Question 900
A quality engineer is writing a regression case after the incident where existing vectors are queried with incompatible assumptions. The underlying business scenario is: A search index changes embedding model and chunking approach. What should the test exercise?
- Ask about identical case names from isolated tenants
- Rebuild a test index and switch traffic under observation
- Invoke both versions with boundary category samples
- Promote a candidate in staging and inspect generated audit entries
- Exercise both old and new agent builds against staged endpoints
Explanation
Correct Answer
B:Existing vectors are queried with incompatible assumptions; a release test must establish that this no longer occurs; rebuilding a test index and switch traffic under observation measures the specific behavior for which explicit index versions support compatible retrieval and rollback.
Incorrect Answers
A:In a solution where two customers store similarly named support incidents, asking about identical case names from isolated tenants could catch the defect that assistant retrieves another tenant's incident detail; the current release gate must challenge the failure where existing vectors are queried with incompatible assumptions.
C:Although the exercise might catch downstream apps interpret correct predictions as the wrong category, it would not directly validate the mechanism that explicit index versions support compatible retrieval and rollback in the target implementation.
D:In the alternate setting where a model is replaced without a traceable change ticket, the test exercises deployment audit links actor, approved artifact and release time; even a pass would not show that existing vectors are queried with incompatible assumptions has been prevented.
E:In a solution where a new action version changes how cases are updated, exercising both old and new agent builds against staged endpoints could catch the defect that rolling back the agent leaves an incompatible backend action; the current release gate must challenge the failure where existing vectors are queried with incompatible assumptions.
Question 901
An investment review asks what measurable result would justify continuing an AI solution where agents begin citing an obsolete article after an ingestion job. The original issue is: No log identifies which source version was indexed. Which indicator should be tracked?
- Retrieved chunks traceable to authoritative source versions
- Model changes with complete actor and approval records
- Unauthorized grounding documents included in answers
- Time needed to recover a known-good model/data pair
Explanation
Correct Answer
A:The use case involves agents begin citing an obsolete article after an ingestion job and the baseline failure that no log identifies which source version was indexed; the KPI tracks retrieved chunks traceable to authoritative source versions, so improvement has an interpretable operational meaning.
Incorrect Answers
B:For the workflow where a model is replaced without a traceable change ticket, model changes with complete actor and approval records is an actionable measure; it does not establish whether no log identifies which source version was indexed in the situation under review.
C:The premise of this indicator is retrieval applies effective source entitlements to candidate documents; the architecture being evaluated instead serves agents begin citing an obsolete article after an ingestion job, and its success is measured by retrieved chunks traceable to authoritative source versions.
D:Time needed to recover a known-good model/data pair; that indicator describes a distinct workflow where a preprocessing release changes category labels incorrectly; it does not quantify retrieved chunks traceable to authoritative source versions, which is needed for the stated case.
Question 902
A pilot scorecard needs evidence of business improvement rather than only invocation counts. The use case is: Peak load can saturate a primary model endpoint. Known defect: Requests fail despite another approved deployment being available. Which KPI should be selected?
- Monthly run-rate per active automated workflow
- Successful requests during primary endpoint failures
- Accuracy-adjusted cost per completed business task
- Percentage of production agents with staffed owners and SLOs
Explanation
Correct Answer
B:Because health-aware fallback uses a compatible approved model when needed, the outcome worth following is successful requests during primary endpoint failures; the metric is tied to the reported issue—requests fail despite another approved deployment being available—for the scenario where peak load can saturate a primary model endpoint.
Incorrect Answers
A: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 successful requests during primary endpoint failures.
C:The proposed metric concerns real task mix differs from prelaunch performance tests, whereas the failure in this case is that requests fail despite another approved deployment being available; the indicator needed here is successful requests during primary endpoint failures.
D:Although percentage of production agents with staffed owners and SLOs is observable, the affected process involves peak load can saturate a primary model endpoint; its acceptance outcome is successful requests during primary endpoint failures, not the alternative measurement.
