Question 320
The pilot needs a durable correction: Logs cannot reconstruct which API mutation followed an answer. This affects the workflow in which a request crosses Copilot Studio, a flow and the CRM backend. Which design choice should the architect approve?
- Connect Foundry agent tracing to Application Insights
- Define end-to-end correlation across agent and business APIs
- Measure tool success, retries and side-effect duplication
- Test Finance-to-Supply-Chain exception handling after partial failure
Explanation
Correct Answer
B:The mechanism that correlation identifiers link the distributed operational path fits the scenario where a request crosses Copilot Studio, a flow and the CRM backend; it addresses the defect that logs cannot reconstruct which API mutation followed an answer.
Incorrect Answers
A:In the case where operations need to correlate prompts, model calls and tools in a Foundry workflow, the alternate design depends on agent trace spans provide component-level execution evidence through connected telemetry; it does not correct the target failure, logs cannot reconstruct which API mutation followed an answer.
C:When agents modify records through external APIs, tool metrics reveal mutation failures and retry consequences; in the stated workflow, however, logs cannot reconstruct which API mutation followed an answer; the relevant architectural behavior is that correlation identifiers link the distributed operational path.
D:The alternative addresses a network fault leaves the two apps inconsistent by a design in which failure-injection tests verify retries and compensation between systems; it does not supply the missing relationship for a request crosses Copilot Studio, a flow and the CRM backend, where logs cannot reconstruct which API mutation followed an answer.
Question 321
A release must prove that historical traces remain available longer than approved retention will not recur. The application is used in this situation: Debugging logs include prompts containing customer data. Which acceptance test most directly checks that failure mode?
- Promote to staging and confirm correct connection targets
- Create a synthetic trace and verify planned expiry behavior
- Pilot a connector with synthetic data and review destinations
- Attempt to publish agents from two risk zones
Explanation
Correct Answer
B:Creating a synthetic trace and verifying planned expiry behavior is aligned with the failure that historical traces remain available longer than approved retention; its causal relevance is that retention schedules and minimization control telemetry exposure for the situation where debugging logs include prompts containing customer data.
Incorrect Answers
A:The suggested test serves a flow moved to production still points to a test endpoint, in which agent writes real customer data into the sandbox; the present scenario instead concerns the reported problem where historical traces remain available longer than approved retention.
C:The verification concerning employees install external agents with unknown data pathways checks procurement and security review evaluate permissions and data flows against an unreviewed plugin exports sensitive tenant information; its result says little about the observed issue where historical traces remain available longer than approved retention.
D:A pass on that verification would speak to personal read-only assistants and payment agents have different risk; it would not show that retention schedules and minimization control telemetry exposure, which is needed to avoid the observed defect historical traces remain available longer than approved retention.
Question 322
The pilot needs a durable correction: It autonomously contacts a supplier without approval. This affects the workflow in which a bot only needs to check warranty and prepare a draft. Which design choice should the architect approve?
- Use explicit state tracking for multi-step task completion
- Choose bounded task-specific instructions over open-ended autonomy
- Enforce identity confirmation before account-specific actions
- Use a structured response contract for downstream consumption
Explanation
Correct Answer
B:Because limited intent and tool scope prevent unnecessary actions, this approach corrects the reported breakdown where it autonomously contacts a supplier without approval; the affected situation is that a bot only needs to check warranty and prepare a draft.
Incorrect Answers
A:The technical behavior of this alternative is that state tracking captures required inputs before any action; the user case being assessed is a bot only needs to check warranty and prepare a draft with a different failure: it autonomously contacts a supplier without approval.
C:The technical behavior of this alternative is that authentication and authorization gates separate general help from account control; the user case being assessed is a bot only needs to check warranty and prepare a draft with a different failure: it autonomously contacts a supplier without approval.
D:For another type of architecture, structured outputs define required fields and allowable values is useful; this design must instead support a bot only needs to check warranty and prepare a draft despite the problem where it autonomously contacts a supplier without approval.
Question 323
Before production promotion, the team needs evidence for this setting: The same agent is offered in voice and asynchronous chat. A prior regression caused the following: Voice responses contain unreadable links and lengthy tables. Which validation is most persuasive?
- Request the same account help by voice and chat
- Attempt authorized and denied inventory calls through the client
- Test empty and malformed input and validate response shape
- Test edge cases with conflicting documents and incomplete facts
Explanation
Correct Answer
A:The scenario is that the same agent is offered in voice and asynchronous chat; the defect to rule out is that voice responses contain unreadable links and lengthy tables; testing via requesting the same account help by voice and chat challenges the design dependency where channel-aware response design adapts wording and modality.
Incorrect Answers
B:Attempting authorized and denied inventory calls through the client verifies an adjacent objective where several agents need the same inventory query operation; it leaves untested the causal path in which channel-aware response design adapts wording and modality for this case.
