Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 114

Question 1130

A staging test has to detect the specific defect that delayed reconciliation hides material fulfillment blockers. In the workflow where invoice holds must trigger exception handling without polling, which exercise verifies the corrective behavior?

  1. Demonstrate a custom multi-step workflow and failure recovery
  2. Run contract tests against both current and previous schema versions
  3. Attempt low-risk and high-risk actions and inspect approval enforcement
  4. Emit duplicate and delayed events and verify idempotent processing

Explanation

Correct Answer

D:

Emiting duplicate and delayed events and verifying idempotent processing is aligned with the failure that delayed reconciliation hides material fulfillment blockers; its causal relevance is that business events and durable orchestration propagate changes asynchronously for the situation where invoice holds must trigger exception handling without polling.

Incorrect Answers

A:

The verification concerning the solution demands its own model selection and hosting policies checks custom-engine architecture controls runtime, model calls and integrations against host-managed orchestration cannot satisfy a unique execution plan; its result says little about the observed issue where delayed reconciliation hides material fulfillment blockers.

B:

In a solution where multiple AI teams consume a changing source API, running contract tests against both current and previous schema versions could catch the defect that an upstream field rename silently breaks downstream grounding; the current release gate must challenge the failure where delayed reconciliation hides material fulfillment blockers.

C:

For teams propose that an agent perform both low-risk lookups and financial postings, the alternate test is about whether tiered permissions require progressively stronger approvals for riskier actions; it cannot prove the outcome needed after the reported regression where delayed reconciliation hides material fulfillment blockers.

Question 1131

One enterprise rollout has 2 separately testable requirements. (1) Business use: Retraining uses revised labels and a different source snapshot. Known defect: Model behavior changes without reproducible training provenance. (2) Business use: A candidate appears accurate on its training examples. Known defect: Overfitting hides failures on new business records. Which 2 architectural choices directly satisfy these stated cases? Choose TWO.

  1. Version models and datasets together for controlled adaptation
  2. Choose speech processing for transcribed contact-center audio
  3. Use independent holdout data to select a tuned model
  4. Create a curated evaluation set for retrieval quality

Explanation

Correct Answers

A:

For the independent requirement that retraining uses revised labels and a different source snapshot, dataset lineage and model versioning support repeatable assessment; it addresses the defect model behavior changes without reproducible training provenance, while a candidate appears accurate on its training examples needs its own architecture.

C:

For the independent requirement that a candidate appears accurate on its training examples, separation of training and evaluation yields realistic generalization checks; it addresses the defect overfitting hides failures on new business records, while retraining uses revised labels and a different source snapshot needs its own architecture.

Incorrect Answers

B:

This design is associated with the different failure where manual notes omit important customer statements in a workflow where quality teams need searchable insights from recorded calls; it does not directly solve the listed conditions: model behavior changes without reproducible training provenance; overfitting hides failures on new business records.

D:

This design is associated with the different failure where generator tuning fails because the required documents were never retrieved in a workflow where stakeholders report hallucinations but cannot isolate search errors; it does not directly solve the listed conditions: model behavior changes without reproducible training provenance; overfitting hides failures on new business records.

Question 1132

The operating team must monitor whether an architectural correction works. Business context: Agent handles threshold-based return decisions. Historical symptom: Human testers overlook values just above approval limits. Which outcome measure is the best match?

  1. Security contract conditions represented in vetted tests
  2. Requirement boundaries exercised by accepted test cases
  3. Multi-turn tasks resolved without entity confusion
  4. Quality-adjusted cost per approved business result

Explanation

Correct Answer

B:

The use case involves agent handles threshold-based return decisions and the baseline failure that human testers overlook values just above approval limits; the KPI tracks requirement boundaries exercised by accepted test cases, so improvement has an interpretable operational meaning.

Incorrect Answers

A:

For the situation where agents can call APIs with customer IDs and amount values, security contract conditions represented in vetted tests is a sensible indicator; the problem that human testers overlook values just above approval limits requires measurement of requirement boundaries exercised by accepted test cases.

C:

The premise of this indicator is multi-turn tests assess state tracking, entity separation and follow-up questions; the architecture being evaluated instead serves agent handles threshold-based return decisions, and its success is measured by requirement boundaries exercised by accepted test cases.

D:

The premise of this indicator is stratified evaluation guides thresholds for difficult requests; the architecture being evaluated instead serves agent handles threshold-based return decisions, and its success is measured by requirement boundaries exercised by accepted test cases.

Question 1133

The architecture board has evidence that Copilot attaches one customer's risk note to another account. The affected situation is: Two systems share company names but not primary keys. What change most directly establishes the missing behavior?

  1. Normalize account identifiers across CRM and external records
  2. Publish an approved business glossary for customer-service concepts
  3. Customize service triage around business-specific severity criteria
  4. Restrict Sales connector fields to approved data classifications

Explanation

Correct Answer

A:

Two systems share company names but not primary keys; its failing behavior is that Copilot attaches one customer's risk note to another account; applying the underlying design principle—a mapping service resolves stable identities before enrichment—connects the requirement to an implementable control.

