Question 950
A product owner must measure whether an AI-enabled change actually improves this process: Compliance letters must preserve approved language and disclaimers. The earlier problem was: The agent invents discretionary promises. Which outcome indicator is most relevant?
- Regulated replies with complete mandatory clauses
- Ticket updates completed with attributable credentials
- Prompt action responses conforming to output contract
- Drafts meeting structure and source-accuracy criteria
Explanation
Correct Answer
A:For the use case where compliance letters must preserve approved language and disclaimers, the indicator regulated replies with complete mandatory clauses is worth baselining against the earlier issue where the agent invents discretionary promises.
Incorrect Answers
B:The alternative tracks ticket updates completed with attributable credentials, arising from the mechanism where an authorized connector action invokes an operational service; its movement alone says little about the failure that the agent invents discretionary promises.
C:The premise of this indicator is prompt actions encapsulate reusable inputs and expected outputs; the architecture being evaluated instead serves compliance letters must preserve approved language and disclaimers, and its success is measured by regulated replies with complete mandatory clauses.
D:That scorecard concerns the activity in which marketing specialists need tailored descriptions from existing briefings and its result of drafts meeting structure and source-accuracy criteria; the reported shortcoming here is the agent invents discretionary promises.
Question 951
A test plan focuses on a previously missed failure: A separate application breaks established productivity workflows. It occurred when staff need a role-specific conversation within Copilot and Teams. Which check directly validates that behavior?
- Delay a specialist and confirm timeout and retry behavior
- Install the agent for a pilot audience and test supported experiences
- Test the custom action with existing identity and audit policy
- Prototype the distinguishing behavior under production constraints
Explanation
Correct Answer
B:The historical failure is that a separate application breaks established productivity workflows; for the setting where staff need a role-specific conversation within Copilot and Teams, installing the agent for a pilot audience and testing supported experiences verifies the dependency where declarative agents use host orchestration with tailored instructions and actions.
Incorrect Answers
A:The verification concerning downstream research agents run longer than the interactive request checks task IDs, status transitions, deadlines and durable handoff state support coordination against orchestrator returns a partial result as though the workflow finished; its result says little about the observed issue where a separate application breaks established productivity workflows.
C:The verification concerning current users need one governed custom action inside their workflow checks extension retains familiar user experience and adds bounded custom behavior against full replacement increases retraining and migration risk; its result says little about the observed issue where a separate application breaks established productivity workflows.
D:The suggested test serves competitive requirements depend on unique multi-model control, in which a boxed product cannot express the execution and ownership rules; the present scenario instead concerns the reported problem where a separate application breaks established productivity workflows.
Question 952
A design adds autonomous behavior to a process where agents need structured case summaries at handoff. This creates a documented risk: A case handoff could include personal attributes irrelevant to the receiving team's work. Which enforced control is the best fit?
- Exclude sensitive attributes not needed by the receiving agent
- Confirm identity before sharing protected case details
- Block unclassified high-risk fields at the connector boundary
- Reject or escalate uncertain cross-system matches
Explanation
Correct Answer
A:Avoiding the consequence where a case handoff could include personal attributes irrelevant to the receiving team's work requires a control for agents need structured case summaries at handoff; in this architecture, tailored instructions and contextual fields standardize case handoff content, which is the point at which the boundary can be applied.
Incorrect Answers
B:A voice bot cannot resolve a complex eligibility dispute; its guardrail addresses the possibility that a transferred voice conversation might disclose protected context before verifying identity; the present architecture must constrain the different threat that a case handoff could include personal attributes irrelevant to the receiving team's work.
C:This restriction is associated with sales copilots should not expose sensitive margin assumptions and addresses the exposure that an ERP connector might expose protected margin columns in its response payload; it does not specifically protect the target workflow against a case handoff could include personal attributes irrelevant to the receiving team's work.
D:Where two systems share company names but not primary keys, the guard can address a guessed cross-system account match might attach credit risk to the wrong customer; the exposure that a case handoff could include personal attributes irrelevant to the receiving team's work remains relevant to the present assessment.
Question 953
After deployment, stakeholders need an interpretable measure for this use case: Many teams copy slightly different instructions for the same task. The pilot showed that a successful prompt change cannot be reproduced after rollout. Which metric captures the needed result?
- Prompt regressions detected before release
- Error-rate delta between canary and control cohorts
- Correct classification rate for ambiguous boundary cases
- Reuse rate of registered AI components across business units
Explanation
Correct Answer
A:Prompt regressions detected before release; this outcome measure belongs to the workflow where many teams copy slightly different instructions for the same task; a change in it can help detect whether the defect that a successful prompt change cannot be reproduced after rollout is resolved.
Incorrect Answers
B:The alternative tracks error-rate delta between canary and control cohorts, arising from the mechanism where controlled release gates restrict blast radius and support rollback; its movement alone says little about the failure that a successful prompt change cannot be reproduced after rollout.
C:The alternative tracks correct classification rate for ambiguous boundary cases, arising from the mechanism where representative examples teach decision boundaries without changing underlying facts; its movement alone says little about the failure that a successful prompt change cannot be reproduced after rollout.
