Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 70

Question 690

Security and business owners want to keep this capability: The same agent must move from development to test and production. They must also prevent the following: A maker might import or publish an agent without release approval. Which guardrail should the design incorporate?

  1. Require a tested human escalation path for unmatched cases
  2. Stop release on unresolved breaking contract mismatches
  3. Enforce approval before production import and publication
  4. Keep prior compatible API routes during release transition

Explanation

Correct Answer

C:

The business setting is the same agent must move from development to test and production, with the exposure that a maker might import or publish an agent without release approval; the enforceable protection must correspond to the path where solution-based ALM transports supported components under controlled releases.

Incorrect Answers

A:

The alternate boundary mitigates a new support agent might route urgent cases to a general queue, which is relevant to a support agent changes how it classifies priority intents; it leaves the current workflow vulnerable to a maker might import or publish an agent without release approval.

B:

The alternate boundary mitigates a breaking connector schema might be released despite failed contract tests, which is relevant to an external API changes a required field used by an agent; it leaves the current workflow vulnerable to a maker might import or publish an agent without release approval.

D:

For the use case where a new action version changes how cases are updated, the control addresses the threat that a forced API cutover might make rollback impossible for an old agent version; the specified agent must be protected against a maker might import or publish an agent without release approval.

Question 691

The fault affects an AI-powered workflow in which multiple low-code apps need a common extraction prompt. Specifically, each app drifts to a different unreviewed AI prompt. Which diagnostic investigation best distinguishes the relevant failure path?

  1. Review agent capability manifest against requested functions
  2. Validate indexed records and source ACL mappings
  3. Compare prompt versions used by dependent apps
  4. Compare advertised tools with backend permissions

Explanation

Correct Answer

C:

Each app drifts to a different unreviewed AI prompt; to attribute the defect rather than guess at a fix, comparing prompt versions used by dependent apps examines the dependency where centralized reusable AI capabilities promote consistency and controlled updates.

Incorrect Answers

A:

Reviewing agent capability manifest against requested functions explores a separate connection, where declarative agents define instructions, knowledge and actions in the supported host; the diagnosed requirement is multiple low-code apps need a common extraction prompt, and the actual defect is each app drifts to a different unreviewed AI prompt.

B:

Copilot cannot find approved wiki pages by natural language; examining that other path does not determine whether centralized reusable AI capabilities promote consistency and controlled updates, which is what distinguishes the issue in the stated use case.

D:

For the failure that each agent builds a different unaudited HTTP wrapper, comparing advertised tools with backend permissions is relevant; this question instead reports that each app drifts to a different unreviewed AI prompt, so different evidence is needed.

Question 692

During integration testing, teams learn about problems only when someone asks the agent. The associated business context is: Shipment exceptions should initiate investigation without a chat message. Which architectural remedy is the strongest match?

  1. Provide source-linked explanations for analytical responses
  2. Specify stop conditions and budgets for long-running agents
  3. Use agent reasoning for uncertain multi-step analysis with bounded tools
  4. Trigger an autonomous agent from a governed business event

Explanation

Correct Answer

D:

Shipment exceptions should initiate investigation without a chat message; its failing behavior is that teams learn about problems only when someone asks the agent; applying the underlying design principle—an event trigger starts the bounded agent workflow on state change—connects the requirement to an implementable control.

Incorrect Answers

A:

This design would make sense if finance asks why a variance was flagged, particularly where answers assert causes without referencing the ledger; the requested architecture instead needs the dependency that an event trigger starts the bounded agent workflow on state change because teams learn about problems only when someone asks the agent.

B:

Time, step and spend budgets terminate unproductive loops; its main purpose is a research agent repeatedly queries sources on difficult cases, not correction of the reported defect where teams learn about problems only when someone asks the agent in the target process.

C:

A simple scripted topic cannot handle legitimate variants; the alternate design can address that situation because reasoning can plan through evidence and tools within enforced boundaries, whereas this question needs a solution to teams learn about problems only when someone asks the agent.

Question 693

A quality engineer is writing a regression case after the incident where the agent cites a superseded rule despite a newer policy existing. The underlying business scenario is: Policy documents contain draft, superseded, and currently effective revisions. What should the test exercise?

  1. Compare matched assisted and unassisted case cohorts
  2. Ask about a revised policy and verify an allowed citation
  3. Replace a source document without training and verify updated answers
  4. Re-evaluate the router on production-like stratified examples
  5. Query immediately before and after an effective-date change

Explanation

Correct Answer

E:

The historical failure is that the agent cites a superseded rule despite a newer policy existing; for the setting where policy documents contain draft, superseded, and currently effective revisions, querying immediately before and after an effective-date change verifies the dependency where version and effective-date metadata select authoritative content at retrieval time.

Incorrect Answers

A:

Comparing matched assisted and unassisted case cohorts verifies an adjacent objective where AI drafts requests quickly but specialists must correct some outputs; it leaves untested the causal path in which version and effective-date metadata select authoritative content at retrieval time for this case.

B:

The suggested test serves employee policies reside in governed Microsoft 365 documents, in which answers omit current employee handbook changes; the present scenario instead concerns the reported problem where the agent cites a superseded rule despite a newer policy existing.

