Question 370
A policy review covers an agent for which general models repeatedly misread rare engineering labels. Its most relevant exposure is that a custom model might be trained on unlicensed or unrepresentative documents. What safeguard should be required before release?
- Validate source access after ranking and before response
- Block use without representative training rights and validation
- Block release when critical evidence is consistently omitted
- Require user review before committing suggested classifications
Explanation
Correct Answer
B:The business setting is general models repeatedly misread rare engineering labels, with the exposure that a custom model might be trained on unlicensed or unrepresentative documents; the enforceable protection must correspond to the path where domain-trained models can be evaluated when generic baselines fall short.
Incorrect Answers
A:The proposed control targets a reranker might return a relevant passage that the user cannot access, while the given risk scenario is a custom model might be trained on unlicensed or unrepresentative documents; these exposures arise in different operational contexts.
C:This restriction is associated with stakeholders report hallucinations but cannot isolate search errors and addresses the exposure that a release might proceed despite consistently missing the required evidence passage; it does not specifically protect the target workflow against a custom model might be trained on unlicensed or unrepresentative documents.
D:Inspectors submit a short description and need suggested categorization; its guardrail addresses the possibility that aI-assisted classifications might be saved without the inspector's review; the present architecture must constrain the different threat that a custom model might be trained on unlicensed or unrepresentative documents.
Question 371
Security and business owners want to keep this capability: A representative must initiate an exception workflow from the case. They must also prevent the following: Repeated agent calls might create duplicate escalation records or excess privileges. Which guardrail should the design incorporate?
- Limit creation privileges and prevent duplicate escalations
- Require explicit consent for consequential create or update operations
- Block unclassified high-risk fields at the connector boundary
- Reject or escalate uncertain cross-system matches
Explanation
Correct Answer
A:In the scenario where a representative must initiate an exception workflow from the case, the explicit risk is that repeated agent calls might create duplicate escalation records or excess privileges; enforcement must follow the mechanism by which an authorized action creates a governed workflow item from the current case.
Incorrect Answers
B:The safeguard for a Teams user should query and update a supported external service deals with a different threat: a plugin might create business records without user consent or approval; even after it is enforced, the exposure that repeated agent calls might create duplicate escalation records or excess privileges remains possible in the target workflow where an authorized action creates a governed workflow item from the current case.
C:The proposed control targets an ERP connector might expose protected margin columns in its response payload, while the given risk scenario is repeated agent calls might create duplicate escalation records or excess privileges; these exposures arise in different operational contexts.
D:That protection is tied to the mechanism that a mapping service resolves stable identities before enrichment; it does not prevent the distinct risk where repeated agent calls might create duplicate escalation records or excess privileges in the system under review.
Question 372
Security and business owners want to keep this capability: An updated agent performs differently months after release. They must also prevent the following: Old evaluation evidence might be overwritten to conceal a failed release gate. Which guardrail should the design incorporate?
- Require reviewed datasets and independent promotion sign-off
- Limit retroactive alteration of approved QA evidence
- Prevent mixed-version retrieval without compatibility checks
- Restrict promotion to reviewed compatible configurations
Explanation
Correct Answer
B:Avoiding the consequence where old evaluation evidence might be overwritten to conceal a failed release gate requires a control for an updated agent performs differently months after release; in this architecture, versioned evaluation sets, model IDs and thresholds document release claims, which is the point at which the boundary can be applied.
Incorrect Answers
A:That protection is tied to the mechanism that separated data-curation and release permissions preserve integrity; it does not prevent the distinct risk where old evaluation evidence might be overwritten to conceal a failed release gate in the system under review.
C:Another process could need this guardrail because mixed embedding or index versions might corrupt retrieval quality; the design being assessed supports an updated agent performs differently months after release and has the explicit exposure that old evaluation evidence might be overwritten to conceal a failed release gate.
D:The proposed control targets an agent release might pair instructions with incompatible tool or model versions, while the given risk scenario is old evaluation evidence might be overwritten to conceal a failed release gate; these exposures arise in different operational contexts.
Question 373
A product owner must measure whether an AI-enabled change actually improves this process: Callers provide account numbers and amounts by speech. The earlier problem was: Transcription confusion risks wrong account modifications. Which outcome indicator is most relevant?
- Critical voice actions with verified entity confirmation
- Fresh authorized account insights from external systems
- Correctly completed legacy UI tasks without unsafe side effects
- Out-of-policy payments blocked prior to posting
Explanation
Correct Answer
A:Critical voice actions with verified entity confirmation; this outcome measure belongs to the workflow where callers provide account numbers and amounts by speech; a change in it can help detect whether the defect that transcription confusion risks wrong account modifications is resolved.
Incorrect Answers
B:The proposed metric concerns sellers need current credit exposure held outside CRM, whereas the failure in this case is that transcription confusion risks wrong account modifications; the indicator needed here is critical voice actions with verified entity confirmation.