C:The verification concerning a summarizer sends output into a ticket ingestion service checks structured outputs define required fields and allowable values against free-form paragraphs break required schema parsing; its result says little about the observed issue where voice responses contain unreadable links and lengthy tables.
D:The verification concerning procurement exceptions require comparing several policies and records checks reasoning can plan through evidence and tools within enforced boundaries against a simple scripted topic cannot handle legitimate variants; its result says little about the observed issue where voice responses contain unreadable links and lengthy tables.
Question 324
Risk reviewers need a control, not merely a monitoring dashboard. The use case is: Benchmarks cover only cooperative user wording. Potential harm: Malformed calls might pass a happy-path suite yet mutate records. Which protection fits?
- Reject requests independent of instruction wording
- Remove personal details before using feedback as test data
- Fail closed on authorization and schema-validation failures
- Block promotion when any mandatory critical threshold fails
- Use idempotency keys and locking where the workflow requires it
Explanation
Correct Answer
C:When malformed calls might pass a happy-path suite yet mutate records, the use case benchmarks cover only cooperative user wording requires a guardrail aligned with the behavior that negative tests expose policy and validation boundaries.
Incorrect Answers
A:The alternate boundary mitigates a persuasive jailbreak might cause the backend to reveal privileged records, which is relevant to a user asks the agent to read an unrelated executive record; it leaves the current workflow vulnerable to malformed calls might pass a happy-path suite yet mutate records.
B:That protection is tied to the mechanism that feedback-derived examples convert incidents into lasting safeguards; it does not prevent the distinct risk where malformed calls might pass a happy-path suite yet mutate records in the system under review.
D:The proposed control targets a candidate might pass overall accuracy while failing a mandatory critical metric, while the given risk scenario is malformed calls might pass a happy-path suite yet mutate records; these exposures arise in different operational contexts.
E:The alternate guard addresses simultaneous events might produce duplicate financial side effects for several events refer to one business order at once; the present threat, malformed calls might pass a happy-path suite yet mutate records, needs a different enforcement boundary.
Question 325
Business leaders reject an activity-only metric for this process: The same agent must move from development to test and production. They need to detect the issue where manual recreation causes mismatched topics and actions. Which performance indicator fits?
- Production calls routed to approved production dependencies
- Critical cases assigned to the correct escalation group
- Connector schema regressions caught before production
- Environment promotions without configuration drift
Explanation
Correct Answer
D:For the use case where the same agent must move from development to test and production, the indicator environment promotions without configuration drift is worth baselining against the earlier issue where manual recreation causes mismatched topics and actions.
Incorrect Answers
A:The proposed metric concerns a flow moved to production still points to a test endpoint, whereas the failure in this case is that manual recreation causes mismatched topics and actions; the indicator needed here is environment promotions without configuration drift.
B:Although critical cases assigned to the correct escalation group is observable, the affected process involves the same agent must move from development to test and production; its acceptance outcome is environment promotions without configuration drift, not the alternative measurement.
C:For the workflow where an external API changes a required field used by an agent, connector schema regressions caught before production is an actionable measure; it does not establish whether manual recreation causes mismatched topics and actions in the situation under review.
Question 326
An incident review has one confirmed symptom: The test suite spends most cases on low-risk FAQs. For the use case where agent supports refunds, order status and generic FAQs, which targeted diagnostic check would reveal the most pertinent evidence?
- Inspect test distribution against business consequence tiers
- Compare generated tests to authentication and schema contracts
- Review the signed acceptance matrix and results
- Review ranking against incident impact and affected populations
Explanation
Correct Answer
A:The reported fault is that the test suite spends most cases on low-risk FAQs in a system where agent supports refunds, order status and generic FAQs; evidence from inspecting test distribution against business consequence tiers can show whether risk-weighting prioritizes high-impact actions and failure paths.
Incorrect Answers
B:That probe evaluates specification-guided generation covers missing fields and denied roles in a workflow where agents can call APIs with customer IDs and amount values; the reported failure concerns the path for which risk-weighting prioritizes high-impact actions and failure paths instead.
C:That probe evaluates predefined gates weigh accuracy, latency and business risk in a workflow where teams promote a custom model because it feels better; the reported failure concerns the path for which risk-weighting prioritizes high-impact actions and failure paths instead.
D:Product owners favor easy fixes over high-consequence failures; its matching diagnostic check is reasonable for a rare data leak risk remains below cosmetic response issues; the present investigation needs evidence for the different failure where the test suite spends most cases on low-risk FAQs.
Question 327
A solution architecture is being reviewed for this business situation: Hostile instructions arrive both as user input and retrieved documents. The observed defect is that the agent is exposed to repeated jailbreak patterns. Which architectural decision addresses the underlying requirement?
