Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 102

Question 1010

One enterprise rollout has 3 separately testable requirements. (1) Business use: New behavior must be evaluated without exposing the full workforce. Known defect: A prompt update changes decisions for all users at once. (2) Business use: Teams propose that an agent perform both low-risk lookups and financial postings. Known defect: The same agent can read an invoice and commit a payment unchecked. (3) Business use: Copilot, Dataverse, Dynamics and custom APIs share one process. Known defect: Failures cannot be attributed to any single integration owner. Which 3 architectural choices directly satisfy these stated cases? Choose THREE.

  1. Define cross-service boundaries and ownership before integration
  2. Use staged rollout with feature flags and pilot cohorts
  3. Use a controlled baseline to measure causal impact
  4. Define autonomy tiers by consequence and reversibility

Explanation

Correct Answers

A:

For the independent requirement that copilot, Dataverse, Dynamics and custom APIs share one process, contracts and ownership boundaries identify responsibility for each state change; it addresses the defect failures cannot be attributed to any single integration owner, while new behavior must be evaluated without exposing the full workforce; teams propose that an agent perform both low-risk lookups and financial postings needs its own architecture.

B:

For the independent requirement that new behavior must be evaluated without exposing the full workforce, controlled release gates restrict blast radius and support rollback; it addresses the defect a prompt update changes decisions for all users at once, while teams propose that an agent perform both low-risk lookups and financial postings; copilot, Dataverse, Dynamics and custom APIs share one process needs its own architecture.

D:

For the independent requirement that teams propose that an agent perform both low-risk lookups and financial postings, tiered permissions require progressively stronger approvals for riskier actions; it addresses the defect the same agent can read an invoice and commit a payment unchecked, while new behavior must be evaluated without exposing the full workforce; copilot, Dataverse, Dynamics and custom APIs share one process needs its own architecture.

Incorrect Answers

C:

This design is associated with the different failure where seasonality produces the same change in unassisted teams in a workflow where management attributes a sales lift to Copilot without a comparison; it does not directly solve the listed conditions: a prompt update changes decisions for all users at once; the same agent can read an invoice and commit a payment unchecked; failures cannot be attributed to any single integration owner.

Question 1011

The pilot needs a durable correction: Shared admin credentials return restricted cases to ordinary users. This affects the workflow in which agent looks up records on behalf of employees with different rights. Which design choice should the architect approve?

  1. Audit privileged agent actions with correlation to user intent
  2. Use delegated user identity for user-scoped data actions
  3. Test prompt changes on paraphrases and edge conditions
  4. Use managed identities for service-to-service agent components

Explanation

Correct Answer

B:

The architecture must reconcile two facts: agent looks up records on behalf of employees with different rights, yet shared admin credentials return restricted cases to ordinary users; the selected pattern does so through the mechanism that delegated authorization evaluates each caller's effective permissions.

Incorrect Answers

A:

End-to-end audit binds user, tool, affected record and authorization event; its main purpose is a tool deletes a record following an ambiguous conversation, not correction of the reported defect where shared admin credentials return restricted cases to ordinary users in the target process.

C:

In the case where a rewritten prompt improves a single demonstration, the alternate design depends on regression sets cover lexical variation and task boundaries; it does not correct the target failure, shared admin credentials return restricted cases to ordinary users.

D:

Long-lived API keys leak from configuration and diagnostics; the alternate design can address that situation because managed identity provides credential-less workload authentication where supported, whereas this question needs a solution to shared admin credentials return restricted cases to ordinary users.

Question 1012

Operations has not established the cause of this defect: A rigid menu fails whenever a request contains two intents. The associated setting is: Employees express several intents and need interactive clarification. Which inspection should precede remediation?

  1. Inspect pending tasks and final-result callbacks by correlation ID
  2. Review intent traces for misunderstood or unhandled requests
  3. Inspect task envelopes and agent capability declarations
  4. Capture advertised capabilities and actual authorized tool calls

Explanation

Correct Answer

B:

Where a rigid menu fails whenever a request contains two intents, the useful evidence comes from reviewing intent traces for misunderstood or unhandled requests; its focus matches the architectural mechanism that generative orchestration interprets intent and can combine authorized tools and knowledge in the process where employees express several intents and need interactive clarification.