Question 903
A solution architecture is being reviewed for this business situation: Teams propose that an agent perform both low-risk lookups and financial postings. The observed defect is that the same agent can read an invoice and commit a payment unchecked. Which architectural decision addresses the underlying requirement?
- Define autonomy tiers by consequence and reversibility
- Define cross-service boundaries and ownership before integration
- Create an executive-sponsored cross-functional AI strategy team
- Implement an abstain-or-escalate response policy for uncertain outputs
Explanation
Correct Answer
A:The mechanism that tiered permissions require progressively stronger approvals for riskier actions fits the scenario where teams propose that an agent perform both low-risk lookups and financial postings; it addresses the defect that the same agent can read an invoice and commit a payment unchecked.
Incorrect Answers
B:In the case where copilot, Dataverse, Dynamics and custom APIs share one process, the alternate design depends on contracts and ownership boundaries identify responsibility for each state change; it does not correct the target failure, the same agent can read an invoice and commit a payment unchecked.
C:When business, platform, security and legal teams make conflicting AI decisions, a named strategy team assigns decision rights and resolves cross-function tradeoffs; in the stated workflow, however, the same agent can read an invoice and commit a payment unchecked; the relevant architectural behavior is that tiered permissions require progressively stronger approvals for riskier actions.
D:In the case where the system cannot always find evidence for regulated guidance, the alternate design depends on confidence and evidence checks route insufficiently supported cases to humans; it does not correct the target failure, the same agent can read an invoice and commit a payment unchecked.
Question 904
An investment review asks what measurable result would justify continuing an AI solution where a tool response includes HTML or links supplied by third parties. The original issue is: Agent-generated pages render an unsafe instruction or link. Which indicator should be tracked?
- Unsafe tool-response content executed or trusted
- Unapproved external destinations reached through agent tools
- Unauthorized action rate in adversarial evaluations
- Wrong-tool invocation rate on representative intents
Explanation
Correct Answer
A:The use case involves a tool response includes HTML or links supplied by third parties and the baseline failure that agent-generated pages render an unsafe instruction or link; the KPI tracks unsafe tool-response content executed or trusted, so improvement has an interpretable operational meaning.
Incorrect Answers
B:That scorecard concerns the activity in which a prompt injection requests data export to an outside address and its result of unapproved external destinations reached through agent tools; the reported shortcoming here is agent-generated pages render an unsafe instruction or link.
C:For the situation where benchmarks cover only cooperative user wording, unauthorized action rate in adversarial evaluations is a sensible indicator; the problem that agent-generated pages render an unsafe instruction or link requires measurement of unsafe tool-response content executed or trusted.
D:For the workflow where planner calls the wrong tool among overlapping capabilities, wrong-tool invocation rate on representative intents is an actionable measure; it does not establish whether agent-generated pages render an unsafe instruction or link in the situation under review.
Question 905
An enterprise team is making a platform design decision. Business context: Employee policies reside in governed Microsoft 365 documents. Present failure: Answers omit current employee handbook changes. Which approach resolves the mismatch?
- Tag documents with ownership, sensitivity, and provenance metadata
- Use A2A for agent-to-agent collaboration across runtimes
- Ground conversational answers in SharePoint knowledge with permissions
- Use Dataverse knowledge for structured business-record questions
Explanation
Correct Answer
C:Because supported SharePoint grounding retrieves organizational content within access controls, this approach corrects the reported breakdown where answers omit current employee handbook changes; the affected situation is that employee policies reside in governed Microsoft 365 documents.
Incorrect Answers
A:For another type of architecture, metadata enables filtering, review, retention and traceable answers is useful; this design must instead support employee policies reside in governed Microsoft 365 documents despite the problem where answers omit current employee handbook changes.
