Microsoft AB-100 (Agentic AI Business Solutions Architect) Practice Test 111

Question 1100

A risk analysis highlights a threat in this workflow: An AI-driven replenishment proposal influences procurement. Exposure: Autonomous replenishment proposals might place purchases beyond approved limits. Which enforceable safeguard addresses it at the relevant boundary?

  1. Do not allow autonomous purchasing without approved limits
  2. Stage compatible transitions and protect authoritative fields
  3. Do not approve procurement while financial hold status is unknown
  4. Prevent sensitive data from entering prompts unless required

Explanation

Correct Answer

A:

When autonomous replenishment proposals might place purchases beyond approved limits, the use case an AI-driven replenishment proposal influences procurement requires a guardrail aligned with the behavior that a pilot cohort and rollback path isolate operational impact.

Incorrect Answers

B:

The alternate boundary mitigates cross-Dynamics releases might overwrite authoritative customer fields, which is relevant to a customer handoff depends on shared record and term mappings; it leaves the current workflow vulnerable to autonomous replenishment proposals might place purchases beyond approved limits.

C:

Another process could need this guardrail because a procurement approval might proceed before a finance hold is resolved; the design being assessed supports an AI-driven replenishment proposal influences procurement and has the explicit exposure that autonomous replenishment proposals might place purchases beyond approved limits.

D:

The proposed control targets a model might expose sensitive user data in uncontrolled prompt output, while the given risk scenario is autonomous replenishment proposals might place purchases beyond approved limits; these exposures arise in different operational contexts.

Question 1101

After deployment, stakeholders need an interpretable measure for this use case: A tuned model claims higher accuracy but adds latency. The pilot showed that direct replacement removes the dependable baseline. Which metric captures the needed result?

  1. Model accuracy under agreed input-quality stress levels
  2. Critical releases with reproducible evaluation and sign-off records
  3. Candidate promotions meeting multimetric acceptance criteria
  4. Deployed models with complete registry evidence

Explanation

Correct Answer

C:

The use case involves a tuned model claims higher accuracy but adds latency and the baseline failure that direct replacement removes the dependable baseline; the KPI tracks candidate promotions meeting multimetric acceptance criteria, so improvement has an interpretable operational meaning.

Incorrect Answers

A:

Model accuracy under agreed input-quality stress levels; that indicator describes a distinct workflow where production documents differ from clean training samples; it does not quantify candidate promotions meeting multimetric acceptance criteria, which is needed for the stated case.

B:

For the workflow where an updated agent performs differently months after release, critical releases with reproducible evaluation and sign-off records is an actionable measure; it does not establish whether direct replacement removes the dependable baseline in the situation under review.

D:

That scorecard concerns the activity in which data scientists cannot explain why a newer classifier behaves differently and its result of deployed models with complete registry evidence; the reported shortcoming here is direct replacement removes the dependable baseline.

Question 1102

One enterprise rollout has 2 separately testable requirements. (1) Business use: Work can proceed from events without human initiation but needs bounded autonomy. Known defect: A background agent has triggered irreversible duplicate actions. (2) Business use: Employees express several intents and need interactive clarification. Known defect: A rigid menu fails whenever a request contains two intents. Which 2 architectural choices directly satisfy these stated cases? Choose TWO.

  1. Define autonomy tiers by consequence and reversibility
  2. Use explicit handoff contracts for asynchronous specialist work
  3. Use a Copilot Studio conversational agent for variable natural-language requests
  4. Use an autonomous agent only for bounded reversible tasks

Explanation

Correct Answers

C:

For the independent requirement that employees express several intents and need interactive clarification, generative orchestration interprets intent and can combine authorized tools and knowledge; it addresses the defect a rigid menu fails whenever a request contains two intents, while work can proceed from events without human initiation but needs bounded autonomy needs its own architecture.