Incorrect Answers

A:

In the separate scenario where downstream research agents run longer than the interactive request, the same investigation could explain orchestrator returns a partial result as though the workflow finished; it does not isolate the reported cause in a system where employees express several intents and need interactive clarification.

C:

The evidence this method produces applies to a workflow where independently hosted agents must exchange tasks and results; it cannot by itself establish why a rigid menu fails whenever a request contains two intents in the current scenario.

D:

The proposed inspection is associated with a Copilot Studio agent needs a reusable interface to governed operations, where MCP presents supported tools and resources through a common integration contract; it would not expose the failure mode that a rigid menu fails whenever a request contains two intents with the required specificity.

Question 1013

Troubleshooting starts from a business complaint: Feedback tickets lack the evidence needed to reproduce them. The affected process is: Users report stale answers after a policy refresh. Which investigation is most directly tied to the reported defect?

  1. Compare query, filters and top retrieved source identifiers
  2. Review virtual-entity metadata and source-field changes
  3. Compare complaint time with retrieved source and model versions
  4. Review critical-path durations for slow requests

Explanation

Correct Answer

C:

The reported fault is that feedback tickets lack the evidence needed to reproduce them in a system where users report stale answers after a policy refresh; evidence from comparing complaint time with retrieved source and model versions can show whether trace and content-version linkage makes failures testable.

Incorrect Answers

A:

In the process where an assistant responds fluently but cites unrelated support articles, the diagnostic procedure could explain surface quality scores appear healthy despite wrong documents; this particular incident instead turns on whether trace and content-version linkage makes failures testable and why feedback tickets lack the evidence needed to reproduce them.

B:

AI assistance reads operational records through Dataverse; its matching diagnostic check is reasonable for an update changes a field mapping and grounded answers become stale; the present investigation needs evidence for the different failure where feedback tickets lack the evidence needed to reproduce them.

D:

The proposed probe concerns agents regularly exceed an interactive response-time objective, where per-span latency isolates model work, network waits and tool execution; this incident instead requires evidence about trace and content-version linkage makes failures testable and the failure that feedback tickets lack the evidence needed to reproduce them.

Question 1014

A product owner must measure whether an AI-enabled change actually improves this process: Standard capabilities draft customer communications but policy requires approval. The earlier problem was: Automatically generated notices are sent without required review. Which outcome indicator is most relevant?

  1. Cross-app case-state agreement rate
  2. Share of grounding records with complete governance metadata
  3. Avoidable custom-maintenance cost per standard task
  4. Outbound notices released with recorded approval

Explanation

Correct Answer

D:

The observed shortcoming is that automatically generated notices are sent without required review; measuring outbound notices released with recorded approval gives business owners a directly relevant indicator for the process where standard capabilities draft customer communications but policy requires approval.

Incorrect Answers

A:

For the workflow where sales and service agents need the same customer and case context, cross-app case-state agreement rate is an actionable measure; it does not establish whether automatically generated notices are sent without required review in the situation under review.

B:

A change in share of grounding records with complete governance metadata is informative for search spans internal manuals with different owners and permissions; it does not establish whether the current process achieves outbound notices released with recorded approval.

C:

Avoidable custom-maintenance cost per standard task; that indicator describes a distinct workflow where users need ordinary meeting summaries and follow-up drafts; it does not quantify outbound notices released with recorded approval, which is needed for the stated case.

Question 1015

One enterprise rollout has 2 separately testable requirements. (1) Business use: A model is replaced without a traceable change ticket. Known defect: Investigators cannot explain when behavior changed. (2) Business use: An updated agent performs differently months after release. Known defect: Old scores cannot be reproduced because prompts and tests changed. Which 2 regression tests directly satisfy these stated cases? Choose TWO.