C:

In a solution where answers depend on policies revised every week, replacing a source document without training and verifying updated answers could catch the defect that a trained model remembers obsolete guidance after source changes; the current release gate must challenge the failure where the agent cites a superseded rule despite a newer policy existing.

D:

In the alternate setting where real task mix differs from prelaunch performance tests, the test exercises feedback-driven thresholds rebalance accuracy, latency and spend; even a pass would not show that the agent cites a superseded rule despite a newer policy existing has been prevented.

Question 694

The pilot needs a durable correction: Model behavior changes without reproducible training provenance. This affects the workflow in which retraining uses revised labels and a different source snapshot. Which design choice should the architect approve?

  1. Select a custom model for evidence-backed specialized extraction
  2. Chunk long documents with stable metadata and source links
  3. Apply entity-aware filtering before grounding customer answers
  4. Version models and datasets together for controlled adaptation

Explanation

Correct Answer

D:

The defect is that model behavior changes without reproducible training provenance; the expected business case is that retraining uses revised labels and a different source snapshot; the design fits because dataset lineage and model versioning support repeatable assessment.

Incorrect Answers

A:

In the case where general models repeatedly misread rare engineering labels, the alternate design depends on domain-trained models can be evaluated when generic baselines fall short; it does not correct the target failure, model behavior changes without reproducible training provenance.

B:

For retrieved policy fragments lose context when documents are split, this other approach relies on the fact that semantically coherent chunks with provenance preserve context during retrieval; the issue here is model behavior changes without reproducible training provenance; that mechanism does not establish the required dataset lineage and model versioning support repeatable assessment.

C:

The alternative addresses retrieval mixes records from two organizations by a design in which tenant and entity filters constrain candidate records before generation; it does not supply the missing relationship for retraining uses revised labels and a different source snapshot, where model behavior changes without reproducible training provenance.

Question 695

In a business situation where peak load can saturate a primary model endpoint, a previous build showed this defect: Requests fail despite another approved deployment being available. Which reproducible test belongs in the release gate?

  1. Simulate endpoint failure and verify continuity without policy bypass
  2. Simulate a production regression and locate the accountable response team
  3. Prototype a unique scheduling dependency with the custom runtime
  4. Stress-test cost at peak demand and realistic retries

Explanation

Correct Answer

A:

Simulating endpoint failure and verifying continuity without policy bypass challenges the path where health-aware fallback uses a compatible approved model when needed for the business setting peak load can saturate a primary model endpoint; the specific defect to exclude is that requests fail despite another approved deployment being available.

Incorrect Answers

B:

This check exercises a different system behavior: long-lived product ownership funds ongoing evaluation and operations; a pass there cannot establish that requests fail despite another approved deployment being available no longer occurs in the required workflow.

C:

Prototyping a unique scheduling dependency with the custom runtime verifies an adjacent objective where a workflow needs proprietary scheduling, model choice and runtime control; it leaves untested the causal path in which health-aware fallback uses a compatible approved model when needed for this case.

D:

The verification concerning an agent scales from a trial to enterprise usage checks workload forecasts include model usage, connectors, storage and observability against token and action costs exceed the pilot estimate; its result says little about the observed issue where requests fail despite another approved deployment being available.

Question 696

One enterprise rollout has 2 separately testable requirements. (1) Business use: A blocked invoice triggers a hold on related procurement actions. Existing shortcoming: A network fault leaves the two apps inconsistent. (2) Business use: A won opportunity should create a properly owned onboarding case. Existing shortcoming: AI recommends an action but downstream case data is incomplete. Which 2 success metrics directly satisfy these stated cases? Choose TWO.

  1. Cross-app exceptions ending in consistent state
  2. Unauthorized finance postings in release tests
  3. Complete onboarding cases with accurate Sales context
  4. Regional resolution and escalation change after deployment

Explanation

Correct Answers

A:

The outcome cross-app exceptions ending in consistent state measures whether the workflow where a blocked invoice triggers a hold on related procurement actions is improving; the scorecard separately needs complete onboarding cases with accurate Sales context.

C:

The outcome complete onboarding cases with accurate Sales context measures whether the workflow where a won opportunity should create a properly owned onboarding case is improving; the scorecard separately needs cross-app exceptions ending in consistent state.

Incorrect Answers

B:

The alternative measures unauthorized finance postings in release tests for the situation where agents suggest payment adjustments and inventory holds; it cannot substitute for performance outcomes in the separate named cases: a blocked invoice triggers a hold on related procurement actions; a won opportunity should create a properly owned onboarding case.

D:

The alternative measures regional resolution and escalation change after deployment for the situation where a conversational change affects only one support geography; it cannot substitute for performance outcomes in the separate named cases: a blocked invoice triggers a hold on related procurement actions; a won opportunity should create a properly owned onboarding case.

Question 697

An investment review asks what measurable result would justify continuing an AI solution where two systems share company names but not primary keys. The original issue is: Copilot attaches one customer's risk note to another account. Which indicator should be tracked?