D:

For the independent requirement that work can proceed from events without human initiation but needs bounded autonomy, event triggers and tool permissions can automate predefined safe steps within explicit limits; it addresses the defect a background agent has triggered irreversible duplicate actions, while employees express several intents and need interactive clarification needs its own architecture.

Incorrect Answers

A:

This design is associated with the different failure where the same agent can read an invoice and commit a payment unchecked in a workflow where teams propose that an agent perform both low-risk lookups and financial postings; it does not directly solve the listed conditions: a background agent has triggered irreversible duplicate actions; a rigid menu fails whenever a request contains two intents.

B:

This design is associated with the different failure where orchestrator returns a partial result as though the workflow finished in a workflow where downstream research agents run longer than the interactive request; it does not directly solve the listed conditions: a background agent has triggered irreversible duplicate actions; a rigid menu fails whenever a request contains two intents.

Question 1103

An incident review has one confirmed symptom: An external model call exports sensitive prompts to an unapproved region. For the use case where a regulated workload must keep data in permitted locations, which targeted diagnostic check would reveal the most pertinent evidence?

  1. Trace model version to source manifest and transformation hash
  2. Match release manifest to transformation and model checksums
  3. Reconcile active model versions with promotion events
  4. Inspect endpoint regions and connected-service data flow records

Explanation

Correct Answer

D:

The reported fault is that an external model call exports sensitive prompts to an unapproved region in a system where a regulated workload must keep data in permitted locations; evidence from inspecting endpoint regions and connected-service data flow records can show whether data-flow architecture identifies processing, storage and telemetry locations.

Incorrect Answers

A:

For the failure that no one can identify which dataset produced a model, tracing model version to source manifest and transformation hash is relevant; this question instead reports that an external model call exports sensitive prompts to an unapproved region, so different evidence is needed.

B:

That probe evaluates versioned data transformations and model artifacts enable consistent rollback in a workflow where a preprocessing release changes category labels incorrectly; the reported failure concerns the path for which data-flow architecture identifies processing, storage and telemetry locations instead.

C:

Investigators cannot explain when behavior changed; examining that other path does not determine whether data-flow architecture identifies processing, storage and telemetry locations, which is what distinguishes the issue in the stated use case.

Question 1104

Security and business owners want to keep this capability: Agent experience must appear in selected Microsoft 365 contexts. They must also prevent the following: A Teams-targeted agent package might be made available on unapproved hosts. Which guardrail should the design incorporate?

  1. Fail closed on unresolved caller identity
  2. Limit deployment to approved users and surfaces
  3. Provide accessible alternatives and honor channel policy
  4. Stop and compensate when a required approval fails

Explanation

Correct Answer

B:

In the context where agent experience must appear in selected Microsoft 365 contexts, the exposure is that a Teams-targeted agent package might be made available on unapproved hosts; containing it requires a guardrail in the operation where host-specific support, packaging and admin approval shape deployment.

Incorrect Answers

A:

For the use case where anonymous callers can ask generic questions but not change orders, the control addresses the threat that an anonymous caller might execute an account mutation after providing only a guessed ID; the specified agent must be protected against a Teams-targeted agent package might be made available on unapproved hosts.

C:

That protection is tied to the mechanism that channel-aware response design adapts wording and modality; it does not prevent the distinct risk where a Teams-targeted agent package might be made available on unapproved hosts in the system under review.

D:

Where approval processing requires sequential verification and notification, the guard can address a flow might notify the customer even though a required approval failed; the exposure that a Teams-targeted agent package might be made available on unapproved hosts remains relevant to the present assessment.

Question 1105

A release must prove that a replaced model package introduces unknown behavior will not recur. The application is used in this situation: Deployment automation pulls model artifacts from unverified origins. Which acceptance test most directly checks that failure mode?