  1. Introduce a known distribution shift and verify review triggers
  2. Re-run a sampled historical release on its frozen corpus
  3. Submit a malicious mislabeled training record and inspect review
  4. Promote a candidate in staging and inspect generated audit entries

Explanation

Correct Answers

B:

For the business workflow where an updated agent performs differently months after release, re-runing a sampled historical release on its frozen corpus challenges the failure that old scores cannot be reproduced because prompts and tests changed; the release also needs a separate check for a model is replaced without a traceable change ticket.

D:

For the business workflow where a model is replaced without a traceable change ticket, promoting a candidate in staging and inspecting generated audit entries challenges the failure that investigators cannot explain when behavior changed; the release also needs a separate check for an updated agent performs differently months after release.

Incorrect Answers

A:

This regression test is associated with the different failure where team retrains monthly despite unchanged model behavior in a workflow where customer terminology changes over time; it does not directly solve the listed conditions: investigators cannot explain when behavior changed; old scores cannot be reproduced because prompts and tests changed.

C:

This regression test is associated with the different failure where unvetted labeled data poisons a retrained classifier in a workflow where business analysts can submit examples but not change production behavior; it does not directly solve the listed conditions: investigators cannot explain when behavior changed; old scores cannot be reproduced because prompts and tests changed.

Question 1016

Troubleshooting starts from a business complaint: The automated suite approves the same defect it should detect. The affected process is: Copilot writes a test expecting a plausible but wrong answer. Which investigation is most directly tied to the reported defect?

  1. Reconcile audit IDs across each application
  2. Compare expected answers against source-of-truth rules
  3. Inspect imported connection references and environment variables
  4. Inspect clustered examples for shared underlying cause

Explanation

Correct Answer

B:

In the setting where Copilot writes a test expecting a plausible but wrong answer, comparing expected answers against source-of-truth rules can isolate the failure in which the automated suite approves the same defect it should detect, because independent acceptance references prevent self-confirming evaluation.

Incorrect Answers

A:

The proposed inspection is associated with service recommends a price adjustment requiring finance approval, where transaction-path tests confirm recommendation, review and posting records; it would not expose the failure mode that the automated suite approves the same defect it should detect with the required specificity.

C:

The proposed probe concerns a flow moved to production still points to a test endpoint, where environment configuration separates deployable logic from endpoint bindings; this incident instead requires evidence about independent acceptance references prevent self-confirming evaluation and the failure that the automated suite approves the same defect it should detect.

D:

In the separate scenario where thousands of transcripts contain recurring misunderstandings, the same investigation could explain analysts overlook a widespread intent failure among unique phrasings; it does not isolate the reported cause in a system where Copilot writes a test expecting a plausible but wrong answer.

Question 1017

A quality engineer is writing a regression case after the incident where the conversation alone authorizes a permanent business change. The underlying business scenario is: The agent may refund large amounts or deactivate accounts. What should the test exercise?

  1. Propose a concession above limit and confirm approval enforcement
  2. Test adversarial lookalike cases not present in prompt examples
  3. Replay boundary cases and verify the same branch and approver are selected
  4. Deliver duplicate events and confirm single reversible updates
  5. Request a high-value refund and verify approval blocks execution

Explanation

Correct Answer

E:

Requesting a high-value refund and verifying approval blocks execution is aligned with the failure that the conversation alone authorizes a permanent business change; its causal relevance is that confirmation and human approval separate advice from execution authority for the situation where the agent may refund large amounts or deactivate accounts.

Incorrect Answers

A:

In the alternate setting where service, sales and finance jointly decide high-value concessions, the test exercises cross-app workflow routes the recommendation to the accountable approver; even a pass would not show that the conversation alone authorizes a permanent business change has been prevented.

B:

Although the exercise might catch a generic instruction cannot express subtle category boundaries, it would not directly validate the mechanism that confirmation and human approval separate advice from execution authority in the target implementation.

C:

For the task follows fixed approval branches with no interpretation needed, the alternate test is about whether explicit conditions execute known business rules predictably and leave an auditable branch history; it cannot prove the outcome needed after the reported regression where the conversation alone authorizes a permanent business change.

D:

The verification concerning work can proceed from events without human initiation but needs bounded autonomy checks event triggers and tool permissions can automate predefined safe steps within explicit limits against a background agent has triggered irreversible duplicate actions; its result says little about the observed issue where the conversation alone authorizes a permanent business change.