B:When independently hosted agents must exchange tasks and results, an agent interoperability protocol supports explicit task delegation between agents; in the stated workflow, however, answers omit current employee handbook changes; the relevant architectural behavior is that supported SharePoint grounding retrieves organizational content within access controls.
D:The technical behavior of this alternative is that Dataverse grounding exposes appropriate structured records to the agent; the user case being assessed is employee policies reside in governed Microsoft 365 documents with a different failure: answers omit current employee handbook changes.
Question 906
An incident review has one confirmed symptom: Rolling back the agent leaves an incompatible backend action. For the use case where a new action version changes how cases are updated, which targeted diagnostic check would reveal the most pertinent evidence?
- Audit token use and secret inventory for agent services
- Compare action privileges before and after package import
- Inspect active action versions and backward-compatibility tests
- Inspect solution layers for unmanaged production changes
Explanation
Correct Answer
C:Rolling back the agent leaves an incompatible backend action; to attribute the defect rather than guess at a fix, inspecting active action versions and backward-compatibility tests examines the dependency where coordinated versioning preserves consumer/provider compatibility.
Incorrect Answers
A:The proposed inspection is associated with a backend agent must call an approved Azure service without embedded secrets, where managed identity provides credential-less workload authentication where supported; it would not expose the failure mode that rolling back the agent leaves an incompatible backend action with the required specificity.
B:That probe evaluates authorization regression gates verify posting authority remains constrained in a workflow where agents suggest payment adjustments and inventory holds; the reported failure concerns the path for which coordinated versioning preserves consumer/provider compatibility instead.
D:The proposed probe concerns production components should not be edited casually, where managed solution deployment supports controlled downstream customization boundaries; this incident instead requires evidence about coordinated versioning preserves consumer/provider compatibility and the failure that rolling back the agent leaves an incompatible backend action.
Question 907
A product owner must measure whether an AI-enabled change actually improves this process: Service recommends a price adjustment requiring finance approval. The earlier problem was: The final order lacks traceability to the AI suggestion and approver. Which outcome indicator is most relevant?
- Concessions with full recommendation-to-posting audit chain
- Customer handoffs free from integration mismatches
- Complete onboarding cases with accurate Sales context
- Outbound regulated messages with required approvals
Explanation
Correct Answer
A:For the business use where service recommends a price adjustment requiring finance approval, the result to quantify is concessions with full recommendation-to-posting audit chain; it connects measurement to the mechanism where transaction-path tests confirm recommendation, review and posting records, not merely to agent traffic.
Incorrect Answers
B:Customer handoffs free from integration mismatches; that indicator describes a distinct workflow where a customer handoff depends on shared record and term mappings; it does not quantify concessions with full recommendation-to-posting audit chain, which is needed for the stated case.
C:A change in complete onboarding cases with accurate Sales context is informative for a won opportunity should create a properly owned onboarding case; it does not establish whether the current process achieves concessions with full recommendation-to-posting audit chain.
D:The alternative tracks outbound regulated messages with required approvals, arising from the mechanism where business approval checks remain independent of agent version; its movement alone says little about the failure that the final order lacks traceability to the AI suggestion and approver.
Question 908
Before introducing a fix, the team must explain why the agent invents discretionary promises. The affected setting is: Compliance letters must preserve approved language and disclaimers. What should be inspected?
- Compare outputs against approved language baselines
- Compare output format across supported channels
- Compare tool inventory with legitimate workflow steps
- Validate indexed records and source ACL mappings
Explanation
Correct Answer
A:In the setting where compliance letters must preserve approved language and disclaimers, comparing outputs against approved language baselines can isolate the failure in which the agent invents discretionary promises, because templating and approved clauses limit generated variation.
Incorrect Answers
B:Voice responses contain unreadable links and lengthy tables; examining that other path does not determine whether templating and approved clauses limit generated variation, which is what distinguishes the issue in the stated use case.
