Question 290
One enterprise rollout has 2 separately testable requirements. (1) Business use: A regulated workload must keep data in permitted locations. Known defect: An external model call exports sensitive prompts to an unapproved region. (2) Business use: A Copilot Studio agent can read corporate data and use public messaging. Known defect: A permitted connector combination enables uncontrolled disclosure. Which 2 architectural choices directly satisfy these stated cases? Choose TWO.
- Map all agent data flows to approved geographic boundaries
- Deploy actions with compatible rollback behavior
- Parameterize environment-specific connection references and variables
- Apply data loss prevention policies to connector combinations
Explanation
Correct Answers
A:For the independent requirement that a regulated workload must keep data in permitted locations, data-flow architecture identifies processing, storage and telemetry locations; it addresses the defect an external model call exports sensitive prompts to an unapproved region, while a Copilot Studio agent can read corporate data and use public messaging needs its own architecture.
D:For the independent requirement that a Copilot Studio agent can read corporate data and use public messaging, Power Platform data policies constrain connector use and movement pathways; it addresses the defect a permitted connector combination enables uncontrolled disclosure, while a regulated workload must keep data in permitted locations needs its own architecture.
Incorrect Answers
B:This design is associated with the different failure where rolling back the agent leaves an incompatible backend action in a workflow where a new action version changes how cases are updated; it does not directly solve the listed conditions: an external model call exports sensitive prompts to an unapproved region; a permitted connector combination enables uncontrolled disclosure.
C:This design is associated with the different failure where agent writes real customer data into the sandbox in a workflow where a flow moved to production still points to a test endpoint; it does not directly solve the listed conditions: an external model call exports sensitive prompts to an unapproved region; a permitted connector combination enables uncontrolled disclosure.
Question 291
A solution architecture is being reviewed for this business situation: A model is replaced without a traceable change ticket. The observed defect is that investigators cannot explain when behavior changed. Which architectural decision addresses the underlying requirement?
- Audit privileged agent actions with correlation to user intent
- Trigger retraining from validated data drift rather than a fixed guess
- Restrict who can supply tuning data and approve model updates
- Record immutable model deployment events with approver and version
Explanation
Correct Answer
D:The defect is that investigators cannot explain when behavior changed; the expected business case is that a model is replaced without a traceable change ticket; the design fits because deployment audit links actor, approved artifact and release time.
Incorrect Answers
A:The alternative addresses logs show deletion but not the requesting principal or approval by a design in which end-to-end audit binds user, tool, affected record and authorization event; it does not supply the missing relationship for a model is replaced without a traceable change ticket, where investigators cannot explain when behavior changed.
B:The alternative addresses team retrains monthly despite unchanged model behavior by a design in which drift monitoring and outcomes justify retraining when performance shifts; it does not supply the missing relationship for a model is replaced without a traceable change ticket, where investigators cannot explain when behavior changed.
C:In the case where business analysts can submit examples but not change production behavior, the alternate design depends on separated data-curation and release permissions preserve integrity; it does not correct the target failure, investigators cannot explain when behavior changed.
Question 292
An incident review has one confirmed symptom: Shared admin credentials return restricted cases to ordinary users. For the use case where agent looks up records on behalf of employees with different rights, which targeted diagnostic check would reveal the most pertinent evidence?
- Correlate conversation request IDs with business API audit entries
- Trace token acquisition and authorization results by environment
- Review endpoint access logs and allowed network paths
- Compare backend access decisions for different users
Explanation
Correct Answer
D:The reported fault is that shared admin credentials return restricted cases to ordinary users in a system where agent looks up records on behalf of employees with different rights; evidence from comparing backend access decisions for different users can show whether delegated authorization evaluates each caller's effective permissions.
Incorrect Answers
A:Correlating conversation request ids with business api audit entries explores a separate connection, where end-to-end audit binds user, tool, affected record and authorization event; the diagnosed requirement is agent looks up records on behalf of employees with different rights, and the actual defect is shared admin credentials return restricted cases to ordinary users.
B:That probe evaluates production consent, identity and connection setup are distinct release checks in a workflow where an imported agent can authenticate in dev but not production; the reported failure concerns the path for which delegated authorization evaluates each caller's effective permissions instead.
C:A sensitive model endpoint is reachable using a broadly shared key; its matching diagnostic check is reasonable for untrusted clients submit requests and exfiltrate outputs; the present investigation needs evidence for the different failure where shared admin credentials return restricted cases to ordinary users.
Question 293
A pilot scorecard needs evidence of business improvement rather than only invocation counts. The use case is: A pilot reports gross labor savings but omits operating expenses. Known defect: The investment case overstates value by ignoring review and hosting. Which KPI should be selected?
- Realized annualized net benefit versus validated baseline
- Median elapsed time from design decision to approval
- Monthly run-rate per active automated workflow
- First-wave use cases meeting readiness gates
Explanation
Correct Answer
A:Because net benefit compares realized outcomes with implementation and recurring costs, the outcome worth following is realized annualized net benefit versus validated baseline; the metric is tied to the reported issue—the investment case overstates value by ignoring review and hosting—for the scenario where a pilot reports gross labor savings but omits operating expenses.