C:A legacy portal has no integration API for a clerical workflow; its intended measurement is correctly completed legacy UI tasks without unsafe side effects; substituting that result here would not capture whether transcription confusion risks wrong account modifications has been corrected.
D:Although out-of-policy payments blocked prior to posting is observable, the affected process involves callers provide account numbers and amounts by speech; its acceptance outcome is critical voice actions with verified entity confirmation, not the alternative measurement.
Question 374
Governance approves the business purpose but requires control of this exposure: An API might accept a valid-looking but unauthorized generated identifier. The agent will be used when the orchestrator chooses between similar invoice and credit tools. Which boundary is most defensible?
- Validate types and authorization at the called service
- Remove unneeded attributes before model invocation
- Protect recordings and redact sensitive transcript elements
- Treat sentiment as advisory rather than a sole eligibility decision
Explanation
Correct Answer
A:When an API might accept a valid-looking but unauthorized generated identifier, the use case the orchestrator chooses between similar invoice and credit tools requires a guardrail aligned with the behavior that specific tool metadata helps the planner choose and parameterize actions.
Incorrect Answers
B:A canvas app passes customer records to a model action; its guardrail addresses the possibility that a canvas app might send unnecessary protected fields into an inference prompt; the present architecture must constrain the different threat that an API might accept a valid-looking but unauthorized generated identifier.
C:The alternate guard addresses recorded customer speech or transcript fields might be exposed to unintended viewers for quality teams need searchable insights from recorded calls; the present threat, an API might accept a valid-looking but unauthorized generated identifier, needs a different enforcement boundary.
D:The alternate guard addresses a sentiment score might become the only basis for an eligibility or disciplinary decision for the system must detect customer sentiment and named entities at scale; the present threat, an API might accept a valid-looking but unauthorized generated identifier, needs a different enforcement boundary.
Question 375
Before introducing a fix, the team must explain why rewriting the entire conversational layer adds little value. The affected setting is: Standard conversation needs a proprietary scoring service. What should be inspected?
- Trace latency and authorization across the managed/custom boundary
- Review available tool inventory versus approved actions
- Audit decision state across sales and finance transactions
- Check whether the required API can be surfaced through supported actions
Explanation
Correct Answer
A:Rewriting the entire conversational layer adds little value; to attribute the defect rather than guess at a fix, tracing latency and authorization across the managed/custom boundary examines the dependency where a managed host can invoke an independently governed domain API.
Incorrect Answers
B:For the failure that the agent can invoke an unintended high-privilege operation, reviewing available tool inventory versus approved actions is relevant; this question instead reports that rewriting the entire conversational layer adds little value, so different evidence is needed.
C:For the failure that a suggested discount bypasses delegated approval authority, auditing decision state across sales and finance transactions is relevant; this question instead reports that rewriting the entire conversational layer adds little value, so different evidence is needed.
D:That probe evaluates supported extensibility adds an action while retaining the host's orchestration in a workflow where users need familiar Copilot behavior plus an internal status lookup; the reported failure concerns the path for which a managed host can invoke an independently governed domain API instead.
Question 376
The architecture board has evidence that former owners leave behind accessible service accounts. The affected situation is: Unused agents retain tokens and active integrations. What change most directly establishes the missing behavior?
- Design a documented agent retirement and offboarding process
- Use progressive deployment and rollback for a stable agent endpoint
- Maintain an enterprise inventory of production agents and owners
- Use managed solutions for governed downstream environments
Explanation
Correct Answer
A:Because decommissioning revokes credentials, connectors and publishing rights, this approach corrects the reported breakdown where former owners leave behind accessible service accounts; the affected situation is that unused agents retain tokens and active integrations.
Incorrect Answers
B:For another type of architecture, canary traffic and versioned rollback reduce release risk is useful; this design must instead support unused agents retain tokens and active integrations despite the problem where former owners leave behind accessible service accounts.
C:This design would make sense if business units create agents with no central visibility, particularly where security cannot identify which agents use protected connectors; the requested architecture instead needs the dependency that decommissioning revokes credentials, connectors and publishing rights because former owners leave behind accessible service accounts.
D:The technical behavior of this alternative is that managed solution deployment supports controlled downstream customization boundaries; the user case being assessed is unused agents retain tokens and active integrations with a different failure: former owners leave behind accessible service accounts.
Question 377
One enterprise rollout has 2 separately testable requirements. (1) Business use: An agent should summarize account data but not alter it. Existing shortcoming: Retrieval configuration accidentally grants write capability. (2) Business use: Prompt wording persuades the agent to submit a privileged operation. Existing shortcoming: API trusts the model's text assertion of permission. Which 2 success metrics directly satisfy these stated cases? Choose TWO.
- Connector schema regressions caught before production
- Unauthorized write operations executed by read-only agents
- Unsupported claims emitted in low-evidence tests
- Privileged mutations without verified principal authorization
Explanation
Correct Answers
B:The outcome unauthorized write operations executed by read-only agents measures whether the workflow where an agent should summarize account data but not alter it is improving; the scorecard separately needs privileged mutations without verified principal authorization.
