Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 94

Question 930

An incident review has one confirmed symptom: Copilot attaches one customer's risk note to another account. For the use case where two systems share company names but not primary keys, which targeted diagnostic check would reveal the most pertinent evidence?

  1. Review confusion pairs against the customer-service glossary
  2. Compare predicted severity against signed service-impact rules
  3. Trace each suggested promise to the customer's entitlement and policy
  4. Inspect match decisions and false-positive entity joins

Explanation

Correct Answer

D:

For the case where two systems share company names but not primary keys, the diagnostic path must explain the failure that Copilot attaches one customer's risk note to another account; inspecting match decisions and false-positive entity joins targets the mechanism by which a mapping service resolves stable identities before enrichment.

Incorrect Answers

A:

The proposed inspection is associated with agents must distinguish an incident, complaint, inquiry and entitlement, where business-term definitions align user vocabulary with governed service entities; it would not expose the failure mode that Copilot attaches one customer's risk note to another account with the required specificity.

B:

The proposed probe concerns a specialized support team uses a nonstandard severity matrix, where custom business rules align AI-assisted triage with actual impact definitions; this incident instead requires evidence about a mapping service resolves stable identities before enrichment and the failure that Copilot attaches one customer's risk note to another account.

C:

Replies cite generic guidance instead of contract-specific policy; examining that other path does not determine whether a mapping service resolves stable identities before enrichment, which is what distinguishes the issue in the stated use case.

Question 931

Before production promotion, the team needs evidence for this setting: Sales copilots should not expose sensitive margin assumptions. A prior regression caused the following: A connector returns the full ERP payload including protected fields. Which validation is most persuasive?

  1. Change a test case status and inspect audit events
  2. Query as multiple roles and confirm only authorized fields appear
  3. Try the action with a user lacking record access
  4. Try status queries and mutations before and after verification

Explanation

Correct Answer

B:

Querying as multiple roles and confirming only authorized fields appear is aligned with the failure that a connector returns the full ERP payload including protected fields; its causal relevance is that field projection and endpoint authorization minimize exposed data for the situation where sales copilots should not expose sensitive margin assumptions.

Incorrect Answers

A:

In a solution where an agent needs to update a governed ticket status, changing a test case status and inspecting audit events could catch the defect that its knowledge source can explain but cannot commit a database change; the current release gate must challenge the failure where a connector returns the full ERP payload including protected fields.

C:

The tested behavior is that server-side authorization and minimized inputs protect data flow; the necessary evidence here concerns field projection and endpoint authorization minimize exposed data and the previously observed a connector returns the full ERP payload including protected fields.

D:

This check exercises a different system behavior: authentication and authorization gates separate general help from account control; a pass there cannot establish that a connector returns the full ERP payload including protected fields no longer occurs in the required workflow.

Question 932

The acceptance board asks how to verify a corrected design. The setting is: Unused agents retain tokens and active integrations. Last time, former owners leave behind accessible service accounts. Which scenario should the team execute?

  1. Test intended roles against each production-bound action
  2. Attempt an unapproved production modification in test governance
  3. Simulate rollback during a nonproduction close cycle
  4. Retire a staged agent and verify permissions are removed

Explanation

Correct Answer

D:

Retiring a staged agent and verifying permissions are removed is aligned with the failure that former owners leave behind accessible service accounts; its causal relevance is that decommissioning revokes credentials, connectors and publishing rights for the situation where unused agents retain tokens and active integrations.

Incorrect Answers

A:

For an imported agent can authenticate in dev but not production, the alternate test is about whether production consent, identity and connection setup are distinct release checks; it cannot prove the outcome needed after the reported regression where former owners leave behind accessible service accounts.

B:

The verification concerning production components should not be edited casually checks managed solution deployment supports controlled downstream customization boundaries against a maker changes production agent logic outside the release process; its result says little about the observed issue where former owners leave behind accessible service accounts.

C:

The suggested test serves invoice posting agents affect critical close processes, in which a prompt change during close generates unexplained exceptions; the present scenario instead concerns the reported problem where former owners leave behind accessible service accounts.

Question 933

The acceptance board asks how to verify a corrected design. The setting is: Two systems share company names but not primary keys. Last time, Copilot attaches one customer's risk note to another account. Which scenario should the team execute?

  1. Test duplicate legal names with different external IDs
  2. Feed outages of different scope and inspect suggested severity
  3. Query synonyms and near-neighbor phrases and check intended classification
  4. Draft replies for differing contract tiers and inspect accuracy

Explanation

Correct Answer

A:

The historical failure is that Copilot attaches one customer's risk note to another account; for the setting where two systems share company names but not primary keys, testing duplicate legal names with different external IDs verifies the dependency where a mapping service resolves stable identities before enrichment.

Incorrect Answers

B:

In a solution where a specialized support team uses a nonstandard severity matrix, feeding outages of different scope and inspecting suggested severity could catch the defect that default summaries treat critical outages as low priority; the current release gate must challenge the failure where Copilot attaches one customer's risk note to another account.