Incorrect Answers

B:

In the case where agents must distinguish an incident, complaint, inquiry and entitlement, the alternate design depends on business-term definitions align user vocabulary with governed service entities; it does not correct the target failure, Copilot attaches one customer's risk note to another account.

C:

When a specialized support team uses a nonstandard severity matrix, custom business rules align AI-assisted triage with actual impact definitions; in the stated workflow, however, Copilot attaches one customer's risk note to another account; the relevant architectural behavior is that a mapping service resolves stable identities before enrichment.

D:

In the case where sales copilots should not expose sensitive margin assumptions, the alternate design depends on field projection and endpoint authorization minimize exposed data; it does not correct the target failure, Copilot attaches one customer's risk note to another account.

Question 1134

The fault affects an AI-powered workflow in which a business process mixes deterministic calculations and open-ended judgments. Specifically, a single agent unpredictably handles policy rules and narrative summarization. Which diagnostic investigation best distinguishes the relevant failure path?

  1. Identify workflow steps with fixed rules versus variable language
  2. Trace customer identifiers and change propagation between apps
  3. Inspect write conflicts and source-of-truth definitions
  4. Trace triggering events and idempotency keys for repeated actions
  5. Review available tool inventory versus approved actions

Explanation

Correct Answer

A:

A single agent unpredictably handles policy rules and narrative summarization; to attribute the defect rather than guess at a fix, identifying workflow steps with fixed rules versus variable language examines the dependency where task decomposition separates deterministic services from bounded agent reasoning.

Incorrect Answers

B:

An opportunity owner sees a stale service escalation status; examining that other path does not determine whether task decomposition separates deterministic services from bounded agent reasoning, which is what distinguishes the issue in the stated use case.

C:

The proposed probe concerns two applications try to overwrite the same customer credit field, where field ownership and integration contracts prevent competing writes; this incident instead requires evidence about task decomposition separates deterministic services from bounded agent reasoning and the failure that a single agent unpredictably handles policy rules and narrative summarization.

D:

For the failure that a background agent has triggered irreversible duplicate actions, tracing triggering events and idempotency keys for repeated actions is relevant; this question instead reports that a single agent unpredictably handles policy rules and narrative summarization, so different evidence is needed.

E:

For the failure that the agent can invoke an unintended high-privilege operation, reviewing available tool inventory versus approved actions is relevant; this question instead reports that a single agent unpredictably handles policy rules and narrative summarization, so different evidence is needed.

Question 1135

The operating team must monitor whether an architectural correction works. Business context: An autonomous agent relies on slow approvals and delayed tool results. Historical symptom: Timeout creates an abandoned workflow with no owner. Which outcome measure is the best match?

  1. Orphaned business workflows after simulated timeouts
  2. Unauthorized retrievals completed in red-team cases
  3. Verified task resolution rate
  4. Release candidates satisfying critical evaluation thresholds

Explanation

Correct Answer

A:

The use case involves an autonomous agent relies on slow approvals and delayed tool results and the baseline failure that timeout creates an abandoned workflow with no owner; the KPI tracks orphaned business workflows after simulated timeouts, so improvement has an interpretable operational meaning.

Incorrect Answers

B:

For the workflow where a user asks the agent to read an unrelated executive record, unauthorized retrievals completed in red-team cases is an actionable measure; it does not establish whether timeout creates an abandoned workflow with no owner in the situation under review.

C:

The premise of this indicator is outcome measures connect agent use with completed business work; the architecture being evaluated instead serves an autonomous agent relies on slow approvals and delayed tool results, and its success is measured by orphaned business workflows after simulated timeouts.

D:

The premise of this indicator is evaluation suites validate quality and safety before publication; the architecture being evaluated instead serves an autonomous agent relies on slow approvals and delayed tool results, and its success is measured by orphaned business workflows after simulated timeouts.

Question 1136

Risk reviewers need a control, not merely a monitoring dashboard. The use case is: Finance users need current operational data in contextual help. Potential harm: Live finance records might become editable or visible to unauthorized user roles. Which protection fits?

  1. Do not auto-save an uncertain prediction as authoritative data
  2. Remove unneeded attributes before model invocation
  3. Apply tenant policies and scoped content permissions
  4. Prevent writes or disclosure outside user permissions

Explanation

Correct Answer

D:

In the scenario where finance users need current operational data in contextual help, the explicit risk is that live finance records might become editable or visible to unauthorized user roles; enforcement must follow the mechanism by which authorized virtual entities can surface current finance and operations data to supported agents.

Incorrect Answers

A:

That protection is tied to the mechanism that clear review interfaces surface provenance and editable proposals; it does not prevent the distinct risk where live finance records might become editable or visible to unauthorized user roles in the system under review.

B:

The alternate boundary mitigates a canvas app might send unnecessary protected fields into an inference prompt, which is relevant to a canvas app passes customer records to a model action; it leaves the current workflow vulnerable to live finance records might become editable or visible to unauthorized user roles.

C:

The safeguard for employees want curated assistance without owning a separate LLM runtime deals with a different threat: a declarative agent might be published beyond permitted tenant groups; even after it is enforced, the exposure that live finance records might become editable or visible to unauthorized user roles remains possible in the target workflow where authorized virtual entities can surface current finance and operations data to supported agents.