D:The outcome privileged mutations without verified principal authorization measures whether the workflow where prompt wording persuades the agent to submit a privileged operation is improving; the scorecard separately needs unauthorized write operations executed by read-only agents.
Incorrect Answers
A:The alternative measures connector schema regressions caught before production for the situation where an external API changes a required field used by an agent; it cannot substitute for performance outcomes in the separate named cases: an agent should summarize account data but not alter it; prompt wording persuades the agent to submit a privileged operation.
C:The alternative measures unsupported claims emitted in low-evidence tests for the situation where policy answers must be tied to a current source; it cannot substitute for performance outcomes in the separate named cases: an agent should summarize account data but not alter it; prompt wording persuades the agent to submit a privileged operation.
Question 378
A solution architecture is being reviewed for this business situation: Service agents draft customer messages but cannot send without review. The observed defect is that new prompt automation bypasses supervisor approval. Which architectural decision addresses the underlying requirement?
- Test audit completeness for AI-assisted customer concessions
- Parameterize environment-specific connection references and variables
- Preserve customer communication approvals across solution releases
- Regression-test case routing after service agent changes
- Package Copilot Studio agent components in a Power Platform solution
Explanation
Correct Answer
C:Without the intended behavior, new prompt automation bypasses supervisor approval; the proposed architecture establishes the specific design principle that business approval checks remain independent of agent version for the process where service agents draft customer messages but cannot send without review.
Incorrect Answers
A:The alternative addresses the final order lacks traceability to the AI suggestion and approver by a design in which transaction-path tests confirm recommendation, review and posting records; it does not supply the missing relationship for service agents draft customer messages but cannot send without review, where new prompt automation bypasses supervisor approval.
B:Environment configuration separates deployable logic from endpoint bindings; its main purpose is a flow moved to production still points to a test endpoint, not correction of the reported defect where new prompt automation bypasses supervisor approval in the target process.
D:There is a valid place for the mechanism that controlled service evaluations validate routing before rollout, namely a support agent changes how it classifies priority intents; it does not remove the specific defect that new prompt automation bypasses supervisor approval from this case.
E:Manual recreation causes mismatched topics and actions; the alternate design can address that situation because solution-based ALM transports supported components under controlled releases, whereas this question needs a solution to new prompt automation bypasses supervisor approval.
Question 379
Before production promotion, the team needs evidence for this setting: A unique domain vocabulary produces persistent measured errors. A prior regression caused the following: Teams propose custom training without any baseline. Which validation is most persuasive?
- Run offline evaluations of the trained candidate against holdouts
- Prototype a unique scheduling dependency with the custom runtime
- Prototype the distinguishing behavior under production constraints
- Run a pilot with a comparable holdout group
Explanation
Correct Answer
A:The check is targeted rather than generic: a measured capability gap, appropriate dataset and MLOps plan justify customization; its pass/fail evidence applies to the workflow where a unique domain vocabulary produces persistent measured errors and to the prior defect that teams propose custom training without any baseline.
Incorrect Answers
B:The suggested test serves a workflow needs proprietary scheduling, model choice and runtime control, in which a predefined host cannot meet the required orchestration contract; the present scenario instead concerns the reported problem where teams propose custom training without any baseline.
C:The verification concerning competitive requirements depend on unique multi-model control checks custom engineering trades more control for more operating responsibility against a boxed product cannot express the execution and ownership rules; its result says little about the observed issue where teams propose custom training without any baseline.
D:This check exercises a different system behavior: matched comparisons separate solution impact from unrelated shifts; a pass there cannot establish that teams propose custom training without any baseline no longer occurs in the required workflow.
Question 380
After deployment, stakeholders need an interpretable measure for this use case: Independent teams choose inconsistent evaluation and security methods. The pilot showed that one division releases agents without risk review. Which metric captures the needed result?
- Interoperable tasks completed without custom point-to-point parsing
- Fully loaded cost per correctly resolved case
- Percentage of agent releases passing common governance gates
- Reuse rate of registered AI components across business units
Explanation
Correct Answer
C:The observed shortcoming is that one division releases agents without risk review; measuring percentage of agent releases passing common governance gates gives business owners a directly relevant indicator for the process where independent teams choose inconsistent evaluation and security methods.
Incorrect Answers
A:A change in interoperable tasks completed without custom point-to-point parsing is informative for independently hosted agents must exchange tasks and results; it does not establish whether the current process achieves percentage of agent releases passing common governance gates.
B:That scorecard concerns the activity in which AI drafts requests quickly but specialists must correct some outputs and its result of fully loaded cost per correctly resolved case; the reported shortcoming here is one division releases agents without risk review.
D:The alternative tracks reuse rate of registered AI components across business units, arising from the mechanism where discovery and ownership metadata support reuse and controlled deprecation; its movement alone says little about the failure that one division releases agents without risk review.