- Test prompt changes on paraphrases and edge conditions
- Apply output filtering and usage restrictions to model interfaces
- Use Prompt Shields or equivalent intervention-point screening
- Constrain high-risk tools with action-level policy checks
Explanation
Correct Answer
C:The mechanism that input and document-attack detection supplies layered protection fits the scenario where hostile instructions arrive both as user input and retrieved documents; it addresses the defect that the agent is exposed to repeated jailbreak patterns.
Incorrect Answers
A:For another type of architecture, regression sets cover lexical variation and task boundaries is useful; this design must instead support hostile instructions arrive both as user input and retrieved documents despite the problem where the agent is exposed to repeated jailbreak patterns.
B:In the case where an internal model sometimes returns confidential text verbatim, the alternate design depends on safety controls and authorization checks limit misuse and disclosure; it does not correct the target failure, the agent is exposed to repeated jailbreak patterns.
D:Model-level refusal alone may not prevent a tool call; the alternate design can address that situation because execution-time destination and permission checks block exfiltration, whereas this question needs a solution to the agent is exposed to repeated jailbreak patterns.
Question 328
An agent-assisted service does not behave as specified: Copilot cannot find approved wiki pages by natural language. It is used in this situation: Workers need internal wiki articles from outside Microsoft 365. Which observable diagnostic procedure is best?
- Compare advertised tools with backend permissions
- Compare prompt versions used by dependent apps
- Trace latency distributions and retries for connector calls
- Validate indexed records and source ACL mappings
Explanation
Correct Answer
D:Where Copilot cannot find approved wiki pages by natural language, the useful evidence comes from validating indexed records and source ACL mappings; its focus matches the architectural mechanism that supported connectors expose external content with discoverability and security metadata in the process where workers need internal wiki articles from outside Microsoft 365.
Incorrect Answers
A:Several agents need the same inventory query operation; its matching diagnostic check is reasonable for each agent builds a different unaudited HTTP wrapper; the present investigation needs evidence for the different failure where Copilot cannot find approved wiki pages by natural language.
B:For the failure that each app drifts to a different unreviewed AI prompt, comparing prompt versions used by dependent apps is relevant; this question instead reports that Copilot cannot find approved wiki pages by natural language, so different evidence is needed.
C:The proposed probe concerns external credit service responds slowly during regional outages, where bounded timeouts and graceful fallback preserve essential user work; this incident instead requires evidence about supported connectors expose external content with discoverability and security metadata and the failure that Copilot cannot find approved wiki pages by natural language.
Question 329
An investment review asks what measurable result would justify continuing an AI solution where business analysts can submit examples but not change production behavior. The original issue is: Unvetted labeled data poisons a retrained classifier. Which indicator should be tracked?
- Cross-tenant retrieved records in evaluation
- Deployed models with complete registry evidence
- Time needed to recover a known-good model/data pair
- Model releases containing unapproved training records
Explanation
Correct Answer
D:Because separated data-curation and release permissions preserve integrity, the outcome worth following is model releases containing unapproved training records; the metric is tied to the reported issue—unvetted labeled data poisons a retrained classifier—for the scenario where business analysts can submit examples but not change production behavior.
Incorrect Answers
A:For the workflow where two customers store similarly named support incidents, cross-tenant retrieved records in evaluation is an actionable measure; it does not establish whether unvetted labeled data poisons a retrained classifier in the situation under review.
B:Although deployed models with complete registry evidence is observable, the affected process involves business analysts can submit examples but not change production behavior; its acceptance outcome is model releases containing unapproved training records, not the alternative measurement.
C:That scorecard concerns the activity in which a preprocessing release changes category labels incorrectly and its result of time needed to recover a known-good model/data pair; the reported shortcoming here is unvetted labeled data poisons a retrained classifier.
Question 330
The operating team must monitor whether an architectural correction works. Business context: A backend agent must call an approved Azure service without embedded secrets. Historical symptom: Long-lived API keys leak from configuration and diagnostics. Which outcome measure is the best match?
- Production agent components using approved workload identity
- Privileged mutations without verified principal authorization
- Unauthorized inference requests successfully blocked
- Authorized tool calls succeeding after promotion
Explanation
Correct Answer
A:For the business use where a backend agent must call an approved Azure service without embedded secrets, the result to quantify is production agent components using approved workload identity; it connects measurement to the mechanism where managed identity provides credential-less workload authentication where supported, not merely to agent traffic.
Incorrect Answers
B:Privileged mutations without verified principal authorization; that indicator describes a distinct workflow where prompt wording persuades the agent to submit a privileged operation; it does not quantify production agent components using approved workload identity, which is needed for the stated case.
C:A change in unauthorized inference requests successfully blocked is informative for a sensitive model endpoint is reachable using a broadly shared key; it does not establish whether the current process achieves production agent components using approved workload identity.
D:The premise of this indicator is production consent, identity and connection setup are distinct release checks; the architecture being evaluated instead serves a backend agent must call an approved Azure service without embedded secrets, and its success is measured by production agent components using approved workload identity.