Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 105

Question 1040

Risk reviewers need a control, not merely a monitoring dashboard. The use case is: Peak load can saturate a primary model endpoint. Potential harm: Fallback routing might send sensitive workloads to a disallowed model location. Which protection fits?

  1. Budget for independent hosting, safety testing and monitoring
  2. Gate launch on ownership, runbooks and support budgets
  3. Set usage budgets, cost alerts and routing limits
  4. Validate residency, safety and output contract on fallback models

Explanation

Correct Answer

D:

In the context where peak load can saturate a primary model endpoint, the exposure is that fallback routing might send sensitive workloads to a disallowed model location; containing it requires a guardrail in the operation where health-aware fallback uses a compatible approved model when needed.

Incorrect Answers

A:

Another process could need this guardrail because a bespoke runtime might be shipped without an accountable security or support team; the design being assessed supports peak load can saturate a primary model endpoint and has the explicit exposure that fallback routing might send sensitive workloads to a disallowed model location.

B:

For the use case where an agent will evolve continuously after its initial project delivery, the control addresses the threat that production agents could launch without an incident owner, runbook or operations funding; the specified agent must be protected against fallback routing might send sensitive workloads to a disallowed model location.

C:

An agent scales from a trial to enterprise usage; its guardrail addresses the possibility that increased model traffic might run beyond the project budget without alerts; the present architecture must constrain the different threat that fallback routing might send sensitive workloads to a disallowed model location.

Question 1041

An agent-assisted service does not behave as specified: Template parsing omits line items on unfamiliar forms. It is used in this situation: Vendors send variable-layout invoices and shipping documents. Which observable diagnostic procedure is best?

  1. Examine query filters and retrieved tenant IDs
  2. Inspect chunk boundaries and parent document links
  3. Examine schema validation logs for missing or invalid properties
  4. Compare extraction failures by document layout

Explanation

Correct Answer

D:

In the setting where vendors send variable-layout invoices and shipping documents, comparing extraction failures by document layout can isolate the failure in which template parsing omits line items on unfamiliar forms, because document processing models extract structured fields from document content.

Incorrect Answers

A:

For the failure that retrieval mixes records from two organizations, examining query filters and retrieved tenant IDs is relevant; this question instead reports that template parsing omits line items on unfamiliar forms, so different evidence is needed.

B:

In the process where retrieved policy fragments lose context when documents are split, the diagnostic procedure could explain answers cite isolated sentences without section qualifiers; this particular incident instead turns on whether document processing models extract structured fields from document content and why template parsing omits line items on unfamiliar forms.

C:

Examining schema validation logs for missing or invalid properties explores a separate connection, where structured outputs define required fields and allowable values; the diagnosed requirement is vendors send variable-layout invoices and shipping documents, and the actual defect is template parsing omits line items on unfamiliar forms.

Question 1042

A business stakeholder describes this operating situation: Candidate processes are attractive but permissions and data quality are weak. Testing instead shows that the first-wave pilot requests broad access to unreliable records. What design approach is most appropriate?

  1. Provide a unified approval handoff for AI-identified account risks
  2. Compute ROI from incremental net benefits and full lifecycle costs
  3. Sequence AI adoption by data and security maturity
  4. Start with documented business outcomes and adoption readiness

Explanation

Correct Answer

C:

The architecture must reconcile two facts: candidate processes are attractive but permissions and data quality are weak, yet the first-wave pilot requests broad access to unreliable records; the selected pattern does so through the mechanism that readiness assessment places safe high-value workloads ahead of risky ones.

Incorrect Answers

A:

Cross-app workflow routes the recommendation to the accountable approver; its main purpose is service, sales and finance jointly decide high-value concessions, not correction of the reported defect where the first-wave pilot requests broad access to unreliable records in the target process.

B:

There is a valid place for the mechanism that net benefit compares realized outcomes with implementation and recurring costs, namely a pilot reports gross labor savings but omits operating expenses; it does not remove the specific defect that the first-wave pilot requests broad access to unreliable records from this case.