C:That probe evaluates limited intent and tool scope prevent unnecessary actions in a workflow where a bot only needs to check warranty and prepare a draft; the reported failure concerns the path for which templating and approved clauses limit generated variation instead.
D:Validating indexed records and source acl mappings explores a separate connection, where supported connectors expose external content with discoverability and security metadata; the diagnosed requirement is compliance letters must preserve approved language and disclaimers, and the actual defect is the agent invents discretionary promises.
Question 909
One enterprise rollout has 3 separately testable requirements. (1) Business use: An agent needs to update a governed ticket status. Known defect: Its knowledge source can explain but cannot commit a database change. (2) Business use: Approval processing requires sequential verification and notification. Known defect: An unconstrained planner sends a notification before authorization. (3) Business use: The orchestrator chooses between similar invoice and credit tools. Known defect: It passes customer IDs into supplier inputs. Which 3 architectural choices directly satisfy these stated cases? Choose THREE.
- Apply security and privacy controls to app-held AI context
- Define robust tool descriptions and input schemas for planner selection
- Choose language analytics for focused text classification
- Use connector actions for authenticated data mutations
- Use an agent flow for repeatable ordered business steps
Explanation
Correct Answers
B:For the independent requirement that the orchestrator chooses between similar invoice and credit tools, specific tool metadata helps the planner choose and parameterize actions; it addresses the defect it passes customer IDs into supplier inputs, while an agent needs to update a governed ticket status; approval processing requires sequential verification and notification needs its own architecture.
D:For the independent requirement that an agent needs to update a governed ticket status, an authorized connector action invokes an operational service; it addresses the defect its knowledge source can explain but cannot commit a database change, while approval processing requires sequential verification and notification; the orchestrator chooses between similar invoice and credit tools needs its own architecture.
E:For the independent requirement that approval processing requires sequential verification and notification, defined flow steps make sequencing and error paths explicit; it addresses the defect an unconstrained planner sends a notification before authorization, while an agent needs to update a governed ticket status; the orchestrator chooses between similar invoice and credit tools needs its own architecture.
Incorrect Answers
A:This design is associated with the different failure where hidden fields from the UI appear in the prompt payload in a workflow where a canvas app passes customer records to a model action; it does not directly solve the listed conditions: its knowledge source can explain but cannot commit a database change; an unconstrained planner sends a notification before authorization; it passes customer IDs into supplier inputs.
C:This design is associated with the different failure where a full agent workflow adds cost to simple text tagging in a workflow where the system must detect customer sentiment and named entities at scale; it does not directly solve the listed conditions: its knowledge source can explain but cannot commit a database change; an unconstrained planner sends a notification before authorization; it passes customer IDs into supplier inputs.
Question 910
The operating team must monitor whether an architectural correction works. Business context: Users have different rights to the same knowledge corpus. Historical symptom: Agent cites a private HR document to unauthorized staff. Which outcome measure is the best match?
- Model releases containing unapproved training records
- Unauthorized grounding documents included in answers
- Unsupported claims emitted in low-evidence tests
- Retrieved chunks traceable to authoritative source versions
Explanation
Correct Answer
B:This scorecard needs an outcome related to users have different rights to the same knowledge corpus; unauthorized grounding documents included in answers. That indicator can be checked against the earlier symptom that agent cites a private HR document to unauthorized staff.
Incorrect Answers
A:For the workflow where business analysts can submit examples but not change production behavior, model releases containing unapproved training records is an actionable measure; it does not establish whether agent cites a private HR document to unauthorized staff in the situation under review.
C:The premise of this indicator is evidence and refusal evaluations enforce source-supported claims; the architecture being evaluated instead serves users have different rights to the same knowledge corpus, and its success is measured by unauthorized grounding documents included in answers.
D:Although retrieved chunks traceable to authoritative source versions is observable, the affected process involves users have different rights to the same knowledge corpus; its acceptance outcome is unauthorized grounding documents included in answers, not the alternative measurement.