Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 45

Question 440

Security and business owners want to keep this capability: Agents must distinguish an incident, complaint, inquiry and entitlement. They must also prevent the following: Unreviewed synonyms could redefine legally significant customer-service classifications. Which guardrail should the design incorporate?

  1. Block unclassified high-risk fields at the connector boundary
  2. Enforce source-system entitlements in the integration
  3. Assign glossary approval to service process owners
  4. Prevent automatic approvals when the credit source is unavailable

Explanation

Correct Answer

C:

In the context where agents must distinguish an incident, complaint, inquiry and entitlement, the exposure is that unreviewed synonyms could redefine legally significant customer-service classifications; containing it requires a guardrail in the operation where business-term definitions align user vocabulary with governed service entities.

Incorrect Answers

A:

The alternate boundary mitigates an ERP connector might expose protected margin columns in its response payload, which is relevant to sales copilots should not expose sensitive margin assumptions; it leaves the current workflow vulnerable to unreviewed synonyms could redefine legally significant customer-service classifications.

B:

Sellers need current credit exposure held outside CRM; its guardrail addresses the possibility that sales enrichment might query external records beyond the seller's entitled scope; the present architecture must constrain the different threat that unreviewed synonyms could redefine legally significant customer-service classifications.

D:

The safeguard for external credit service responds slowly during regional outages deals with a different threat: credit-dependent approvals might proceed while the live credit source is unavailable; even after it is enforced, the exposure that unreviewed synonyms could redefine legally significant customer-service classifications remains possible in the target workflow where business-term definitions align user vocabulary with governed service entities.

Question 441

An enterprise team is making a platform design decision. Business context: External knowledge contains embedded role-change commands. Present failure: The agent obeys text retrieved from a supplier document. Which approach resolves the mismatch?

  1. Use grounded answer evaluation to separate retrieval and generation faults
  2. Inspect tool-selection spans when orchestration chooses wrong actions
  3. Use Prompt Shields or equivalent intervention-point screening
  4. Apply document-level access trimming during retrieval
  5. Treat retrieved documents as untrusted evidence rather than instructions

Explanation

Correct Answer

E:

External knowledge contains embedded role-change commands; its failing behavior is that the agent obeys text retrieved from a supplier document; applying the underlying design principle—trust hierarchy separates lower-trust content from agent instructions—connects the requirement to an implementable control.

Incorrect Answers

A:

When an answer contradicts the handbook despite high model fluency, separate relevance and groundedness tests isolate the failing layer; in the stated workflow, however, the agent obeys text retrieved from a supplier document; the relevant architectural behavior is that trust hierarchy separates lower-trust content from agent instructions.

B:

For another type of architecture, step traces expose planning, selected tools and arguments is useful; this design must instead support external knowledge contains embedded role-change commands despite the problem where the agent obeys text retrieved from a supplier document.

C:

For hostile instructions arrive both as user input and retrieved documents, this other approach relies on the fact that input and document-attack detection supplies layered protection; the issue here is the agent obeys text retrieved from a supplier document; that mechanism does not establish the required trust hierarchy separates lower-trust content from agent instructions.

D:

For users have different rights to the same knowledge corpus, this other approach relies on the fact that retrieval applies effective source entitlements to candidate documents; the issue here is the agent obeys text retrieved from a supplier document; that mechanism does not establish the required trust hierarchy separates lower-trust content from agent instructions.

Question 442

An agent rollout is intended for this situation: Third-party documents include sentences telling the model to ignore rules. The known problem is that answers adopt an instruction embedded in a source document. Which success measure connects to that specific outcome?

  1. Unsupported answer rate during retrieval failures
  2. Correct classification rate for ambiguous boundary cases
  3. Successful prompt-injection override rate
  4. Differentiated requirements satisfied without unsupported workarounds

Explanation

Correct Answer

C:

This scorecard needs an outcome related to third-party documents include sentences telling the model to ignore rules; successful prompt-injection override rate. That indicator can be checked against the earlier symptom that answers adopt an instruction embedded in a source document.