D:

The alternative addresses prototype activity increases but no business value is measured by a design in which cloud adoption planning begins from motivations, outcomes and readiness constraints; it does not supply the missing relationship for candidate processes are attractive but permissions and data quality are weak, where the first-wave pilot requests broad access to unreliable records.

Question 1043

Troubleshooting starts from a business complaint: A document index cannot answer current case ownership reliably. The affected process is: Users ask about governed CRM records rather than prose manuals. Which investigation is most directly tied to the reported defect?

  1. Correlate retrieval timeouts with unsupported responses
  2. Compare errors for historical and newly effective rules
  3. Trace transactions using end-to-end correlation IDs
  4. Inspect entity mappings and security context at retrieval

Explanation

Correct Answer

D:

The reported fault is that a document index cannot answer current case ownership reliably in a system where users ask about governed CRM records rather than prose manuals; evidence from inspecting entity mappings and security context at retrieval can show whether Dataverse grounding exposes appropriate structured records to the agent.

Incorrect Answers

A:

That probe evaluates health checks and explicit no-evidence responses handle missing context safely in a workflow where required knowledge stores are intermittently unreachable; the reported failure concerns the path for which Dataverse grounding exposes appropriate structured records to the agent instead.

B:

That probe evaluates retrieval separates volatile knowledge from model parameters in a workflow where answers depend on policies revised every week; the reported failure concerns the path for which Dataverse grounding exposes appropriate structured records to the agent instead.

C:

Copilot, Dataverse, Dynamics and custom APIs share one process; its matching diagnostic check is reasonable for failures cannot be attributed to any single integration owner; the present investigation needs evidence for the different failure where a document index cannot answer current case ownership reliably.

Question 1044

One enterprise rollout has 2 separately testable requirements. (1) Business use: Complaints use vague labels such as agent is wrong. Exposure: A serious privacy incident might be triaged behind cosmetic chatbot defects. (2) Business use: Fixes are marked complete when a prompt is deployed. Exposure: A published fix might be closed even after causing a new critical regression. Which 2 safeguards directly satisfy these stated cases? Choose TWO.

  1. Reject broad admin consent not required by the workload
  2. Reopen fixes that regress unrelated critical behaviors
  3. Prioritize incidents with safety or privacy impact
  4. Require heightened review for autonomous privileged agents
  5. Require new data rights and labeling checks before training

Explanation

Correct Answers

B:

For the exposure that a published fix might be closed even after causing a new critical regression, outcome verification checks whether the change fixed the user problem explains the protection required when fixes are marked complete when a prompt is deployed; this does not replace control of a serious privacy incident might be triaged behind cosmetic chatbot defects.

C:

For the exposure that a serious privacy incident might be triaged behind cosmetic chatbot defects, structured defect taxonomy assigns the right remediation owner explains the protection required when complaints use vague labels such as agent is wrong; this does not replace control of a published fix might be closed even after causing a new critical regression.

Incorrect Answers

A:

This security safeguard addresses the unrelated exposure that production consent might grant a wider tenant scope than necessary in a workflow where an imported agent can authenticate in dev but not production; the question instead names the specific risks a serious privacy incident might be triaged behind cosmetic chatbot defects; a published fix might be closed even after causing a new critical regression; this choice does not directly mitigate those specific exposures.

D:

This security safeguard addresses the unrelated exposure that a privileged autonomous agent might receive the same review as a private helper in a workflow where personal read-only assistants and payment agents have different risk; the question instead names the specific risks a serious privacy incident might be triaged behind cosmetic chatbot defects; a published fix might be closed even after causing a new critical regression; this choice does not directly mitigate those specific exposures.

E:

This security safeguard addresses the unrelated exposure that retraining might incorporate sensitive examples without renewed data rights review in a workflow where customer terminology changes over time; the question instead names the specific risks a serious privacy incident might be triaged behind cosmetic chatbot defects; a published fix might be closed even after causing a new critical regression; this choice does not directly mitigate those specific exposures.

Question 1045

The pilot needs a durable correction: Supervisors routinely click approve without seeing evidence. This affects the workflow in which AI proposes eligibility decisions affecting customers. Which design choice should the architect approve?

