Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 40

Question 390

The pilot needs a durable correction: Retrieval configuration accidentally grants write capability. This affects the workflow in which an agent should summarize account data but not alter it. Which design choice should the architect approve?

  1. Separate read-only grounding from transactional tools
  2. Test tool-permission escalation attempts at the execution layer
  3. Require independent backend authorization for generated tool calls
  4. Test posting permissions and approvals after agent updates

Explanation

Correct Answer

A:

For the need that an agent should summarize account data but not alter it, the relevant mechanism is that distinct permission surfaces enforce least privilege by task; that architectural dependency addresses the observed failure where retrieval configuration accidentally grants write capability.

Incorrect Answers

B:

Authorization evaluations verify server-side deny behavior; its main purpose is a user asks the agent to read an unrelated executive record, not correction of the reported defect where retrieval configuration accidentally grants write capability in the target process.

C:

For another type of architecture, server-side access checks enforce real actor entitlements is useful; this design must instead support an agent should summarize account data but not alter it despite the problem where retrieval configuration accidentally grants write capability.

D:

There is a valid place for the mechanism that authorization regression gates verify posting authority remains constrained, namely agents suggest payment adjustments and inventory holds; it does not remove the specific defect that retrieval configuration accidentally grants write capability from this case.

Question 391

The pilot needs a durable correction: Summaries omit customer impact and current ownership. This affects the workflow in which agents need structured case summaries at handoff. Which design choice should the architect approve?

  1. Add approved case-summary guidance to the service workflow
  2. Maintain context when transferring voice to human support
  3. Normalize account identifiers across CRM and external records
  4. Customize service triage around business-specific severity criteria

Explanation

Correct Answer

A:

The defect is that summaries omit customer impact and current ownership; the expected business case is that agents need structured case summaries at handoff; the design fits because tailored instructions and contextual fields standardize case handoff content.

Incorrect Answers

B:

For a voice bot cannot resolve a complex eligibility dispute, this other approach relies on the fact that structured handoff passes allowed intent, transcript summary and case context; the issue here is summaries omit customer impact and current ownership; that mechanism does not establish the required tailored instructions and contextual fields standardize case handoff content.

C:

The technical behavior of this alternative is that a mapping service resolves stable identities before enrichment; the user case being assessed is agents need structured case summaries at handoff with a different failure: summaries omit customer impact and current ownership.

D:

For another type of architecture, custom business rules align AI-assisted triage with actual impact definitions is useful; this design must instead support agents need structured case summaries at handoff despite the problem where summaries omit customer impact and current ownership.

Question 392

A quality engineer is writing a regression case after the incident where slightly different wording breaks correct entity selection. The underlying business scenario is: A rewritten prompt improves a single demonstration. What should the test exercise?

  1. Submit urgent and ambiguous cases to the candidate agent
  2. Degrade sample images and test extraction stability
  3. Re-run the same multi-intent corpus on both configurations
  4. Run semantically similar requests with differing critical constraints

Explanation

Correct Answer

D:

For the required workflow where a rewritten prompt improves a single demonstration, the test needs to reproduce or prevent the failure where slightly different wording breaks correct entity selection; running semantically similar requests with differing critical constraints exercises the path where regression sets cover lexical variation and task boundaries.

Incorrect Answers

A:

The verification concerning a support agent changes how it classifies priority intents checks controlled service evaluations validate routing before rollout against high-severity tickets begin routing to general queues; its result says little about the observed issue where slightly different wording breaks correct entity selection.

B:

In a solution where production documents differ from clean training samples, degrading sample images and testing extraction stability could catch the defect that scanned forms suddenly produce inaccurate predictions; the current release gate must challenge the failure where slightly different wording breaks correct entity selection.

C:

The tested behavior is that trace comparisons reveal shifts in plan selection and step order; the necessary evidence here concerns regression sets cover lexical variation and task boundaries and the previously observed slightly different wording breaks correct entity selection.

Question 393

In a business situation where agents and apps need shared definitions of customers and work orders, a previous build showed this defect: Different solutions apply incompatible status and customer fields. Which reproducible test belongs in the release gate?

  1. Ask for a case field and verify the exact authorized record
  2. Query the same entity from two authorized clients and reconcile attributes
  3. Re-run a saved golden set across prompt versions
  4. Emit duplicate and delayed events and verify idempotent processing

Explanation

Correct Answer

B:

The historical failure is that different solutions apply incompatible status and customer fields; for the setting where agents and apps need shared definitions of customers and work orders, querying the same entity from two authorized clients and reconciling attributes verifies the dependency where Dataverse tables and governed relationships provide a reusable business-data surface.

Incorrect Answers

A:

For users ask about governed CRM records rather than prose manuals, the alternate test is about whether Dataverse grounding exposes appropriate structured records to the agent; it cannot prove the outcome needed after the reported regression where different solutions apply incompatible status and customer fields.

