Question 20
A policy review covers an agent for which contact Center agents sometimes cannot resolve complex billing intent. Its most relevant exposure is that a caller might get trapped in repeated fallback without a safe escalation route. What safeguard should be required before release?
- Require explicit consent for consequential create or update operations
- Preserve conversation context and avoid infinite re-prompting
- Limit creation privileges and prevent duplicate escalations
- Limit tool availability and confirmations for risky steps
Explanation
Correct Answer
B:In the scenario where contact Center agents sometimes cannot resolve complex billing intent, the explicit risk is that a caller might get trapped in repeated fallback without a safe escalation route; enforcement must follow the mechanism by which confidence, retry ceilings and escalation routing enable resolution.
Incorrect Answers
A:For the use case where a Teams user should query and update a supported external service, the control addresses the threat that a plugin might create business records without user consent or approval; the specified agent must be protected against a caller might get trapped in repeated fallback without a safe escalation route.
C:For the use case where a representative must initiate an exception workflow from the case, the control addresses the threat that repeated agent calls might create duplicate escalation records or excess privileges; the specified agent must be protected against a caller might get trapped in repeated fallback without a safe escalation route.
D:The alternate boundary mitigates a generative planner might chain tools that exceed the approved workflow scope, which is relevant to users request a status check and an authorized follow-up action; it leaves the current workflow vulnerable to a caller might get trapped in repeated fallback without a safe escalation route.
Question 21
A quality engineer is writing a regression case after the incident where mean response time looks acceptable. The underlying business scenario is: A small portion of calls time out during peak demand. What should the test exercise?
- Compare controlled tests before and after tool curation
- Load-test burst traffic and compare latency percentiles
- Roll out to one region and check segmented resolution metrics
- Send an identical holdout to champion and challenger
Explanation
Correct Answer
B:Load-testing burst traffic and comparing latency percentiles is aligned with the failure that mean response time looks acceptable; its causal relevance is that tail percentiles surface extreme experiences affecting users for the situation where a small portion of calls time out during peak demand.
Incorrect Answers
A:This check exercises a different system behavior: clear capabilities and restricted sets improve selection reliability; a pass there cannot establish that mean response time looks acceptable no longer occurs in the required workflow.
C:The verification concerning a conversational change affects only one support geography checks cohort telemetry supports safe staged validation against global averages hide a regional escalation spike; its result says little about the observed issue where mean response time looks acceptable.
D:A pass on that verification would speak to a tuned model claims higher accuracy but adds latency; it would not show that tail percentiles surface extreme experiences affecting users, which is needed to avoid the observed defect mean response time looks acceptable.
Question 22
Troubleshooting starts from a business complaint: Finance sees savings without the downstream rework cost. The affected process is: A new model reduces price but also degrades exact extractions. Which investigation is most directly tied to the reported defect?
- Review threshold breaches against user-impacting failure times
- Compare release report hashes with preserved test artifacts
- Diff representative before-and-after activity maps
- Compare cost per accepted output across model versions
Explanation
Correct Answer
D:Finance sees savings without the downstream rework cost; to attribute the defect rather than guess at a fix, comparing cost per accepted output across model versions examines the dependency where joint quality and consumption measures capture real efficiency.
Incorrect Answers
A:In the separate scenario where operations must distinguish healthy traffic from quality incidents, the same investigation could explain teams discover failures through complaints several days later; it does not isolate the reported cause in a system where a new model reduces price but also degrades exact extractions.
B:The proposed probe concerns an updated agent performs differently months after release, where versioned evaluation sets, model IDs and thresholds document release claims; this incident instead requires evidence about joint quality and consumption measures capture real efficiency and the failure that finance sees savings without the downstream rework cost.
C:A new topic description changes which agent is invoked; its matching diagnostic check is reasonable for regression appears only for compound user requests; the present investigation needs evidence for the different failure where finance sees savings without the downstream rework cost.
Question 23
A proposed deployment supports this business case: A newly deployed Foundry agent has no detailed diagnostics. Reviewers also warn that tracing might store secrets and personal content without adequate access controls. Which restriction mitigates that concrete risk?
- Mask sensitive input/output fields and control log access
- Require a tested human escalation path for unmatched cases
- Use idempotency on any repeatable mutation action
- Define completion against system-of-record changes where relevant
Explanation
Correct Answer
A:When tracing might store secrets and personal content without adequate access controls, the use case a newly deployed Foundry agent has no detailed diagnostics requires a guardrail aligned with the behavior that connected observability supplies traces and operational metrics.
Incorrect Answers
B:The alternate guard addresses a new support agent might route urgent cases to a general queue for a support agent changes how it classifies priority intents; the present threat, tracing might store secrets and personal content without adequate access controls, needs a different enforcement boundary.
C:The safeguard for agents modify records through external APIs deals with a different threat: retries might cause repeated transactions even though tool calls appear successful; even after it is enforced, the exposure that tracing might store secrets and personal content without adequate access controls remains possible in the target workflow where connected observability supplies traces and operational metrics.