Incorrect Answers
B:The proposed metric concerns business, platform, security and legal teams make conflicting AI decisions, whereas the failure in this case is that the investment case overstates value by ignoring review and hosting; the indicator needed here is realized annualized net benefit versus validated baseline.
C:A change in monthly run-rate per active automated workflow is informative for an agent scales from a trial to enterprise usage; it does not establish whether the current process achieves realized annualized net benefit versus validated baseline.
D:The premise of this indicator is readiness assessment places safe high-value workloads ahead of risky ones; the architecture being evaluated instead serves a pilot reports gross labor savings but omits operating expenses, and its success is measured by realized annualized net benefit versus validated baseline.
Question 294
In a business situation where a pilot reports gross labor savings but omits operating expenses, a previous build showed this defect: The investment case overstates value by ignoring review and hosting. Which reproducible test belongs in the release gate?
- Ask for a case field and verify the exact authorized record
- Compare prioritized roadmap with a risk-adjusted benefit matrix
- Recalculate the case under higher review time and lower adoption
- Run a representative extraction holdout at production concurrency
Explanation
Correct Answer
C:The check is targeted rather than generic: net benefit compares realized outcomes with implementation and recurring costs; its pass/fail evidence applies to the workflow where a pilot reports gross labor savings but omits operating expenses and to the prior defect that the investment case overstates value by ignoring review and hosting.
Incorrect Answers
A:The tested behavior is that Dataverse grounding exposes appropriate structured records to the agent; the necessary evidence here concerns net benefit compares realized outcomes with implementation and recurring costs and the previously observed the investment case overstates value by ignoring review and hosting.
B:The suggested test serves candidate processes are attractive but permissions and data quality are weak, in which the first-wave pilot requests broad access to unreliable records; the present scenario instead concerns the reported problem where the investment case overstates value by ignoring review and hosting.
D:Running a representative extraction holdout at production concurrency verifies an adjacent objective where high-volume narrow extraction needs predictable latency and cost; it leaves untested the causal path in which net benefit compares realized outcomes with implementation and recurring costs for this case.
Question 295
Before production promotion, the team needs evidence for this setting: Support tickets reveal an unusual multistep conversation failure. A prior regression caused the following: The fixed scenario is not added to future test suites. Which validation is most persuasive?
- Reproduce a reported answer from recorded non-sensitive context
- Reproduce a sanitized failure and confirm the new regression catches it
- Run the same corpus in a CI release candidate
- Execute both approved and rejected concession scenarios
Explanation
Correct Answer
B:The scenario is that support tickets reveal an unusual multistep conversation failure; the defect to rule out is that the fixed scenario is not added to future test suites; testing via reproducing a sanitized failure and confirming the new regression catches it challenges the design dependency where feedback-derived examples convert incidents into lasting safeguards.
Incorrect Answers
A:Reproducing a reported answer from recorded non-sensitive context verifies an adjacent objective where users report stale answers after a policy refresh; it leaves untested the causal path in which feedback-derived examples convert incidents into lasting safeguards for this case.
C:Running the same corpus in a ci release candidate verifies an adjacent objective where prompt revisions cause intermittent regressions; it leaves untested the causal path in which feedback-derived examples convert incidents into lasting safeguards for this case.
D:Executing both approved and rejected concession scenarios verifies an adjacent objective where service recommends a price adjustment requiring finance approval; it leaves untested the causal path in which feedback-derived examples convert incidents into lasting safeguards for this case.
Question 296
Security and business owners want to keep this capability: A candidate appears accurate on its training examples. They must also prevent the following: A team might promote a tuned model without meeting a critical holdout threshold. Which guardrail should the design incorporate?
- Sandbox high-risk interactions and require approvals
- Require predefined acceptance thresholds before promotion
- Block release when critical evidence is consistently omitted
- Protect recordings and redact sensitive transcript elements
Explanation
Correct Answer
B:Because a team might promote a tuned model without meeting a critical holdout threshold, a guardrail must constrain the actual system behavior: separation of training and evaluation yields realistic generalization checks; protecting a response alone cannot enforce that boundary.
Incorrect Answers
A:A legacy portal has no integration API for a clerical workflow; its guardrail addresses the possibility that browser-based computer use might click a destructive UI control unexpectedly; the present architecture must constrain the different threat that a team might promote a tuned model without meeting a critical holdout threshold.
C:The proposed control targets a release might proceed despite consistently missing the required evidence passage, while the given risk scenario is a team might promote a tuned model without meeting a critical holdout threshold; these exposures arise in different operational contexts.
D: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, a team might promote a tuned model without meeting a critical holdout threshold, needs a different enforcement boundary.
Question 297
A design adds autonomous behavior to a process where business, platform, security and legal teams make conflicting AI decisions. This creates a documented risk: Cross-function objections might be overruled without an accountable approval decision. Which enforced control is the best fit?