  1. Promote a candidate in staging and inspect generated audit entries
  2. Reconstruct one released model's source and threshold decisions
  3. Replace a candidate hash in staging and verify rejection
  4. Submit a malicious mislabeled training record and inspect review

Explanation

Correct Answer

C:

A replaced model package introduces unknown behavior; a release test must establish that this no longer occurs; replacing a candidate hash in staging and verifying rejection measures the specific behavior for which artifact provenance and integrity checks guard model supply chain.

Incorrect Answers

A:

Although the exercise might catch investigators cannot explain when behavior changed, it would not directly validate the mechanism that artifact provenance and integrity checks guard model supply chain in the target implementation.

B:

For data scientists cannot explain why a newer classifier behaves differently, the alternate test is about whether a model registry ties versions to lineage, evaluations and approval; it cannot prove the outcome needed after the reported regression where a replaced model package introduces unknown behavior.

D:

The suggested test serves business analysts can submit examples but not change production behavior, in which unvetted labeled data poisons a retrained classifier; the present scenario instead concerns the reported problem where a replaced model package introduces unknown behavior.

Question 1106

The acceptance board asks how to verify a corrected design. The setting is: Sales and Finance roles have different record entitlements. Last time, a Copilot suggestion includes restricted invoice details. Which scenario should the team execute?

  1. Alternate ambiguous and corrected identifiers within one session
  2. Simulate rollback during a nonproduction close cycle
  3. Re-run identical agent requests with distinct permissions
  4. Modify a safe test purchase order and verify the mapped output

Explanation

Correct Answer

C:

The historical failure is that a Copilot suggestion includes restricted invoice details; for the setting where sales and Finance roles have different record entitlements, re-runing identical agent requests with distinct permissions verifies the dependency where cross-role tests verify that authorization survives integration.

Incorrect Answers

A:

The verification concerning agent must remember order context while asking for missing details checks multi-turn tests assess state tracking, entity separation and follow-up questions against it substitutes a prior customer's order in a later turn; its result says little about the observed issue where a Copilot suggestion includes restricted invoice details.

B:

This check exercises a different system behavior: coordinated release windows account for transaction criticality; a pass there cannot establish that a Copilot suggestion includes restricted invoice details no longer occurs in the required workflow.

D:

The suggested test serves AI assistance reads operational records through Dataverse, in which an update changes a field mapping and grounded answers become stale; the present scenario instead concerns the reported problem where a Copilot suggestion includes restricted invoice details.

Question 1107

An operational risk could arise because real customer details might leak into evaluation corpora derived from tickets. The AI system supports the business need that support tickets reveal an unusual multistep conversation failure. Which guardrail is designed for this threat?

  1. Redact customer data before making examples available to developers
  2. Remove personal details before using feedback as test data
  3. Require a tested human escalation path for unmatched cases
  4. Forbid direct mutation before recorded delegated approval
  5. Roll back a routing change when a critical segment regresses

Explanation

Correct Answer

B:

Avoiding the consequence where real customer details might leak into evaluation corpora derived from tickets requires a control for support tickets reveal an unusual multistep conversation failure; in this architecture, feedback-derived examples convert incidents into lasting safeguards, which is the point at which the boundary can be applied.

Incorrect Answers

A:

The alternate guard addresses user transcripts pasted into developer tickets might disclose customer details for users report stale answers after a policy refresh; the present threat, real customer details might leak into evaluation corpora derived from tickets, needs a different enforcement boundary.

C:

The proposed control targets a new support agent might route urgent cases to a general queue, while the given risk scenario is real customer details might leak into evaluation corpora derived from tickets; these exposures arise in different operational contexts.

D:

The alternate guard addresses a suggested customer concession might post before delegated finance approval for service recommends a price adjustment requiring finance approval; the present threat, real customer details might leak into evaluation corpora derived from tickets, needs a different enforcement boundary.

E:

Another process could need this guardrail because a model router change might silently degrade the highest-risk task segment; the design being assessed supports support tickets reveal an unusual multistep conversation failure and has the explicit exposure that real customer details might leak into evaluation corpora derived from tickets.