D:That protection is tied to the mechanism that outcome measures connect agent use with completed business work; it does not prevent the distinct risk where tracing might store secrets and personal content without adequate access controls in the system under review.
Question 24
Support engineers report this failure: Submission screens freeze when inference fails. The business context is: Field staff must keep working during a model outage. What evidence should be collected to isolate the cause?
- Trace dependency failures and app completion paths
- Inspect field-level source and suggestion labels
- Trace page data bindings and invoked record APIs
- Record recall at K and citation-grounding errors separately
Explanation
Correct Answer
A:For the situation where field staff must keep working during a model outage, the defect is that submission screens freeze when inference fails; tracing dependency failures and app completion paths helps verify the cause-and-effect path: timeouts and graceful degradation protect core app operations.
Incorrect Answers
B:The evidence this method produces applies to a workflow where users need to distinguish verified fields from AI suggestions; it cannot by itself establish why submission screens freeze when inference fails in the current scenario.
C:The proposed inspection is associated with staff need a contextual workspace assembled around one customer case, where generated interface components can be grounded in governed app data and defined actions; it would not expose the failure mode that submission screens freeze when inference fails with the required specificity.
D:Stakeholders report hallucinations but cannot isolate search errors; its matching diagnostic check is reasonable for generator tuning fails because the required documents were never retrieved; the present investigation needs evidence for the different failure where submission screens freeze when inference fails.
Question 25
The architecture board has evidence that the agent fabricates an answer during a search outage. The affected situation is: Required knowledge stores are intermittently unreachable. What change most directly establishes the missing behavior?
- Check source availability and fallback when grounding is unavailable
- Ground conversational answers in SharePoint knowledge with permissions
- Route simple classification requests to a validated inexpensive model
- Synchronize an indexed source on an explicit freshness schedule
Explanation
Correct Answer
A:Without the intended behavior, the agent fabricates an answer during a search outage; the proposed architecture establishes the specific design principle that health checks and explicit no-evidence responses handle missing context safely for the process where required knowledge stores are intermittently unreachable.
Incorrect Answers
B:For employee policies reside in governed Microsoft 365 documents, this other approach relies on the fact that supported SharePoint grounding retrieves organizational content within access controls; the issue here is the agent fabricates an answer during a search outage; that mechanism does not establish the required health checks and explicit no-evidence responses handle missing context safely.
C:When many calls require only a small fixed-label output, task-aware routing chooses economical models after quality gates; in the stated workflow, however, the agent fabricates an answer during a search outage; the relevant architectural behavior is that health checks and explicit no-evidence responses handle missing context safely.
D:For another type of architecture, incremental refresh and source-to-index lag monitoring reduce staleness is useful; this design must instead support required knowledge stores are intermittently unreachable despite the problem where the agent fabricates an answer during a search outage.
Question 26
A quality engineer is writing a regression case after the incident where teams propose browser clicks despite an available stable API. The underlying business scenario is: A legacy portal has no integration API for a clerical workflow. What should the test exercise?
- Show a conflicting suggestion beside the authoritative value
- Change page layout in a test environment and observe recovery
- Score unseen cases representing each critical segment
- Escalate a multi-step voice conversation and check context continuity
Explanation
Correct Answer
B:Changing page layout in a test environment and observing recovery challenges the path where uI automation can operate application interfaces but requires stability checks for the business setting a legacy portal has no integration API for a clerical workflow; the specific defect to exclude is that teams propose browser clicks despite an available stable API.
Incorrect Answers
A:A pass on that verification would speak to users need to distinguish verified fields from AI suggestions; it would not show that uI automation can operate application interfaces but requires stability checks, which is needed to avoid the observed defect teams propose browser clicks despite an available stable API.
C:Scoring unseen cases representing each critical segment verifies an adjacent objective where a candidate appears accurate on its training examples; it leaves untested the causal path in which uI automation can operate application interfaces but requires stability checks for this case.
D:Escalating a multi-step voice conversation and checking context continuity verifies an adjacent objective where a voice bot cannot resolve a complex eligibility dispute; it leaves untested the causal path in which uI automation can operate application interfaces but requires stability checks for this case.
Question 27
An incident review has one confirmed symptom: Invalid record IDs produce updates to the wrong business object. For the use case where the model submits malformed identifiers to an operational API, which targeted diagnostic check would reveal the most pertinent evidence?
- Interview sponsors to locate missing success criteria
- Inspect action failures for bad types and entity mismatches
- Inspect write conflicts and source-of-truth definitions
- Audit rollout settings and incident timelines for uncontrolled changes
- Document specific requirements unmet by available platforms
Explanation
Correct Answer
B:For the situation where the model submits malformed identifiers to an operational API, the defect is that invalid record IDs produce updates to the wrong business object; inspecting action failures for bad types and entity mismatches helps verify the cause-and-effect path: typed inputs and server validation reject malformed or out-of-scope calls.
Incorrect Answers
A:Prototype activity increases but no business value is measured; examining that other path does not determine whether typed inputs and server validation reject malformed or out-of-scope calls, which is what distinguishes the issue in the stated use case.