  1. Design meaningful human oversight for consequential recommendations
  2. Provide clear disclosure when users interact with AI
  3. Monitor newly released service agent behavior by customer cohort
  4. Regression-test case routing after service agent changes

Explanation

Correct Answer

A:

For the need that AI proposes eligibility decisions affecting customers, the relevant mechanism is that explainable review interfaces support accountable judgment; that architectural dependency addresses the observed failure where supervisors routinely click approve without seeing evidence.

Incorrect Answers

B:

This design would make sense if customers believe generated advice is a human decision, particularly where users cannot identify automated recommendations; the requested architecture instead needs the dependency that explainable review interfaces support accountable judgment because supervisors routinely click approve without seeing evidence.

C:

For another type of architecture, cohort telemetry supports safe staged validation is useful; this design must instead support AI proposes eligibility decisions affecting customers despite the problem where supervisors routinely click approve without seeing evidence.

D:

High-severity tickets begin routing to general queues; the alternate design can address that situation because controlled service evaluations validate routing before rollout, whereas this question needs a solution to supervisors routinely click approve without seeing evidence.

Question 1046

Before production promotion, the team needs evidence for this setting: Stakeholders report hallucinations but cannot isolate search errors. A prior regression caused the following: Generator tuning fails because the required documents were never retrieved. Which validation is most persuasive?

  1. Reproduce a candidate evaluation from the recorded dataset
  2. Evaluate difficult queries against annotated source passages
  3. Change a permitted credit field and verify updated authorized retrieval
  4. Open the page for different roles and verify context limits

Explanation

Correct Answer

B:

Evaluating difficult queries against annotated source passages is aligned with the failure that generator tuning fails because the required documents were never retrieved; its causal relevance is that labeled queries and expected evidence distinguish retrieval from synthesis for the situation where stakeholders report hallucinations but cannot isolate search errors.

Incorrect Answers

A:

In a solution where retraining uses revised labels and a different source snapshot, reproducing a candidate evaluation from the recorded dataset could catch the defect that model behavior changes without reproducible training provenance; the current release gate must challenge the failure where generator tuning fails because the required documents were never retrieved.

C:

Although the exercise might catch Copilot relies on last quarter's manually imported account report, it would not directly validate the mechanism that labeled queries and expected evidence distinguish retrieval from synthesis in the target implementation.

D:

The tested behavior is that generated interface components can be grounded in governed app data and defined actions; the necessary evidence here concerns labeled queries and expected evidence distinguish retrieval from synthesis and the previously observed generator tuning fails because the required documents were never retrieved.

Question 1047

Release reviewers will not accept a generic success message. The business scenario is that average accuracy hides failures on complex finance questions, and the historical failure is: Lower-cost routing degrades a business-critical subgroup. Which test is appropriately targeted?

  1. Submit urgent and ambiguous cases to the candidate agent
  2. Replay complex and simple cases through revised thresholds
  3. Execute both approved and rejected concession scenarios
  4. Simulate expected launch concurrency and validate headroom

Explanation

Correct Answer

B:

For the required workflow where average accuracy hides failures on complex finance questions, the test needs to reproduce or prevent the failure where lower-cost routing degrades a business-critical subgroup; replaying complex and simple cases through revised thresholds exercises the path where stratified evaluation guides thresholds for difficult requests.

Incorrect Answers

A:

For a support agent changes how it classifies priority intents, the alternate test is about whether controlled service evaluations validate routing before rollout; it cannot prove the outcome needed after the reported regression where lower-cost routing degrades a business-critical subgroup.

C:

A pass on that verification would speak to service recommends a price adjustment requiring finance approval; it would not show that stratified evaluation guides thresholds for difficult requests, which is needed to avoid the observed defect lower-cost routing degrades a business-critical subgroup.

D:

A pass on that verification would speak to a regional launch multiplies concurrent agent requests; it would not show that stratified evaluation guides thresholds for difficult requests, which is needed to avoid the observed defect lower-cost routing degrades a business-critical subgroup.