Incorrect Answers

A:

The proposed metric concerns required knowledge stores are intermittently unreachable, whereas the failure in this case is that answers adopt an instruction embedded in a source document; the indicator needed here is successful prompt-injection override rate.

B:

The classifier confuses complaints with refund requests; its intended measurement is correct classification rate for ambiguous boundary cases; substituting that result here would not capture whether answers adopt an instruction embedded in a source document has been corrected.

D:

For the situation where competitive requirements depend on unique multi-model control, differentiated requirements satisfied without unsupported workarounds is a sensible indicator; the problem that answers adopt an instruction embedded in a source document requires measurement of successful prompt-injection override rate.

Question 443

A proposed deployment supports this business case: A new topic description changes which agent is invoked. Reviewers also warn that a new orchestration release might suddenly include unexpected privileged tool paths. Which restriction mitigates that concrete risk?

  1. Scope the identity to only required resource permissions
  2. Roll back if privileged actions appear in unexpected paths
  3. Restore the previous prompt if critical intent selection regresses
  4. Refuse cost optimization that breaches business-quality thresholds

Explanation

Correct Answer

B:

When a new orchestration release might suddenly include unexpected privileged tool paths, the use case a new topic description changes which agent is invoked requires a guardrail aligned with the behavior that trace comparisons reveal shifts in plan selection and step order.

Incorrect Answers

A:

Where a backend agent must call an approved Azure service without embedded secrets, the guard can address long-lived credentials might grant an agent unnecessary Azure resource privileges; the exposure that a new orchestration release might suddenly include unexpected privileged tool paths remains relevant to the present assessment.

C:

That protection is tied to the mechanism that regression sets cover lexical variation and task boundaries; it does not prevent the distinct risk where a new orchestration release might suddenly include unexpected privileged tool paths in the system under review.

D:

This restriction is associated with a new model reduces price but also degrades exact extractions and addresses the exposure that lower-cost model routing might undermine critical extraction accuracy; it does not specifically protect the target workflow against a new orchestration release might suddenly include unexpected privileged tool paths.

Question 444

Before production promotion, the team needs evidence for this setting: Hostile instructions arrive both as user input and retrieved documents. A prior regression caused the following: The agent is exposed to repeated jailbreak patterns. Which validation is most persuasive?

  1. Test known attack patterns at user and tool-response boundaries
  2. Rebuild a staged agent from recorded versions and run smoke tests
  3. Run semantically similar requests with differing critical constraints
  4. Test extraction and secret-disclosure prompts with synthetic fixtures

Explanation

Correct Answer

A:

Testing known attack patterns at user and tool-response boundaries challenges the path where input and document-attack detection supplies layered protection for the business setting hostile instructions arrive both as user input and retrieved documents; the specific defect to exclude is that the agent is exposed to repeated jailbreak patterns.

Incorrect Answers

B:

Although the exercise might catch a release uses stale prompts against a new tool signature, it would not directly validate the mechanism that input and document-attack detection supplies layered protection in the target implementation.

C:

In a solution where a rewritten prompt improves a single demonstration, running semantically similar requests with differing critical constraints could catch the defect that slightly different wording breaks correct entity selection; the current release gate must challenge the failure where the agent is exposed to repeated jailbreak patterns.

D:

Although the exercise might catch a client can request arbitrary portions of training examples, it would not directly validate the mechanism that input and document-attack detection supplies layered protection in the target implementation.

Question 445

A release must prove that default summaries treat critical outages as low priority will not recur. The application is used in this situation: A specialized support team uses a nonstandard severity matrix. Which acceptance test most directly checks that failure mode?

  1. Test duplicate legal names with different external IDs
  2. Delay the external API and verify useful partial handling
  3. Pilot each intended host rather than assuming parity
  4. Feed outages of different scope and inspect suggested severity

Explanation

Correct Answer

D:

Default summaries treat critical outages as low priority; a release test must establish that this no longer occurs; feeding outages of different scope and inspecting suggested severity measures the specific behavior for which custom business rules align AI-assisted triage with actual impact definitions.