C:For the failure that parallel agent updates create conflicting authoritative values, inspecting write conflicts and source-of-truth definitions is relevant; this question instead reports that invalid record IDs produce updates to the wrong business object, so different evidence is needed.
D:A prompt update changes decisions for all users at once; examining that other path does not determine whether typed inputs and server validation reject malformed or out-of-scope calls, which is what distinguishes the issue in the stated use case.
E:Documenting specific requirements unmet by available platforms explores a separate connection, where custom engineering trades more control for more operating responsibility; the diagnosed requirement is the model submits malformed identifiers to an operational API, and the actual defect is invalid record IDs produce updates to the wrong business object.
Question 28
An operational risk could arise because an owned custom runtime might lack the host's identity and safety governance. The AI system supports the business need that the solution demands its own model selection and hosting policies. Which guardrail is designed for this threat?
- Operate independent identity, safety and observability controls
- Scope connector permissions to specific business operations
- Apply privacy review and model-version rollback before release
- Use approved extension points rather than unsupported modifications
Explanation
Correct Answer
A:The solution demands its own model selection and hosting policies; its risk scenario is that an owned custom runtime might lack the host's identity and safety governance; the safeguard fits because custom-engine architecture controls runtime, model calls and integrations and the restriction addresses that specific exposure.
Incorrect Answers
B:Where users need familiar Copilot behavior plus an internal status lookup, the guard can address an extension could receive broad write permissions beyond its single new action; the exposure that an owned custom runtime might lack the host's identity and safety governance remains relevant to the present assessment.
C:A unique domain vocabulary produces persistent measured errors; its guardrail addresses the possibility that customization might promote a model trained on data without a privacy review; the present architecture must constrain the different threat that an owned custom runtime might lack the host's identity and safety governance.
D:Where current users need one governed custom action inside their workflow, the guard can address an unsupported product customization might break after vendor updates; the exposure that an owned custom runtime might lack the host's identity and safety governance remains relevant to the present assessment.
Question 29
One enterprise rollout has 2 separately testable requirements. (1) Business use: A request spans procurement, finance and policy expertise. Existing shortcoming: One general agent conflates expertise and may call the wrong tools. (2) Business use: Downstream research agents run longer than the interactive request. Existing shortcoming: Orchestrator returns a partial result as though the workflow finished. Which 2 success metrics directly satisfy these stated cases? Choose TWO.
- Task completion without switching application context
- Custom workflow completion against the required execution contract
- Cross-domain tasks routed to the correct specialist
- Percentage of delegated tasks completed within agreed deadlines
Explanation
Correct Answers
C:The outcome cross-domain tasks routed to the correct specialist measures whether the workflow where a request spans procurement, finance and policy expertise is improving; the scorecard separately needs percentage of delegated tasks completed within agreed deadlines.
D:The outcome percentage of delegated tasks completed within agreed deadlines measures whether the workflow where downstream research agents run longer than the interactive request is improving; the scorecard separately needs cross-domain tasks routed to the correct specialist.
Incorrect Answers
A:The alternative measures task completion without switching application context for the situation where users must remain in a proprietary work-order application; it cannot substitute for performance outcomes in the separate named cases: a request spans procurement, finance and policy expertise; downstream research agents run longer than the interactive request.
B:The alternative measures custom workflow completion against the required execution contract for the situation where the solution demands its own model selection and hosting policies; it cannot substitute for performance outcomes in the separate named cases: a request spans procurement, finance and policy expertise; downstream research agents run longer than the interactive request.
Question 30
Governance approves the business purpose but requires control of this exposure: An unapproved change to the approval matrix could silently alter authorization decisions. The agent will be used when the task follows fixed approval branches with no interpretation needed. Which boundary is most defensible?
- Deny unapproved tools regardless of model-generated arguments
- Reject malformed output before writing business records
- Validate sending-agent identity and delegated authorization
- Require policy-owner approval for any rule change
Explanation
Correct Answer
D:The task follows fixed approval branches with no interpretation needed; its risk scenario is that an unapproved change to the approval matrix could silently alter authorization decisions; the safeguard fits because explicit conditions execute known business rules predictably and leave an auditable branch history and the restriction addresses that specific exposure.
Incorrect Answers
A:The alternate boundary mitigates planner-generated arguments might invoke tools that the user was never permitted to call, which is relevant to agent instructions mention many APIs but only two are permitted; it leaves the current workflow vulnerable to an unapproved change to the approval matrix could silently alter authorization decisions.
B:The alternate boundary mitigates invalid generated JSON might be written directly into a business record, which is relevant to agents return inconsistent fields for downstream integrations; it leaves the current workflow vulnerable to an unapproved change to the approval matrix could silently alter authorization decisions.
C:That protection is tied to the mechanism that an agent interoperability protocol supports explicit task delegation between agents; it does not prevent the distinct risk where an unapproved change to the approval matrix could silently alter authorization decisions in the system under review.