Question 1048

Risk reviewers need a control, not merely a monitoring dashboard. The use case is: An intake task requires several prerequisite fields. Potential harm: Missing order fields might be converted into an unvalidated mutation. Which protection fits?

  1. Label unsupported explanations as hypotheses or abstain
  2. Reject action execution until mandatory values are validated
  3. Stop and compensate when a required approval fails
  4. Require human sign-off when regulated language is modified

Explanation

Correct Answer

B:

In the context where an intake task requires several prerequisite fields, the exposure is that missing order fields might be converted into an unvalidated mutation; containing it requires a guardrail in the operation where state tracking captures required inputs before any action.

Incorrect Answers

A:

The safeguard for finance asks why a variance was flagged deals with a different threat: an analytical response might present an unsupported hypothesis as a ledger fact; even after it is enforced, the exposure that missing order fields might be converted into an unvalidated mutation remains possible in the target workflow where state tracking captures required inputs before any action.

C:

That protection is tied to the mechanism that defined flow steps make sequencing and error paths explicit; it does not prevent the distinct risk where missing order fields might be converted into an unvalidated mutation in the system under review.

D:

The alternate guard addresses a model might alter mandatory legal wording in customer correspondence for compliance letters must preserve approved language and disclaimers; the present threat, missing order fields might be converted into an unvalidated mutation, needs a different enforcement boundary.

Question 1049

A design adds autonomous behavior to a process where sales users say deal stage differently across regions. This creates a documented risk: Informal sales-stage wording might trigger a high-value opportunity state change. Which enforced control is the best fit?

  1. Require confirmation before changing high-value opportunity state
  2. Reject or escalate uncertain cross-system matches
  3. Block unclassified high-risk fields at the connector boundary
  4. Honor source access when syncing or federating content

Explanation

Correct Answer

A:

The business setting is sales users say deal stage differently across regions, with the exposure that informal sales-stage wording might trigger a high-value opportunity state change; the enforceable protection must correspond to the path where canonical business terminology maps language to underlying Dynamics records.

Incorrect Answers

B:

The proposed control targets a guessed cross-system account match might attach credit risk to the wrong customer, while the given risk scenario is informal sales-stage wording might trigger a high-value opportunity state change; these exposures arise in different operational contexts.

C:

The alternate guard addresses an ERP connector might expose protected margin columns in its response payload for sales copilots should not expose sensitive margin assumptions; the present threat, informal sales-stage wording might trigger a high-value opportunity state change, needs a different enforcement boundary.

D:

The proposed control targets an external connector might index documents without honoring original ACLs, while the given risk scenario is informal sales-stage wording might trigger a high-value opportunity state change; these exposures arise in different operational contexts.

Question 1050

A solution architecture is being reviewed for this business situation: An AI-driven replenishment proposal influences procurement. The observed defect is that recommendation changes spread to all planners without evaluation. Which architectural decision addresses the underlying requirement?

  1. Plan reversible rollout for supply-chain decision assistance
  2. Apply output filtering and usage restrictions to model interfaces
  3. Test Finance-to-Supply-Chain exception handling after partial failure
  4. Validate finance and operations virtual-entity mappings after change

Explanation

Correct Answer

A:

The architecture must reconcile two facts: an AI-driven replenishment proposal influences procurement, yet recommendation changes spread to all planners without evaluation; the selected pattern does so through the mechanism that a pilot cohort and rollback path isolate operational impact.

Incorrect Answers

B:

A client can request arbitrary portions of training examples; the alternate design can address that situation because safety controls and authorization checks limit misuse and disclosure, whereas this question needs a solution to recommendation changes spread to all planners without evaluation.

C:

In the case where a blocked invoice triggers a hold on related procurement actions, the alternate design depends on failure-injection tests verify retries and compensation between systems; it does not correct the target failure, recommendation changes spread to all planners without evaluation.

D:

The technical behavior of this alternative is that integration tests protect virtual-entity schema and source semantics; the user case being assessed is an AI-driven replenishment proposal influences procurement with a different failure: recommendation changes spread to all planners without evaluation.

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