C:

For agents must distinguish an incident, complaint, inquiry and entitlement, the alternate test is about whether business-term definitions align user vocabulary with governed service entities; it cannot prove the outcome needed after the reported regression where Copilot attaches one customer's risk note to another account.

D:

The suggested test serves service representatives draft responses from authorized knowledge, in which replies cite generic guidance instead of contract-specific policy; the present scenario instead concerns the reported problem where Copilot attaches one customer's risk note to another account.

Question 934

An agent rollout is intended for this situation: Production components should not be edited casually. The known problem is that a maker changes production agent logic outside the release process. Which success measure connects to that specific outcome?

  1. Outbound regulated messages with required approvals
  2. Unauthorized production customization events
  3. Unauthorized cross-environment requests blocked
  4. Environment promotions without configuration drift

Explanation

Correct Answer

B:

The observed shortcoming is that a maker changes production agent logic outside the release process; measuring unauthorized production customization events gives business owners a directly relevant indicator for the process where production components should not be edited casually.

Incorrect Answers

A:

Although outbound regulated messages with required approvals is observable, the affected process involves production components should not be edited casually; its acceptance outcome is unauthorized production customization events, not the alternative measurement.

C:

Test automation must not reach live customer records; its intended measurement is unauthorized cross-environment requests blocked; substituting that result here would not capture whether a maker changes production agent logic outside the release process has been corrected.

D:

The premise of this indicator is solution-based ALM transports supported components under controlled releases; the architecture being evaluated instead serves production components should not be edited casually, and its success is measured by unauthorized production customization events.

Question 935

Business leaders reject an activity-only metric for this process: Required knowledge stores are intermittently unreachable. They need to detect the issue where the agent fabricates an answer during a search outage. Which performance indicator fits?

  1. Wrong-entity retrieval rate in curated evaluation cases
  2. Accuracy-adjusted cost per completed business task
  3. Unsupported answer rate during retrieval failures
  4. Successful prompt-injection override rate

Explanation

Correct Answer

C:

For the business use where required knowledge stores are intermittently unreachable, the result to quantify is unsupported answer rate during retrieval failures; it connects measurement to the mechanism where health checks and explicit no-evidence responses handle missing context safely, not merely to agent traffic.

Incorrect Answers

A:

For the workflow where multiple systems use inconsistent customer identifiers and duplicate cases, wrong-entity retrieval rate in curated evaluation cases is an actionable measure; it does not establish whether the agent fabricates an answer during a search outage in the situation under review.

B:

For the situation where real task mix differs from prelaunch performance tests, accuracy-adjusted cost per completed business task is a sensible indicator; the problem that the agent fabricates an answer during a search outage requires measurement of unsupported answer rate during retrieval failures.

D:

That scorecard concerns the activity in which third-party documents include sentences telling the model to ignore rules and its result of successful prompt-injection override rate; the reported shortcoming here is the agent fabricates an answer during a search outage.

Question 936

The rollout supports this workflow: Several agents need the same inventory query operation. Leaders want a metric that responds to the reported shortcoming that each agent builds a different unaudited HTTP wrapper. Which measure best reflects success?

  1. Complex exception tasks resolved with supported reasoning traces
  2. Tool operations executed with valid delegated identity
  3. Channel-appropriate responses passing usability review
  4. Prompt action responses conforming to output contract
  5. Role-specific tasks completed in supported Copilot surfaces

Explanation

Correct Answer

B:

The use case involves several agents need the same inventory query operation and the baseline failure that each agent builds a different unaudited HTTP wrapper; the KPI tracks tool operations executed with valid delegated identity, so improvement has an interpretable operational meaning.

Incorrect Answers

A:

Procurement exceptions require comparing several policies and records; its intended measurement is complex exception tasks resolved with supported reasoning traces; substituting that result here would not capture whether each agent builds a different unaudited HTTP wrapper has been corrected.

C:

Although channel-appropriate responses passing usability review is observable, the affected process involves several agents need the same inventory query operation; its acceptance outcome is tool operations executed with valid delegated identity, not the alternative measurement.

D:

Prompt action responses conforming to output contract; that indicator describes a distinct workflow where an agent must turn service notes into a fixed JSON summary; it does not quantify tool operations executed with valid delegated identity, which is needed for the stated case.

E:

The premise of this indicator is declarative agents define instructions, knowledge and actions in the supported host; the architecture being evaluated instead serves several agents need the same inventory query operation, and its success is measured by tool operations executed with valid delegated identity.

Question 937

Operations has not established the cause of this defect: Approved data crosses an unreviewed processor. The associated setting is: A third-party AI tool changes where submitted documents are processed. Which inspection should precede remediation?

  1. Match release manifest to transformation and model checksums
  2. Review data processing agreements and network endpoints
  3. Correlate conversation request IDs with business API audit entries
  4. Diff tool traces and evaluation outcomes against baseline

Explanation

Correct Answer

B:

For the case where a third-party AI tool changes where submitted documents are processed, the diagnostic path must explain the failure that approved data crosses an unreviewed processor; reviewing data processing agreements and network endpoints targets the mechanism by which contract and privacy assessment cover service location and subprocessor use.

