidence
  2. Model changes with complete actor and approval records
  3. True performance drift detected before critical SLO breach
  4. Time needed to recover a known-good model/data pair

Explanation

Correct Answer

C:

For the business use where customer terminology changes over time, the result to quantify is true performance drift detected before critical SLO breach; it connects measurement to the mechanism where drift monitoring and outcomes justify retraining when performance shifts, not merely to agent traffic.

Incorrect Answers

A:

For the workflow where data scientists cannot explain why a newer classifier behaves differently, deployed models with complete registry evidence is an actionable measure; it does not establish whether team retrains monthly despite unchanged model behavior in the situation under review.

B:

The premise of this indicator is deployment audit links actor, approved artifact and release time; the architecture being evaluated instead serves customer terminology changes over time, and its success is measured by true performance drift detected before critical SLO breach.

D:

For the workflow where a preprocessing release changes category labels incorrectly, time needed to recover a known-good model/data pair is an actionable measure; it does not establish whether team retrains monthly despite unchanged model behavior in the situation under review.

Question 511

A product owner must measure whether an AI-enabled change actually improves this process: Operational records update faster than the agent's indexed copy. The earlier problem was: Shipping estimates reflect last week's warehouse data. Which outcome indicator is most relevant?

  1. Cost per accepted classification at required accuracy
  2. First-wave use cases meeting readiness gates
  3. Citations referencing in-force policy versions
  4. Ninety-fifth percentile source-to-answer freshness delay

Explanation

Correct Answer

D:

For the use case where operational records update faster than the agent's indexed copy, the indicator ninety-fifth percentile source-to-answer freshness delay is worth baselining against the earlier issue where shipping estimates reflect last week's warehouse data.

Incorrect Answers

A:

The alternative tracks cost per accepted classification at required accuracy, arising from the mechanism where task-aware routing chooses economical models after quality gates; its movement alone says little about the failure that shipping estimates reflect last week's warehouse data.

B:

For the workflow where candidate processes are attractive but permissions and data quality are weak, first-wave use cases meeting readiness gates is an actionable measure; it does not establish whether shipping estimates reflect last week's warehouse data in the situation under review.

C:

Although citations referencing in-force policy versions is observable, the affected process involves operational records update faster than the agent's indexed copy; its acceptance outcome is ninety-fifth percentile source-to-answer freshness delay, not the alternative measurement.

Question 512

A pilot scorecard needs evidence of business improvement rather than only invocation counts. The use case is: The model submits malformed identifiers to an operational API. Known defect: Invalid record IDs produce updates to the wrong business object. Which KPI should be selected?

  1. Invalid tool-call requests rejected before mutation
  2. Unauthorized tool invocation success rate
  3. Conflicting cross-application writes per thousand transactions
  4. Portion of pilots tied to measurable business objectives

Explanation

Correct Answer

A:

Because typed inputs and server validation reject malformed or out-of-scope calls, the outcome worth following is invalid tool-call requests rejected before mutation; the metric is tied to the reported issue—invalid record IDs produce updates to the wrong business object—for the scenario where the model submits malformed identifiers to an operational API.

Incorrect Answers

B:

The proposed metric concerns agent instructions mention many APIs but only two are permitted, whereas the failure in this case is that invalid record IDs produce updates to the wrong business object; the indicator needed here is invalid tool-call requests rejected before mutation.

C:

Although conflicting cross-application writes per thousand transactions is observable, the affected process involves the model submits malformed identifiers to an operational API; its acceptance outcome is invalid tool-call requests rejected before mutation, not the alternative measurement.

D:

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 invalid record IDs produce updates to the wrong business object has been corrected.

Question 513

Release reviewers will not accept a generic success message. The business scenario is that users request a status check and an authorized follow-up action, and the historical failure is: A single-topic design cannot combine both operations naturally. Which test is appropriately targeted?

  1. Create a test request and verify backend audit
  2. Replay a timeout after committing a purchase
  3. Run representative notes and validate schema and exclusions
  4. Simulate failed verification and confirm no notification
  5. Test mixed intents with missing action parameters

Explanation

Correct Answer

E:

Testing mixed intents with missing action parameters challenges the path where the planner can select supported knowledge and tools for compound intents for the business setting users request a status check and an authorized follow-up action; the specific defect to exclude is that a single-topic design cannot combine both operations naturally.

Incorrect Answers

A:

Although the exercise might catch read-only knowledge alone cannot create a service request, it would not directly validate the mechanism that the planner can select supported knowledge and tools for compound intents in the target implementation.

B:

Replaying a timeout after committing a purchase verifies an adjacent objective where network retries can repeat a single purchase request; it leaves untested the causal path in which the planner can select supported knowledge and tools for compound intents for this case.

C:

In a solution where an agent must turn service notes into a fixed JSON summary, running representative notes and validating schema and exclusions could catch the defect that a general conversational turn produces inconsistent output fields; the current release gate must challenge the failure where a single-topic design cannot combine both operations naturally.