- Separate recommendation privileges from posting permissions
- Keep a known-good model and agent configuration for rollback
- Require review for breaking changes and publish migration windows
- Require risk sign-off from accountable stakeholders
- Add human approval for consequential recommendations
Explanation
Correct Answer
D:Business, platform, security and legal teams make conflicting AI decisions; its risk scenario is that cross-function objections might be overruled without an accountable approval decision; the safeguard fits because a named strategy team assigns decision rights and resolves cross-function tradeoffs and the restriction addresses that specific exposure.
Incorrect Answers
A:Another process could need this guardrail because an AI recommendation could become an actual posting without an authorized reviewer; the design being assessed supports business, platform, security and legal teams make conflicting AI decisions and has the explicit exposure that cross-function objections might be overruled without an accountable approval decision.
B:The alternate guard addresses a new agent release could regress behavior across the entire employee base for new behavior must be evaluated without exposing the full workforce; the present threat, cross-function objections might be overruled without an accountable approval decision, needs a different enforcement boundary.
C:That protection is tied to the mechanism that versioned schemas and compatibility policy make interfaces predictable; it does not prevent the distinct risk where cross-function objections might be overruled without an accountable approval decision in the system under review.
E:The alternate boundary mitigates an automated recommendation might be executed without human accountability, which is relevant to some cases need cross-document reasoning and detailed justification; it leaves the current workflow vulnerable to cross-function objections might be overruled without an accountable approval decision.
Question 298
An agent-assisted service does not behave as specified: The agent can invoke an unintended high-privilege operation. It is used in this situation: Agent instructions mention many APIs but only two are permitted. Which observable diagnostic procedure is best?
- Audit rollout settings and incident timelines for uncontrolled changes
- Review available tool inventory versus approved actions
- Document specific requirements unmet by available platforms
- Compare feature gaps and vendor operating obligations
Explanation
Correct Answer
B:In the setting where agent instructions mention many APIs but only two are permitted, reviewing available tool inventory versus approved actions can isolate the failure in which the agent can invoke an unintended high-privilege operation, because allowlisting and server-enforced identity restrict actual executable capability.
Incorrect Answers
A:A prompt update changes decisions for all users at once; examining that other path does not determine whether allowlisting and server-enforced identity restrict actual executable capability, which is what distinguishes the issue in the stated use case.
C:That probe evaluates custom engineering trades more control for more operating responsibility in a workflow where competitive requirements depend on unique multi-model control; the reported failure concerns the path for which allowlisting and server-enforced identity restrict actual executable capability instead.
D:The proposed inspection is associated with the requirement is common and product coverage is adequate, where managed functionality shifts routine platform operations to the service; it would not expose the failure mode that the agent can invoke an unintended high-privilege operation with the required specificity.
Question 299
A pilot scorecard needs evidence of business improvement rather than only invocation counts. The use case is: The system cannot always find evidence for regulated guidance. Known defect: An agent guesses a policy instead of escalating. Which KPI should be selected?
- Unauthorized tool invocation success rate
- High-impact actions executed without required approval
- Unsupported regulated answers reaching users
- Schema-valid response rate
Explanation
Correct Answer
C:Because confidence and evidence checks route insufficiently supported cases to humans, the outcome worth following is unsupported regulated answers reaching users; the metric is tied to the reported issue—an agent guesses a policy instead of escalating—for the scenario where the system cannot always find evidence for regulated guidance.
Incorrect Answers
A:A change in unauthorized tool invocation success rate is informative for agent instructions mention many APIs but only two are permitted; it does not establish whether the current process achieves unsupported regulated answers reaching users.
B:For the workflow where teams propose that an agent perform both low-risk lookups and financial postings, high-impact actions executed without required approval is an actionable measure; it does not establish whether an agent guesses a policy instead of escalating in the situation under review.
D:Schema-valid response rate. That indicator describes a distinct workflow where agents return inconsistent fields for downstream integrations; it does not quantify unsupported regulated answers reaching users, which is needed for the stated case.
Question 300
A staging test has to detect the specific defect that an update changes a field mapping and grounded answers become stale. In the workflow where AI assistance reads operational records through Dataverse, which exercise verifies the corrective behavior?
- Interrupt processing after the first app commits
- Execute both approved and rejected concession scenarios
- Modify a safe test purchase order and verify the mapped output
- Try accessing a restricted training corpus from an unrelated project
Explanation
Correct Answer
C:An update changes a field mapping and grounded answers become stale; a release test must establish that this no longer occurs; modifying a safe test purchase order and verifying the mapped output measures the specific behavior for which integration tests protect virtual-entity schema and source semantics.
Incorrect Answers
A:For a blocked invoice triggers a hold on related procurement actions, the alternate test is about whether failure-injection tests verify retries and compensation between systems; it cannot prove the outcome needed after the reported regression where an update changes a field mapping and grounded answers become stale.
B:This check exercises a different system behavior: transaction-path tests confirm recommendation, review and posting records; a pass there cannot establish that an update changes a field mapping and grounded answers become stale no longer occurs in the required workflow.
D:For proprietary records are copied to an open test workspace, the alternate test is about whether segregated datasets and access controls protect sensitive data; it cannot prove the outcome needed after the reported regression where an update changes a field mapping and grounded answers become stale.