Incorrect Answers

A:

For the failure that restoring code alone does not restore previous predictions, matching release manifest to transformation and model checksums is relevant; this question instead reports that approved data crosses an unreviewed processor, so different evidence is needed.

C:

The proposed inspection is associated with a tool deletes a record following an ambiguous conversation, where end-to-end audit binds user, tool, affected record and authorization event; it would not expose the failure mode that approved data crosses an unreviewed processor with the required specificity.

D:

The evidence this method produces applies to a workflow where a hosted agent gains an additional research tool; it cannot by itself establish why approved data crosses an unreviewed processor in the current scenario.

Question 938

A proposed deployment supports this business case: A business process mixes deterministic calculations and open-ended judgments. Reviewers also warn that probabilistic generation might replace authoritative arithmetic used in financial controls. Which restriction mitigates that concrete risk?

  1. Deny unapproved tools regardless of model-generated arguments
  2. Keep critical numeric calculations in tested services
  3. Require policy-owner approval for any rule change
  4. Restrict write APIs to the authoritative owning service

Explanation

Correct Answer

B:

In the context where a business process mixes deterministic calculations and open-ended judgments, the exposure is that probabilistic generation might replace authoritative arithmetic used in financial controls; containing it requires a guardrail in the operation where task decomposition separates deterministic services from bounded agent reasoning.

Incorrect Answers

A:

The proposed control targets planner-generated arguments might invoke tools that the user was never permitted to call, while the given risk scenario is probabilistic generation might replace authoritative arithmetic used in financial controls; these exposures arise in different operational contexts.

C:

The alternate boundary mitigates an unapproved change to the approval matrix could silently alter authorization decisions, which is relevant to the task follows fixed approval branches with no interpretation needed; it leaves the current workflow vulnerable to probabilistic generation might replace authoritative arithmetic used in financial controls.

D:

The alternate boundary mitigates two apps might overwrite each other's supposedly authoritative customer values, which is relevant to two applications try to overwrite the same customer credit field; it leaves the current workflow vulnerable to probabilistic generation might replace authoritative arithmetic used in financial controls.

Question 939

An investment review asks what measurable result would justify continuing an AI solution where a regulated workload must keep data in permitted locations. The original issue is: An external model call exports sensitive prompts to an unapproved region. Which indicator should be tracked?

  1. Sensitive telemetry records retained beyond policy limit
  2. Model releases with reproducible data lineage
  3. Time needed to recover a known-good model/data pair
  4. Sensitive requests crossing unauthorized residency boundaries

Explanation

Correct Answer

D:

Sensitive requests crossing unauthorized residency boundaries; this outcome measure belongs to the workflow where a regulated workload must keep data in permitted locations; a change in it can help detect whether the defect that an external model call exports sensitive prompts to an unapproved region is resolved.

Incorrect Answers

A:

Debugging logs include prompts containing customer data; its intended measurement is sensitive telemetry records retained beyond policy limit; substituting that result here would not capture whether an external model call exports sensitive prompts to an unapproved region has been corrected.

B:

The proposed metric concerns model retraining consumes periodically revised business records, whereas the failure in this case is that an external model call exports sensitive prompts to an unapproved region; the indicator needed here is sensitive requests crossing unauthorized residency boundaries.

C:

Time needed to recover a known-good model/data pair; that indicator describes a distinct workflow where a preprocessing release changes category labels incorrectly; it does not quantify sensitive requests crossing unauthorized residency boundaries, which is needed for the stated case.

Question 940

Governance approves the business purpose but requires control of this exposure: A low-frequency severe harm might persist despite high average satisfaction. The agent will be used when a helpful assistant can still produce unsafe financial or legal guidance. Which boundary is most defensible?

  1. Block publication without accountable owner and classification
  2. Escalate security findings outside the ordinary product backlog
  3. Disable high-risk capabilities until residual risks are accepted
  4. Require heightened review for autonomous privileged agents

Explanation

Correct Answer

C:

In the context where a helpful assistant can still produce unsafe financial or legal guidance, the exposure is that a low-frequency severe harm might persist despite high average satisfaction; containing it requires a guardrail in the operation where risk-based red-team evaluations identify foreseeable harm.

Incorrect Answers

A:

This restriction is associated with business units create agents with no central visibility and addresses the exposure that production agents might have no recorded accountable owner or risk classification; it does not specifically protect the target workflow against a low-frequency severe harm might persist despite high average satisfaction.

B:

The safeguard for product owners favor easy fixes over high-consequence failures deals with a different threat: a security issue might remain in the normal backlog without incident escalation; even after it is enforced, the exposure that a low-frequency severe harm might persist despite high average satisfaction remains possible in the target workflow where risk-based red-team evaluations identify foreseeable harm.

D:

Where personal read-only assistants and payment agents have different risk, the guard can address a privileged autonomous agent might receive the same review as a private helper; the exposure that a low-frequency severe harm might persist despite high average satisfaction remains relevant to the present assessment.

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