C:

The suggested test serves many teams copy slightly different instructions for the same task, in which a successful prompt change cannot be reproduced after rollout; the present scenario instead concerns the reported problem where different solutions apply incompatible status and customer fields.

D:

In a solution where invoice holds must trigger exception handling without polling, emiting duplicate and delayed events and verifying idempotent processing could catch the defect that delayed reconciliation hides material fulfillment blockers; the current release gate must challenge the failure where different solutions apply incompatible status and customer fields.

Question 394

The pilot needs a durable correction: It substitutes a prior customer's order in a later turn. This affects the workflow in which agent must remember order context while asking for missing details. Which design choice should the architect approve?

  1. Use production feedback to propose targeted regression tests
  2. Test multi-turn context retention and clarification
  3. Test concurrent agent runs for duplicate side effects
  4. Automate repeatable evaluation gates before publishing

Explanation

Correct Answer

B:

Agent must remember order context while asking for missing details; its failing behavior is that it substitutes a prior customer's order in a later turn; applying the underlying design principle—multi-turn tests assess state tracking, entity separation and follow-up questions—connects the requirement to an implementable control.

Incorrect Answers

A:

For support tickets reveal an unusual multistep conversation failure, this other approach relies on the fact that feedback-derived examples convert incidents into lasting safeguards; the issue here is it substitutes a prior customer's order in a later turn; that mechanism does not establish the required multi-turn tests assess state tracking, entity separation and follow-up questions.

C:

Parallel agents each submit a costly replacement; the alternate design can address that situation because concurrency evaluation tests idempotency and serialization, whereas this question needs a solution to it substitutes a prior customer's order in a later turn.

D:

In the case where prompt revisions cause intermittent regressions, the alternate design depends on stable automated suites compare response quality across versions; it does not correct the target failure, it substitutes a prior customer's order in a later turn.

Question 395

A design adds autonomous behavior to a process where a conversational change affects only one support geography. This creates a documented risk: A regional pilot might produce a severe local regression hidden by global averages. Which enforced control is the best fit?

  1. Limit production edit roles and monitor drift
  2. Roll back or withhold publication if acceptance checks fail
  3. Disable the candidate for a cohort on critical regression
  4. Block promotion when authorization or data-leak tests fail

Explanation

Correct Answer

C:

The business setting is a conversational change affects only one support geography, with the exposure that a regional pilot might produce a severe local regression hidden by global averages; the enforceable protection must correspond to the path where cohort telemetry supports safe staged validation.

Incorrect Answers

A:

This restriction is associated with production components should not be edited casually and addresses the exposure that direct production customization might diverge from the approved solution; it does not specifically protect the target workflow against a regional pilot might produce a severe local regression hidden by global averages.

B:

That protection is tied to the mechanism that post-import validation includes publishing and channel behavior; it does not prevent the distinct risk where a regional pilot might produce a severe local regression hidden by global averages in the system under review.

D:

The proposed control targets a new Foundry workflow might fail safety evaluation but still be published, while the given risk scenario is a regional pilot might produce a severe local regression hidden by global averages; these exposures arise in different operational contexts.

Question 396

The fault affects an AI-powered workflow in which several events refer to one business order at once. Specifically, parallel agents each submit a costly replacement. Which diagnostic investigation best distinguishes the relevant failure path?

  1. Correlate request rates with throttles and queue depth
  2. Compare operation keys with final business transaction counts
  3. Reconcile session outcomes to actual completed records
  4. Review evaluator results by task segment and release
  5. Inspect test distribution against business consequence tiers

Explanation

Correct Answer

B:

Parallel agents each submit a costly replacement; to attribute the defect rather than guess at a fix, comparing operation keys with final business transaction counts examines the dependency where concurrency evaluation tests idempotency and serialization.

Incorrect Answers

A:

That probe evaluates capacity metrics and throttling errors show saturation risk in a workflow where a regional launch multiplies concurrent agent requests; the reported failure concerns the path for which concurrency evaluation tests idempotency and serialization instead.

C:

A service bot handles many chats but few cases are resolved; its matching diagnostic check is reasonable for a dashboard calls the bot successful because messages increased; the present investigation needs evidence for the different failure where parallel agents each submit a costly replacement.

D:

Reviewing evaluator results by task segment and release explores a separate connection, where stable automated suites compare response quality across versions; the diagnosed requirement is several events refer to one business order at once, and the actual defect is parallel agents each submit a costly replacement.

E:

Inspecting test distribution against business consequence tiers explores a separate connection, where risk-weighting prioritizes high-impact actions and failure paths; the diagnosed requirement is several events refer to one business order at once, and the actual defect is parallel agents each submit a costly replacement.

Question 397

Risk reviewers need a control, not merely a monitoring dashboard. The use case is: Agents begin citing an obsolete article after an ingestion job. Potential harm: A disputed answer might be untraceable to its source and index revision. Which protection fits?