Incorrect Answers

A:

In the alternate setting where two systems share company names but not primary keys, the test exercises a mapping service resolves stable identities before enrichment; even a pass would not show that default summaries treat critical outages as low priority has been prevented.

B:

The verification concerning external credit service responds slowly during regional outages checks bounded timeouts and graceful fallback preserve essential user work against a sales workflow hangs awaiting optional enrichment; its result says little about the observed issue where default summaries treat critical outages as low priority.

C:

The tested behavior is that host-specific support, packaging and admin approval shape deployment; the necessary evidence here concerns custom business rules align AI-assisted triage with actual impact definitions and the previously observed default summaries treat critical outages as low priority.

Question 446

The pilot needs a durable correction: Delayed reconciliation hides material fulfillment blockers. This affects the workflow in which invoice holds must trigger exception handling without polling. Which design choice should the architect approve?

  1. Publish discoverable data contracts with schema versions
  2. Use Dataverse as the governed handoff for customer cases
  3. Use event-driven integration for finance-to-supply-chain exceptions
  4. Create a custom-engine agent for bespoke model orchestration

Explanation

Correct Answer

C:

For the need that invoice holds must trigger exception handling without polling, the relevant mechanism is that business events and durable orchestration propagate changes asynchronously; that architectural dependency addresses the observed failure where delayed reconciliation hides material fulfillment blockers.

Incorrect Answers

A:

Versioned schemas and compatibility policy make interfaces predictable; its main purpose is multiple AI teams consume a changing source API, not correction of the reported defect where delayed reconciliation hides material fulfillment blockers in the target process.

B:

An opportunity owner sees a stale service escalation status; the alternate design can address that situation because shared authorized records support coordinated business context, whereas this question needs a solution to delayed reconciliation hides material fulfillment blockers.

D:

This design would make sense if the solution demands its own model selection and hosting policies, particularly where host-managed orchestration cannot satisfy a unique execution plan; the requested architecture instead needs the dependency that business events and durable orchestration propagate changes asynchronously because delayed reconciliation hides material fulfillment blockers.

Question 447

The fault affects an AI-powered workflow in which agents return inconsistent fields for downstream integrations. Specifically, narrative answers cannot be parsed by consuming workflows. Which diagnostic investigation best distinguishes the relevant failure path?

  1. Audit whether failures concern style rather than missing current facts
  2. Inspect approvals against the approved decision table
  3. Inspect parse failures against missing schema instructions
  4. Inspect entity mappings and security context at retrieval

Explanation

Correct Answer

C:

Narrative answers cannot be parsed by consuming workflows; to attribute the defect rather than guess at a fix, inspecting parse failures against missing schema instructions examines the dependency where clear instructions and structured outputs reduce avoidable ambiguity.

Incorrect Answers

A:

Auditing whether failures concern style rather than missing current facts explores a separate connection, where fine-tuning can adapt recurring behavior using vetted examples; the diagnosed requirement is agents return inconsistent fields for downstream integrations, and the actual defect is narrative answers cannot be parsed by consuming workflows.

B:

In the separate scenario where the task follows fixed approval branches with no interpretation needed, the same investigation could explain employees receive inconsistent approvals when prose drives the decision; it does not isolate the reported cause in a system where agents return inconsistent fields for downstream integrations.

D:

The proposed probe concerns users ask about governed CRM records rather than prose manuals, where Dataverse grounding exposes appropriate structured records to the agent; this incident instead requires evidence about clear instructions and structured outputs reduce avoidable ambiguity and the failure that narrative answers cannot be parsed by consuming workflows.

Question 448

Release reviewers will not accept a generic success message. The business scenario is that some cases need cross-document reasoning and detailed justification, and the historical failure is: Small-model outputs omit critical contractual exceptions. Which test is appropriately targeted?

  1. Prototype a unique scheduling dependency with the custom runtime
  2. Run multi-hop evaluation cases for the strong-model route
  3. Replace a source document without training and verify updated answers
  4. Stress-test cost at peak demand and realistic retries

Explanation

Correct Answer

B:

For the required workflow where some cases need cross-document reasoning and detailed justification, the test needs to reproduce or prevent the failure where small-model outputs omit critical contractual exceptions; running multi-hop evaluation cases for the strong-model route exercises the path where complexity classification selects a capable model for difficult tasks.