Question 1108

Business leaders reject an activity-only metric for this process: Agents and apps need shared definitions of customers and work orders. They need to detect the issue where different solutions apply incompatible status and customer fields. Which performance indicator fits?

  1. Percentage of agent metrics matching reconciled source calculations
  2. Time from source exception to downstream acknowledgement
  3. Share of failed workflows with an identified responsible component
  4. Cross-application entity consistency rate

Explanation

Correct Answer

D:

This scorecard needs an outcome related to agents and apps need shared definitions of customers and work orders; cross-application entity consistency rate. That indicator can be checked against the earlier symptom that different solutions apply incompatible status and customer fields.

Incorrect Answers

A:

A change in percentage of agent metrics matching reconciled source calculations is informative for sales and finance agents must agree on revenue and margin calculations; it does not establish whether the current process achieves cross-application entity consistency rate.

B:

Although time from source exception to downstream acknowledgement is observable, the affected process involves agents and apps need shared definitions of customers and work orders; its acceptance outcome is cross-application entity consistency rate, not the alternative measurement.

C:

Copilot, Dataverse, Dynamics and custom APIs share one process; its intended measurement is share of failed workflows with an identified responsible component; substituting that result here would not capture whether different solutions apply incompatible status and customer fields has been corrected.

Question 1109

An incident review has one confirmed symptom: Average handling-time savings disappear during rework. For the use case where AI drafts requests quickly but specialists must correct some outputs, which targeted diagnostic check would reveal the most pertinent evidence?

  1. Correlate throttling errors with fallback activation
  2. Time sample assisted cases including corrective work
  3. Segment failures by route, token cost and difficulty
  4. Analyze route decisions by task class and observed error

Explanation

Correct Answer

B:

The reported fault is that average handling-time savings disappear during rework in a system where AI drafts requests quickly but specialists must correct some outputs; evidence from timing sample assisted cases including corrective work can show whether end-to-end unit cost includes review, escalation and retry effort.

Incorrect Answers

A:

In the process where peak load can saturate a primary model endpoint, the diagnostic procedure could explain requests fail despite another approved deployment being available; this particular incident instead turns on whether end-to-end unit cost includes review, escalation and retry effort and why average handling-time savings disappear during rework.

C:

The proposed inspection is associated with real task mix differs from prelaunch performance tests, where feedback-driven thresholds rebalance accuracy, latency and spend; it would not expose the failure mode that average handling-time savings disappear during rework with the required specificity.

D:

Many calls require only a small fixed-label output; its matching diagnostic check is reasonable for universal use of a premium model inflates cost; the present investigation needs evidence for the different failure where average handling-time savings disappear during rework.

Question 1110

A product owner must measure whether an AI-enabled change actually improves this process: Sellers need current credit exposure held outside CRM. The earlier problem was: Copilot relies on last quarter's manually imported account report. Which outcome indicator is most relevant?

  1. Sales task completion under external API impairment
  2. Fresh authorized account insights from external systems
  3. Handoffs with complete approved case context
  4. Correct identification of governed service intents

Explanation

Correct Answer

B:

The use case involves sellers need current credit exposure held outside CRM and the baseline failure that Copilot relies on last quarter's manually imported account report; the KPI tracks fresh authorized account insights from external systems, so improvement has an interpretable operational meaning.

Incorrect Answers

A:

The premise of this indicator is bounded timeouts and graceful fallback preserve essential user work; the architecture being evaluated instead serves sellers need current credit exposure held outside CRM, and its success is measured by fresh authorized account insights from external systems.

C:

Agents need structured case summaries at handoff; its intended measurement is handoffs with complete approved case context; substituting that result here would not capture whether Copilot relies on last quarter's manually imported account report has been corrected.

D:

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 relies on last quarter's manually imported account report in the situation under review.

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