  1. Retain audit records of source replacements and withdrawals
  2. Reject cached results that omit original security context
  3. Do not share private cached outputs across principals
  4. Require reviewed datasets and independent promotion sign-off

Explanation

Correct Answer

A:

In the scenario where agents begin citing an obsolete article after an ingestion job, the explicit risk is that a disputed answer might be untraceable to its source and index revision; enforcement must follow the mechanism by which content lineage records update, transformation and index revision.

Incorrect Answers

B:

For the use case where users have different rights to the same knowledge corpus, the control addresses the threat that shared answer caches might drop source ACLs and leak a private document; the specified agent must be protected against a disputed answer might be untraceable to its source and index revision.

C:

That protection is tied to the mechanism that cache keys and access validation prevent security-context leakage; it does not prevent the distinct risk where a disputed answer might be untraceable to its source and index revision in the system under review.

D:

The alternate guard addresses unreviewed training examples might poison a promoted model for business analysts can submit examples but not change production behavior; the present threat, a disputed answer might be untraceable to its source and index revision, needs a different enforcement boundary.

Question 398

A release must prove that regression appears only for compound user requests will not recur. The application is used in this situation: A new topic description changes which agent is invoked. Which acceptance test most directly checks that failure mode?

  1. Re-run the same multi-intent corpus on both configurations
  2. Rotate away from a test secret and verify identity-based access
  3. Re-evaluate extraction accuracy at forecast traffic levels
  4. Run semantically similar requests with differing critical constraints

Explanation

Correct Answer

A:

For the required workflow where a new topic description changes which agent is invoked, the test needs to reproduce or prevent the failure where regression appears only for compound user requests; re-runing the same multi-intent corpus on both configurations exercises the path where trace comparisons reveal shifts in plan selection and step order.

Incorrect Answers

B:

For a backend agent must call an approved Azure service without embedded secrets, the alternate test is about whether managed identity provides credential-less workload authentication where supported; it cannot prove the outcome needed after the reported regression where regression appears only for compound user requests.

C:

In a solution where a new model reduces price but also degrades exact extractions, re-evaluating extraction accuracy at forecast traffic levels could catch the defect that finance sees savings without the downstream rework cost; the current release gate must challenge the failure where regression appears only for compound user requests.

D:

The suggested test serves a rewritten prompt improves a single demonstration, in which slightly different wording breaks correct entity selection; the present scenario instead concerns the reported problem where regression appears only for compound user requests.

Question 399

After deployment, stakeholders need an interpretable measure for this use case: A Teams user should query and update a supported external service. The pilot showed that read-only knowledge alone cannot create a service request. Which metric captures the needed result?

  1. Successful authorized external service actions
  2. External knowledge retrievals respecting access permissions
  3. Fresh authorized account insights from external systems
  4. Current finance record answers with correct access trimming

Explanation

Correct Answer

A:

For the use case where a Teams user should query and update a supported external service, the indicator successful authorized external service actions is worth baselining against the earlier issue where read-only knowledge alone cannot create a service request.

Incorrect Answers

B:

Although external knowledge retrievals respecting access permissions is observable, the affected process involves a Teams user should query and update a supported external service; its acceptance outcome is successful authorized external service actions, not the alternative measurement.

C:

The proposed metric concerns sellers need current credit exposure held outside CRM, whereas the failure in this case is that read-only knowledge alone cannot create a service request; the indicator needed here is successful authorized external service actions.

D:

The proposed metric concerns finance users need current operational data in contextual help, whereas the failure in this case is that read-only knowledge alone cannot create a service request; the indicator needed here is successful authorized external service actions.

Question 400

A risk analysis highlights a threat in this workflow: An internal helper produces attractive but unsupported policy answers. Exposure: An agent might invent a policy answer when no authoritative evidence exists. Which enforceable safeguard addresses it at the relevant boundary?

  1. Escalate questions where no authoritative source exists
  2. Mask sensitive input/output fields and control log access
  3. Fail closed for unsupported sensitive answers
  4. Limit retroactive alteration of approved QA evidence

Explanation

Correct Answer

A:

In the context where an internal helper produces attractive but unsupported policy answers, the exposure is that an agent might invent a policy answer when no authoritative evidence exists; containing it requires a guardrail in the operation where expert-checked evaluation examples quantify evidence fidelity.

Incorrect Answers

B:

This restriction is associated with a newly deployed Foundry agent has no detailed diagnostics and addresses the exposure that tracing might store secrets and personal content without adequate access controls; it does not specifically protect the target workflow against an agent might invent a policy answer when no authoritative evidence exists.

C:

For the use case where an assistant responds fluently but cites unrelated support articles, the control addresses the threat that missing retrieval evidence might still generate definitive sensitive advice; the specified agent must be protected against an agent might invent a policy answer when no authoritative evidence exists.

D:

The alternate boundary mitigates old evaluation evidence might be overwritten to conceal a failed release gate, which is relevant to an updated agent performs differently months after release; it leaves the current workflow vulnerable to an agent might invent a policy answer when no authoritative evidence exists.

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