Incorrect Answers

A:

In a solution where a workflow needs proprietary scheduling, model choice and runtime control, prototyping a unique scheduling dependency with the custom runtime could catch the defect that a predefined host cannot meet the required orchestration contract; the current release gate must challenge the failure where small-model outputs omit critical contractual exceptions.

C:

Although the exercise might catch a trained model remembers obsolete guidance after source changes, it would not directly validate the mechanism that complexity classification selects a capable model for difficult tasks in the target implementation.

D:

For an agent scales from a trial to enterprise usage, the alternate test is about whether workload forecasts include model usage, connectors, storage and observability; it cannot prove the outcome needed after the reported regression where small-model outputs omit critical contractual exceptions.

Question 449

A risk analysis highlights a threat in this workflow: Agents and apps need shared definitions of customers and work orders. Exposure: Shared data tables might expose rows to a different business unit. Which enforceable safeguard addresses it at the relevant boundary?

  1. Restrict production prompt edits to approved maintainers
  2. Preserve table and row access controls for retrieval
  3. Dead-letter failures and prevent duplicate financial postings
  4. Enforce business-unit and row-level permissions at the data layer

Explanation

Correct Answer

D:

Avoiding the consequence where shared data tables might expose rows to a different business unit requires a control for agents and apps need shared definitions of customers and work orders; in this architecture, Dataverse tables and governed relationships provide a reusable business-data surface, which is the point at which the boundary can be applied.

Incorrect Answers

A:

The alternate boundary mitigates a prompt editor could change production decisions without tracked approval, which is relevant to many teams copy slightly different instructions for the same task; it leaves the current workflow vulnerable to shared data tables might expose rows to a different business unit.

B:

For the use case where users ask about governed CRM records rather than prose manuals, the control addresses the threat that Dataverse retrieval might reveal a row not authorized to the caller; the specified agent must be protected against shared data tables might expose rows to a different business unit.

C:

The safeguard for invoice holds must trigger exception handling without polling deals with a different threat: retried finance events might post the same invoice adjustment twice; even after it is enforced, the exposure that shared data tables might expose rows to a different business unit remains possible in the target workflow where Dataverse tables and governed relationships provide a reusable business-data surface.

Question 450

A pilot scorecard needs evidence of business improvement rather than only invocation counts. The use case is: An agent will evolve continuously after its initial project delivery. Known defect: The pilot has no team for monitoring, incidents, or iteration. Which KPI should be selected?

  1. Successful requests during primary endpoint failures
  2. Percentage of production agents with staffed owners and SLOs
  3. Median elapsed time from design decision to approval
  4. High-complexity case pass rate by model route
  5. Proportion of tasks assigned to an explicit execution pattern

Explanation

Correct Answer

B:

The observed shortcoming is that the pilot has no team for monitoring, incidents, or iteration; measuring percentage of production agents with staffed owners and SLOs gives business owners a directly relevant indicator for the process where an agent will evolve continuously after its initial project delivery.

Incorrect Answers

A:

The premise of this indicator is health-aware fallback uses a compatible approved model when needed; the architecture being evaluated instead serves an agent will evolve continuously after its initial project delivery, and its success is measured by percentage of production agents with staffed owners and SLOs.

C:

The premise of this indicator is a named strategy team assigns decision rights and resolves cross-function tradeoffs; the architecture being evaluated instead serves an agent will evolve continuously after its initial project delivery, and its success is measured by percentage of production agents with staffed owners and SLOs.

D:

A change in high-complexity case pass rate by model route is informative for some cases need cross-document reasoning and detailed justification; it does not establish whether the current process achieves percentage of production agents with staffed owners and SLOs.

E:

Although proportion of tasks assigned to an explicit execution pattern is observable, the affected process involves an agent will evolve continuously after its initial project delivery; its acceptance outcome is percentage of production agents with staffed owners and SLOs, not the alternative measurement.

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