D:

This check exercises a different system behavior: defined flow steps make sequencing and error paths explicit; a pass there cannot establish that a single-topic design cannot combine both operations naturally no longer occurs in the required workflow.

Question 514

Troubleshooting starts from a business complaint: Voice responses contain unreadable links and lengthy tables. The affected process is: The same agent is offered in voice and asynchronous chat. Which investigation is most directly tied to the reported defect?

  1. Examine schema validation logs for missing or invalid properties
  2. Compare output format across supported channels
  3. Compare summaries with mandated case-handoff sections
  4. Review plans for unauthorized leaps or tool chaining errors

Explanation

Correct Answer

B:

For the case where the same agent is offered in voice and asynchronous chat, the diagnostic path must explain the failure that voice responses contain unreadable links and lengthy tables; comparing output format across supported channels targets the mechanism by which channel-aware response design adapts wording and modality.

Incorrect Answers

A:

Free-form paragraphs break required schema parsing; examining that other path does not determine whether channel-aware response design adapts wording and modality, which is what distinguishes the issue in the stated use case.

C:

For the failure that summaries omit customer impact and current ownership, comparing summaries with mandated case-handoff sections is relevant; this question instead reports that voice responses contain unreadable links and lengthy tables, so different evidence is needed.

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 channel-aware response design adapts wording and modality and the failure that voice responses contain unreadable links and lengthy tables.

Question 515

A policy review covers an agent for which a summarizer sends output into a ticket ingestion service. Its most relevant exposure is that a malformed structured summary might create corrupt service tickets. What safeguard should be required before release?

  1. Ensure the backend rejects repeat writes for the same intent
  2. Reject unvalidated model output before ticket creation
  3. Give the agent only the required return operation
  4. Separate draft generation from external communication authority

Explanation

Correct Answer

B:

In the scenario where a summarizer sends output into a ticket ingestion service, the explicit risk is that a malformed structured summary might create corrupt service tickets; enforcement must follow the mechanism by which structured outputs define required fields and allowable values.

Incorrect Answers

A:

Network retries can repeat a single purchase request; its guardrail addresses the possibility that an API retry might create a second purchase for the same user instruction; the present architecture must constrain the different threat that a malformed structured summary might create corrupt service tickets.

C:

Where an employee requests a standard equipment return initiation, the guard can address a single-purpose return agent might invoke unrelated inventory or purchasing tools; the exposure that a malformed structured summary might create corrupt service tickets remains relevant to the present assessment.

D:

The alternate boundary mitigates a drafting tool might send an external supplier message without authority, which is relevant to a bot only needs to check warranty and prepare a draft; it leaves the current workflow vulnerable to a malformed structured summary might create corrupt service tickets.

Question 516

A product owner must measure whether an AI-enabled change actually improves this process: Service, sales and finance jointly decide high-value concessions. The earlier problem was: A suggested discount bypasses delegated approval authority. Which outcome indicator is most relevant?

  1. Cross-domain tasks routed to the correct specialist
  2. High-value concessions with complete approval trace
  3. Irreversible changes with verified approver attribution
  4. Cross-app case-state agreement rate

Explanation

Correct Answer

B:

Because cross-app workflow routes the recommendation to the accountable approver, the outcome worth following is high-value concessions with complete approval trace; the metric is tied to the reported issue—a suggested discount bypasses delegated approval authority—for the scenario where service, sales and finance jointly decide high-value concessions.

Incorrect Answers

A:

A request spans procurement, finance and policy expertise; its intended measurement is cross-domain tasks routed to the correct specialist; substituting that result here would not capture whether a suggested discount bypasses delegated approval authority has been corrected.

C:

The alternative tracks irreversible changes with verified approver attribution, arising from the mechanism where confirmation and human approval separate advice from execution authority; its movement alone says little about the failure that a suggested discount bypasses delegated approval authority.

D:

Although cross-app case-state agreement rate is observable, the affected process involves service, sales and finance jointly decide high-value concessions; its acceptance outcome is high-value concessions with complete approval trace, not the alternative measurement.

Question 517

A risk analysis highlights a threat in this workflow: Employees express several intents and need interactive clarification. Exposure: A conversational tool could invoke consequential changes from an ambiguous request. Which enforceable safeguard addresses it at the relevant boundary?

  1. Validate sending-agent identity and delegated authorization
  2. Narrow allowed tool capabilities and verify server-side authorization
  3. Require human review for high-impact low-confidence classifications
  4. Restrict available actions and require confirmation for consequential changes

Explanation

Correct Answer

D:

Avoiding the consequence where a conversational tool could invoke consequential changes from an ambiguous request requires a control for employees express several intents and need interactive clarification; in this architecture, generative orchestration interprets intent and can combine authorized tools and knowledge, which is the point at which the boundary can be applied.