D:The proposed metric concerns teams rebuild similar assistants without seeing existing assets, whereas the failure in this case is that a successful prompt change cannot be reproduced after rollout; the indicator needed here is prompt regressions detected before release.
Question 954
A policy review covers an agent for which a specialized support team uses a nonstandard severity matrix. Its most relevant exposure is that automated severity suggestions might override an incident commander's decision. What safeguard should be required before release?
- Limit deployment to approved users and surfaces
- Let authorized incident commanders override recommendations
- Prevent automatic approvals when the credit source is unavailable
- Reject or escalate uncertain cross-system matches
- Block unclassified high-risk fields at the connector boundary
Explanation
Correct Answer
B:Because automated severity suggestions might override an incident commander's decision, a guardrail must constrain the actual system behavior: custom business rules align AI-assisted triage with actual impact definitions; protecting a response alone cannot enforce that boundary.
Incorrect Answers
A:The alternate boundary mitigates a Teams-targeted agent package might be made available on unapproved hosts, which is relevant to agent experience must appear in selected Microsoft 365 contexts; it leaves the current workflow vulnerable to automated severity suggestions might override an incident commander's decision.
C:The safeguard for external credit service responds slowly during regional outages deals with a different threat: credit-dependent approvals might proceed while the live credit source is unavailable; even after it is enforced, the exposure that automated severity suggestions might override an incident commander's decision remains possible in the target workflow where custom business rules align AI-assisted triage with actual impact definitions.
D:For the use case where two systems share company names but not primary keys, the control addresses the threat that a guessed cross-system account match might attach credit risk to the wrong customer; the specified agent must be protected against automated severity suggestions might override an incident commander's decision.
E:Where sales copilots should not expose sensitive margin assumptions, the guard can address an ERP connector might expose protected margin columns in its response payload; the exposure that automated severity suggestions might override an incident commander's decision remains relevant to the present assessment.
Question 955
A design adds autonomous behavior to a process where an agent should summarize account data but not alter it. This creates a documented risk: A read-only assistant might acquire write operations by sharing a tool identity. Which enforced control is the best fit?
- Block protected actions unless the trusted user and backend authorize
- Require approval and a separate action identity for mutations
- Do not share private cached outputs across principals
- Reject requests independent of instruction wording
Explanation
Correct Answer
B:Avoiding the consequence where a read-only assistant might acquire write operations by sharing a tool identity requires a control for an agent should summarize account data but not alter it; in this architecture, distinct permission surfaces enforce least privilege by task, which is the point at which the boundary can be applied.
Incorrect Answers
A:Where a retrieved email contains instructions to bypass approval, the guard can address injected text might persuade a model to invoke a protected API; the exposure that a read-only assistant might acquire write operations by sharing a tool identity remains relevant to the present assessment.
C:The alternate guard addresses a user might receive another principal's cached protected answer for shared result caching makes answers fast across roles; the present threat, a read-only assistant might acquire write operations by sharing a tool identity, needs a different enforcement boundary.
D:This restriction is associated with a user asks the agent to read an unrelated executive record and addresses the exposure that a persuasive jailbreak might cause the backend to reveal privileged records; it does not specifically protect the target workflow against a read-only assistant might acquire write operations by sharing a tool identity.
Question 956
The acceptance board asks how to verify a corrected design. The setting is: Teams rebuild similar assistants without seeing existing assets. Last time, multiple agents offer overlapping tools with inconsistent permissions. Which scenario should the team execute?
- Search the catalogue by task and verify current ownership
- Attempt concurrent updates through both app interfaces
- Compare matched assisted and unassisted case cohorts
- Update a case and verify an authorized sales workflow sees it
Explanation
Correct Answer
A:Searching the catalogue by task and verifying current ownership is aligned with the failure that multiple agents offer overlapping tools with inconsistent permissions; its causal relevance is that discovery and ownership metadata support reuse and controlled deprecation for the situation where teams rebuild similar assistants without seeing existing assets.
Incorrect Answers
B:A pass on that verification would speak to two applications try to overwrite the same customer credit field; it would not show that discovery and ownership metadata support reuse and controlled deprecation, which is needed to avoid the observed defect multiple agents offer overlapping tools with inconsistent permissions.
C:Although the exercise might catch average handling-time savings disappear during rework, it would not directly validate the mechanism that discovery and ownership metadata support reuse and controlled deprecation in the target implementation.
D:The tested behavior is that shared authorized records support coordinated business context; the necessary evidence here concerns discovery and ownership metadata support reuse and controlled deprecation and the previously observed multiple agents offer overlapping tools with inconsistent permissions.
Question 957
One enterprise rollout has 2 separately testable requirements. (1) Business use: Agents modify records through external APIs. Known defect: Successful chat sessions conceal repeated backend writes. (2) Business use: A service bot handles many chats but few cases are resolved. Known defect: A dashboard calls the bot successful because messages increased. Which 2 regression tests directly satisfy these stated cases? Choose TWO.