Question 1018

A proposed deployment supports this business case: A new action version changes how cases are updated. Reviewers also warn that a forced API cutover might make rollback impossible for an old agent version. Which restriction mitigates that concrete risk?

  1. Limit production edit roles and monitor drift
  2. Block release if privileged writes bypass approvals
  3. Restrict promotion to reviewed compatible configurations
  4. Keep prior compatible API routes during release transition

Explanation

Correct Answer

D:

Because a forced API cutover might make rollback impossible for an old agent version, a guardrail must constrain the actual system behavior: coordinated versioning preserves consumer/provider compatibility; protecting a response alone cannot enforce that boundary.

Incorrect Answers

A:

The proposed control targets direct production customization might diverge from the approved solution, while the given risk scenario is a forced API cutover might make rollback impossible for an old agent version; these exposures arise in different operational contexts.

B:

Agents suggest payment adjustments and inventory holds; its guardrail addresses the possibility that an updated connector might gain a direct privileged posting route; the present architecture must constrain the different threat that a forced API cutover might make rollback impossible for an old agent version.

C:

For the use case where model and tool changes are promoted independently, the control addresses the threat that an agent release might pair instructions with incompatible tool or model versions; the specified agent must be protected against a forced API cutover might make rollback impossible for an old agent version.

Question 1019

An investment review asks what measurable result would justify continuing an AI solution where the task follows fixed approval branches with no interpretation needed. The original issue is: Employees receive inconsistent approvals when prose drives the decision. Which indicator should be tracked?

  1. Percentage of decisions matching the signed decision table
  2. Successful task completion across unseen paraphrases
  3. Interoperable tasks completed without custom point-to-point parsing
  4. Unauthorized tool invocation success rate

Explanation

Correct Answer

A:

For the use case where the task follows fixed approval branches with no interpretation needed, the indicator percentage of decisions matching the signed decision table is worth baselining against the earlier issue where employees receive inconsistent approvals when prose drives the decision.

Incorrect Answers

B:

Although successful task completion across unseen paraphrases is observable, the affected process involves the task follows fixed approval branches with no interpretation needed; its acceptance outcome is percentage of decisions matching the signed decision table, not the alternative measurement.

C:

The proposed metric concerns independently hosted agents must exchange tasks and results, whereas the failure in this case is that employees receive inconsistent approvals when prose drives the decision; the indicator needed here is percentage of decisions matching the signed decision table.

D:

Although unauthorized tool invocation success rate is observable, the affected process involves the task follows fixed approval branches with no interpretation needed; its acceptance outcome is percentage of decisions matching the signed decision table, not the alternative measurement.

Question 1020

An enterprise team is making a platform design decision. Business context: Third-party documents include sentences telling the model to ignore rules. Present failure: Answers adopt an instruction embedded in a source document. Which approach resolves the mismatch?

  1. Synchronize an indexed source on an explicit freshness schedule
  2. Separate retrieved data from trusted system instructions
  3. Build a bespoke agent when differentiated orchestration is essential
  4. Version reusable prompts with owners and evaluation examples

Explanation

Correct Answer

B:

The defect is that answers adopt an instruction embedded in a source document; the expected business case is that third-party documents include sentences telling the model to ignore rules; the design fits because trust boundaries treat source text as evidence rather than authority.

Incorrect Answers

A:

When operational records update faster than the agent's indexed copy, incremental refresh and source-to-index lag monitoring reduce staleness; in the stated workflow, however, answers adopt an instruction embedded in a source document; the relevant architectural behavior is that trust boundaries treat source text as evidence rather than authority.

C:

A boxed product cannot express the execution and ownership rules; the alternate design can address that situation because custom engineering trades more control for more operating responsibility, whereas this question needs a solution to answers adopt an instruction embedded in a source document.

D:

This design would make sense if many teams copy slightly different instructions for the same task, particularly where a successful prompt change cannot be reproduced after rollout; the requested architecture instead needs the dependency that trust boundaries treat source text as evidence rather than authority because answers adopt an instruction embedded in a source document.

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