  1. Fresh authorized account insights from external systems
  2. Correct identification of governed service intents
  3. Unauthorized financial commitment language rate
  4. Correct external-to-Dynamics account match rate

Explanation

Correct Answer

D:

Because a mapping service resolves stable identities before enrichment, the outcome worth following is correct external-to-Dynamics account match rate; the metric is tied to the reported issue—Copilot attaches one customer's risk note to another account—for the scenario where two systems share company names but not primary keys.

Incorrect Answers

A:

Although fresh authorized account insights from external systems is observable, the affected process involves two systems share company names but not primary keys; its acceptance outcome is correct external-to-Dynamics account match rate, not the alternative measurement.

B:

For the workflow where agents must distinguish an incident, complaint, inquiry and entitlement, correct identification of governed service intents is an actionable measure; it does not establish whether Copilot attaches one customer's risk note to another account in the situation under review.

C:

That scorecard concerns the activity in which refund, credit, reimbursement and adjustment have different accounting impacts and its result of unauthorized financial commitment language rate; the reported shortcoming here is Copilot attaches one customer's risk note to another account.

Question 698

The operating team must monitor whether an architectural correction works. Business context: Agent experience must appear in selected Microsoft 365 contexts. Historical symptom: An agent package is assumed to work in every host automatically. Which outcome measure is the best match?

  1. Successful authorized external service actions
  2. Adoption among eligible users in supported hosts
  3. Channel-appropriate responses passing usability review
  4. Sales task completion under external API impairment

Explanation

Correct Answer

B:

The use case involves agent experience must appear in selected Microsoft 365 contexts and the baseline failure that an agent package is assumed to work in every host automatically; the KPI tracks adoption among eligible users in supported hosts, so improvement has an interpretable operational meaning.

Incorrect Answers

A:

For the situation where a Teams user should query and update a supported external service, successful authorized external service actions is a sensible indicator; the problem that an agent package is assumed to work in every host automatically requires measurement of adoption among eligible users in supported hosts.

C:

Although channel-appropriate responses passing usability review is observable, the affected process involves agent experience must appear in selected Microsoft 365 contexts; its acceptance outcome is adoption among eligible users in supported hosts, not the alternative measurement.

D:

A change in sales task completion under external API impairment is informative for external credit service responds slowly during regional outages; it does not establish whether the current process achieves adoption among eligible users in supported hosts.

Question 699

An agent rollout is intended for this situation: An agent must turn service notes into a fixed JSON summary. The known problem is that a general conversational turn produces inconsistent output fields. Which success measure connects to that specific outcome?

  1. Completed tasks with all required input fields
  2. Correct tool selection and input-binding rate
  3. Prompt action responses conforming to output contract
  4. Role-specific tasks completed in supported Copilot surfaces

Explanation

Correct Answer

C:

The use case involves an agent must turn service notes into a fixed JSON summary and the baseline failure that a general conversational turn produces inconsistent output fields; the KPI tracks prompt action responses conforming to output contract, so improvement has an interpretable operational meaning.

Incorrect Answers

A:

For the situation where an intake task requires several prerequisite fields, completed tasks with all required input fields is a sensible indicator; the problem that a general conversational turn produces inconsistent output fields requires measurement of prompt action responses conforming to output contract.

B:

The orchestrator chooses between similar invoice and credit tools; its intended measurement is correct tool selection and input-binding rate; substituting that result here would not capture whether a general conversational turn produces inconsistent output fields has been corrected.

D:

For the workflow where employees want curated assistance without owning a separate LLM runtime, role-specific tasks completed in supported Copilot surfaces is an actionable measure; it does not establish whether a general conversational turn produces inconsistent output fields in the situation under review.

Question 700

Governance approves the business purpose but requires control of this exposure: A flow might notify the customer even though a required approval failed. The agent will be used when approval processing requires sequential verification and notification. Which boundary is most defensible?

  1. Disable write tools and label generated material for review
  2. Reject unvalidated model output before ticket creation
  3. Ensure the backend rejects repeat writes for the same intent
  4. Stop and compensate when a required approval fails

Explanation

Correct Answer

D:

Because a flow might notify the customer even though a required approval failed, a guardrail must constrain the actual system behavior: defined flow steps make sequencing and error paths explicit; protecting a response alone cannot enforce that boundary.

Incorrect Answers

A:

The safeguard for marketing specialists need tailored descriptions from existing briefings deals with a different threat: a text-only drafting assistant might be given unnecessary record-writing powers; even after it is enforced, the exposure that a flow might notify the customer even though a required approval failed remains possible in the target workflow where defined flow steps make sequencing and error paths explicit.

B:

The alternate boundary mitigates a malformed structured summary might create corrupt service tickets, which is relevant to a summarizer sends output into a ticket ingestion service; it leaves the current workflow vulnerable to a flow might notify the customer even though a required approval failed.

C:

The safeguard for network retries can repeat a single purchase request deals with a different threat: an API retry might create a second purchase for the same user instruction; even after it is enforced, the exposure that a flow might notify the customer even though a required approval failed remains possible in the target workflow where defined flow steps make sequencing and error paths explicit.

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