Question 1137

The operating team must monitor whether an architectural correction works. Business context: Quality teams need searchable insights from recorded calls. Historical symptom: Manual notes omit important customer statements. Which outcome measure is the best match?

  1. Correct tool selection and input-binding rate
  2. Correctly transcribed critical utterances
  3. Meaning preservation rate on multilingual service samples
  4. Successful inferences within latency SLO

Explanation

Correct Answer

B:

For the use case where quality teams need searchable insights from recorded calls, the indicator correctly transcribed critical utterances is worth baselining against the earlier issue where manual notes omit important customer statements.

Incorrect Answers

A:

A change in correct tool selection and input-binding rate is informative for the orchestrator chooses between similar invoice and credit tools; it does not establish whether the current process achieves correctly transcribed critical utterances.

C:

That scorecard concerns the activity in which support documents arrive in several customer languages and its result of meaning preservation rate on multilingual service samples; the reported shortcoming here is manual notes omit important customer statements.

D:

The proposed metric concerns a specialized model must serve production apps reliably, whereas the failure in this case is that manual notes omit important customer statements; the indicator needed here is correctly transcribed critical utterances.

Question 1138

Governance approves the business purpose but requires control of this exposure: Partial business commits might remain inconsistent after a later action fails. The agent will be used when an agent creates an order but reservation of inventory fails. Which boundary is most defensible?

  1. Require confirmation when intent remains ambiguous
  2. Require reversible actions or human intervention for partial commits
  3. Prevent writes or disclosure outside user permissions
  4. Do not permit unreviewed AI predictions during outages

Explanation

Correct Answer

B:

In the context where an agent creates an order but reservation of inventory fails, the exposure is that partial business commits might remain inconsistent after a later action fails; containing it requires a guardrail in the operation where compensating operations reconcile earlier successful steps.

Incorrect Answers

A:

Where two topics claim nearly identical scheduling intents, the guard can address overlapping scheduling intents might invoke a change before confirmation; the exposure that partial business commits might remain inconsistent after a later action fails remains relevant to the present assessment.

C:

The alternate boundary mitigates live finance records might become editable or visible to unauthorized user roles, which is relevant to finance users need current operational data in contextual help; it leaves the current workflow vulnerable to partial business commits might remain inconsistent after a later action fails.

D:

Where field staff must keep working during a model outage, the guard can address an AI outage might lead to unchecked predictions in core app records; the exposure that partial business commits might remain inconsistent after a later action fails remains relevant to the present assessment.

Question 1139

A planned AI solution must support this situation: Model retraining consumes periodically revised business records. Its prototype fails because no one can identify which dataset produced a model. Which design best fits the constraint?

  1. Version training datasets with lineage and access records
  2. Promote grounding datasets through quality gates
  3. Stage candidate models behind champion-challenger evaluation
  4. Version agent instructions, tool definitions and model references together

Explanation

Correct Answer

A:

For the need that model retraining consumes periodically revised business records, the relevant mechanism is that immutable snapshots and lineage tie runs to approved sources; that architectural dependency addresses the observed failure where no one can identify which dataset produced a model.

Incorrect Answers

B:

Ingestion gates check classification, freshness and ownership; its main purpose is new documents enter a production source automatically, not correction of the reported defect where no one can identify which dataset produced a model in the target process.

C:

In the case where a tuned model claims higher accuracy but adds latency, the alternate design depends on candidate comparison evaluates quality, latency and safety under matched load; it does not correct the target failure, no one can identify which dataset produced a model.

D:

A complete agent release manifest pins related runtime artifacts; its main purpose is model and tool changes are promoted independently, not correction of the reported defect where no one can identify which dataset produced a model in the target process.

Question 1140

A staging test has to detect the specific defect that it autonomously contacts a supplier without approval. In the workflow where a bot only needs to check warranty and prepare a draft, which exercise verifies the corrective behavior?

  1. Change a test case status and inspect audit events
  2. Generate stressful and adversarial user examples
  3. Ask unrelated questions and verify no supplier action is triggered
  4. Ask for a variance explanation with incomplete data

Explanation

Correct Answer

C:

The historical failure is that it autonomously contacts a supplier without approval; for the setting where a bot only needs to check warranty and prepare a draft, asking unrelated questions and verifying no supplier action is triggered verifies the dependency where limited intent and tool scope prevent unnecessary actions.

Incorrect Answers

A:

Although the exercise might catch its knowledge source can explain but cannot commit a database change, it would not directly validate the mechanism that limited intent and tool scope prevent unnecessary actions in the target implementation.

B:

For compliance letters must preserve approved language and disclaimers, the alternate test is about whether templating and approved clauses limit generated variation; it cannot prove the outcome needed after the reported regression where it autonomously contacts a supplier without approval.

D:

The tested behavior is that grounded explanatory responses distinguish evidence from hypothesis; the necessary evidence here concerns limited intent and tool scope prevent unnecessary actions and the previously observed it autonomously contacts a supplier without approval.

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