- Inject timeout after commit and verify a single write
- Delay an approver past the defined execution deadline
- Send concurrent equivalent events and inspect resulting writes
- Test a conversation that ends politely without resolving the case
Explanation
Correct Answers
A:For the business workflow where agents modify records through external APIs, injecting timeout after commit and verifying a single write challenges the failure that successful chat sessions conceal repeated backend writes; the release also needs a separate check for a service bot handles many chats but few cases are resolved.
D:For the business workflow where a service bot handles many chats but few cases are resolved, testing a conversation that ends politely without resolving the case challenges the failure that a dashboard calls the bot successful because messages increased; the release also needs a separate check for agents modify records through external APIs.
Incorrect Answers
B:This regression test is associated with the different failure where timeout creates an abandoned workflow with no owner in a workflow where an autonomous agent relies on slow approvals and delayed tool results; it does not directly solve the listed conditions: successful chat sessions conceal repeated backend writes; a dashboard calls the bot successful because messages increased.
C:This regression test is associated with the different failure where parallel agents each submit a costly replacement in a workflow where several events refer to one business order at once; it does not directly solve the listed conditions: successful chat sessions conceal repeated backend writes; a dashboard calls the bot successful because messages increased.
Question 958
A product owner must measure whether an AI-enabled change actually improves this process: An agent creates an order but reservation of inventory fails. The earlier problem was: Half-complete transactions leave operational state inconsistent. Which outcome indicator is most relevant?
- Correct topic selection for closely related user intents
- Tool operations executed with valid delegated identity
- Eligible exception events creating one investigation
- Inconsistent workflow states after failed multi-step runs
Explanation
Correct Answer
D:The observed shortcoming is that half-complete transactions leave operational state inconsistent; measuring inconsistent workflow states after failed multi-step runs gives business owners a directly relevant indicator for the process where an agent creates an order but reservation of inventory fails.
Incorrect Answers
A:For the workflow where two topics claim nearly identical scheduling intents, correct topic selection for closely related user intents is an actionable measure; it does not establish whether half-complete transactions leave operational state inconsistent in the situation under review.
B:The proposed metric concerns several agents need the same inventory query operation, whereas the failure in this case is that half-complete transactions leave operational state inconsistent; the indicator needed here is inconsistent workflow states after failed multi-step runs.
C:The proposed metric concerns shipment exceptions should initiate investigation without a chat message, whereas the failure in this case is that half-complete transactions leave operational state inconsistent; the indicator needed here is inconsistent workflow states after failed multi-step runs.
Question 959
The operating team must monitor whether an architectural correction works. Business context: A business process mixes deterministic calculations and open-ended judgments. Historical symptom: A single agent unpredictably handles policy rules and narrative summarization. Which outcome measure is the best match?
- Share of failed workflows with an identified responsible component
- Conflicting cross-application writes per thousand transactions
- Unauthorized tool invocation success rate
- Proportion of tasks assigned to an explicit execution pattern
Explanation
Correct Answer
D:The use case involves a business process mixes deterministic calculations and open-ended judgments and the baseline failure that a single agent unpredictably handles policy rules and narrative summarization; the KPI tracks proportion of tasks assigned to an explicit execution pattern, so improvement has an interpretable operational meaning.
Incorrect Answers
A:For the situation where copilot, Dataverse, Dynamics and custom APIs share one process, share of failed workflows with an identified responsible component is a sensible indicator; the problem that a single agent unpredictably handles policy rules and narrative summarization requires measurement of proportion of tasks assigned to an explicit execution pattern.
B:For the situation where two applications try to overwrite the same customer credit field, conflicting cross-application writes per thousand transactions is a sensible indicator; the problem that a single agent unpredictably handles policy rules and narrative summarization requires measurement of proportion of tasks assigned to an explicit execution pattern.
C:The proposed metric concerns agent instructions mention many APIs but only two are permitted, whereas the failure in this case is that a single agent unpredictably handles policy rules and narrative summarization; the indicator needed here is proportion of tasks assigned to an explicit execution pattern.
Question 960
The rollout supports this workflow: Many calls require only a small fixed-label output. Leaders want a metric that responds to the reported shortcoming that universal use of a premium model inflates cost. Which measure best reflects success?
- Fully loaded cost per correctly resolved case
- Successful requests during primary endpoint failures
- First-wave use cases meeting readiness gates
- Cost per accepted classification at required accuracy
Explanation
Correct Answer
D:For the use case where many calls require only a small fixed-label output, the indicator cost per accepted classification at required accuracy is worth baselining against the earlier issue where universal use of a premium model inflates cost.
Incorrect Answers
A:The proposed metric concerns AI drafts requests quickly but specialists must correct some outputs, whereas the failure in this case is that universal use of a premium model inflates cost; the indicator needed here is cost per accepted classification at required accuracy.
B:Although successful requests during primary endpoint failures is observable, the affected process involves many calls require only a small fixed-label output; its acceptance outcome is cost per accepted classification at required accuracy, not the alternative measurement.
C:Candidate processes are attractive but permissions and data quality are weak; its intended measurement is first-wave use cases meeting readiness gates; substituting that result here would not capture whether universal use of a premium model inflates cost has been corrected.