Incorrect Answers

A:

This restriction is associated with independently hosted agents must exchange tasks and results and addresses the exposure that an external agent might accept delegated tasks from a spoofed sender; it does not specifically protect the target workflow against a conversational tool could invoke consequential changes from an ambiguous request.

B:

The alternate guard addresses a shared MCP server might honor an unauthorized tool invocation for a Copilot Studio agent needs a reusable interface to governed operations; the present threat, a conversational tool could invoke consequential changes from an ambiguous request, needs a different enforcement boundary.

C:

That protection is tied to the mechanism that representative examples teach decision boundaries without changing underlying facts; it does not prevent the distinct risk where a conversational tool could invoke consequential changes from an ambiguous request in the system under review.

Question 518

Troubleshooting starts from a business complaint: Successful chat sessions conceal repeated backend writes. The affected process is: Agents modify records through external APIs. Which investigation is most directly tied to the reported defect?

  1. Compare denial logs and executed actions for attack samples
  2. Inspect backend denials and attempted tool arguments
  3. Compare request IDs to backend transaction counts
  4. Compare errors by relevant justified cohorts with privacy care

Explanation

Correct Answer

C:

Where successful chat sessions conceal repeated backend writes, the useful evidence comes from comparing request IDs to backend transaction counts; its focus matches the architectural mechanism that tool metrics reveal mutation failures and retry consequences in the process where agents modify records through external APIs.

Incorrect Answers

A:

For the failure that a malicious instruction triggers a privileged tool, comparing denial logs and executed actions for attack samples is relevant; this question instead reports that successful chat sessions conceal repeated backend writes, so different evidence is needed.

B:

The model may try a tool outside the user's rights; examining that other path does not determine whether tool metrics reveal mutation failures and retry consequences, which is what distinguishes the issue in the stated use case.

D:

Overall accuracy conceals disproportionate error; examining that other path does not determine whether tool metrics reveal mutation failures and retry consequences, which is what distinguishes the issue in the stated use case.

Question 519

An enterprise team is making a platform design decision. Business context: A stable specialized classification style persists despite prompt examples. Present failure: Few-shot prompts require excessive tokens and yield inconsistent formats. Which approach resolves the mismatch?

  1. Use examples and counterexamples to clarify edge cases
  2. Give prompts explicit role, task, constraints and response schema
  3. Fine-tune only when repeated task behavior warrants adaptation
  4. Select a small task-specific model for constrained workloads

Explanation

Correct Answer

C:

A stable specialized classification style persists despite prompt examples; its failing behavior is that few-shot prompts require excessive tokens and yield inconsistent formats; applying the underlying design principle—fine-tuning can adapt recurring behavior using vetted examples—connects the requirement to an implementable control.

Incorrect Answers

A:

There is a valid place for the mechanism that representative examples teach decision boundaries without changing underlying facts, namely the classifier confuses complaints with refund requests; it does not remove the specific defect that few-shot prompts require excessive tokens and yield inconsistent formats from this case.

B:

When agents return inconsistent fields for downstream integrations, clear instructions and structured outputs reduce avoidable ambiguity; in the stated workflow, however, few-shot prompts require excessive tokens and yield inconsistent formats; the relevant architectural behavior is that fine-tuning can adapt recurring behavior using vetted examples.

D:

For another type of architecture, smaller models can be evaluated for narrow tasks with lower cost is useful; this design must instead support a stable specialized classification style persists despite prompt examples despite the problem where few-shot prompts require excessive tokens and yield inconsistent formats.

Question 520

In a business situation where employees install external agents with unknown data pathways, a previous build showed this defect: An unreviewed plugin exports sensitive tenant information. Which reproducible test belongs in the release gate?

  1. Try accessing a restricted training corpus from an unrelated project
  2. Pilot a connector with synthetic data and review destinations
  3. Submit a malicious mislabeled training record and inspect review
  4. Test synthetic flows and compare actual destinations

Explanation

Correct Answer

B:

For the required workflow where employees install external agents with unknown data pathways, the test needs to reproduce or prevent the failure where an unreviewed plugin exports sensitive tenant information; piloting a connector with synthetic data and reviewing destinations exercises the path where procurement and security review evaluate permissions and data flows.

Incorrect Answers

A:

The suggested test serves proprietary records are copied to an open test workspace, in which training examples become accessible outside intended teams; the present scenario instead concerns the reported problem where an unreviewed plugin exports sensitive tenant information.

C:

In a solution where business analysts can submit examples but not change production behavior, submitting a malicious mislabeled training record and inspecting review could catch the defect that unvetted labeled data poisons a retrained classifier; the current release gate must challenge the failure where an unreviewed plugin exports sensitive tenant information.

D:

For a third-party AI tool changes where submitted documents are processed, the alternate test is about whether contract and privacy assessment cover service location and subprocessor use; it cannot prove the outcome needed after the reported regression where an unreviewed plugin exports sensitive tenant information.

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