Question 770
A planned AI solution must support this situation: Teams want approved external records searchable in Copilot. Its prototype fails because the agent lacks knowledge of non-Microsoft content. Which design best fits the constraint?
- Use explicit handoff contracts for asynchronous specialist work
- Buy a supported managed capability for standard requirements
- Use Microsoft 365 Copilot connectors for external knowledge
- Embed a Copilot-powered interaction inside an existing line-of-business app
Explanation
Correct Answer
C:The defect is that the agent lacks knowledge of non-Microsoft content; the expected business case is that teams want approved external records searchable in Copilot; the design fits because connectors expose external enterprise content to supported Microsoft search/Copilot experiences.
Incorrect Answers
A:Orchestrator returns a partial result as though the workflow finished; the alternate design can address that situation because task IDs, status transitions, deadlines and durable handoff state support coordination, whereas this question needs a solution to the agent lacks knowledge of non-Microsoft content.
B:A bespoke solution creates needless infrastructure overhead; the alternate design can address that situation because managed functionality shifts routine platform operations to the service, whereas this question needs a solution to the agent lacks knowledge of non-Microsoft content.
D:For another type of architecture, app-hosted integration brings relevant AI actions into an owned user interface is useful; this design must instead support teams want approved external records searchable in Copilot despite the problem where the agent lacks knowledge of non-Microsoft content.
Question 771
A business stakeholder describes this operating situation: Several events refer to one business order at once. Testing instead shows that parallel agents each submit a costly replacement. What design approach is most appropriate?
- Test concurrent agent runs for duplicate side effects
- Preserve reproducible evaluation evidence for major model changes
- Track throughput and throttling to plan capacity
- Test tool-permission escalation attempts at the execution layer
Explanation
Correct Answer
A:Several events refer to one business order at once; its failing behavior is that parallel agents each submit a costly replacement; applying the underlying design principle—concurrency evaluation tests idempotency and serialization—connects the requirement to an implementable control.
Incorrect Answers
B:For another type of architecture, versioned evaluation sets, model IDs and thresholds document release claims is useful; this design must instead support several events refer to one business order at once despite the problem where parallel agents each submit a costly replacement.
C:This design would make sense if a regional launch multiplies concurrent agent requests, particularly where rate limiting causes intermittent failures; the requested architecture instead needs the dependency that concurrency evaluation tests idempotency and serialization because parallel agents each submit a costly replacement.
D:This design would make sense if a user asks the agent to read an unrelated executive record, particularly where the model may try a tool outside the user's rights; the requested architecture instead needs the dependency that concurrency evaluation tests idempotency and serialization because parallel agents each submit a costly replacement.
Question 772
Implementation alternatives must be compared for a workflow where a search index changes embedding model and chunking approach. The pilot shows this gap: Existing vectors are queried with incompatible assumptions. Which alternative should lead the solution?
- Deploy actions with compatible rollback behavior
- Version training datasets with lineage and access records
- Record immutable model deployment events with approver and version
- Treat index schema and embeddings as versioned artifacts
Explanation
Correct Answer
D:A search index changes embedding model and chunking approach; its failing behavior is that existing vectors are queried with incompatible assumptions; applying the underlying design principle—explicit index versions support compatible retrieval and rollback—connects the requirement to an implementable control.
Incorrect Answers
A:For another type of architecture, coordinated versioning preserves consumer/provider compatibility is useful; this design must instead support a search index changes embedding model and chunking approach despite the problem where existing vectors are queried with incompatible assumptions.
B:For another type of architecture, immutable snapshots and lineage tie runs to approved sources is useful; this design must instead support a search index changes embedding model and chunking approach despite the problem where existing vectors are queried with incompatible assumptions.
C:In the case where a model is replaced without a traceable change ticket, the alternate design depends on deployment audit links actor, approved artifact and release time; it does not correct the target failure, existing vectors are queried with incompatible assumptions.
Question 773
The architect wants evidence before altering the deployment. In a workflow where an internal helper produces attractive but unsupported policy answers, satisfaction scores hide factual mistakes. What should the team examine first?
- Audit incorrect answers and their cited source spans
- Inspect backend denials and attempted tool arguments
- Compare query, filters and top retrieved source identifiers
- Verify spans and errors arrive in the expected telemetry resource
Explanation
Correct Answer
A:The investigation directly tests the causal connection: expert-checked evaluation examples quantify evidence fidelity; it collects evidence for a workflow where an internal helper produces attractive but unsupported policy answers but the outcome is instead that satisfaction scores hide factual mistakes.
Incorrect Answers
B:For the failure that the model may try a tool outside the user's rights, inspecting backend denials and attempted tool arguments is relevant; this question instead reports that satisfaction scores hide factual mistakes, so different evidence is needed.
C:That probe evaluates retrieval telemetry distinguishes evidence search from answer generation in a workflow where an assistant responds fluently but cites unrelated support articles; the reported failure concerns the path for which expert-checked evaluation examples quantify evidence fidelity instead.
D:That probe evaluates connected observability supplies traces and operational metrics in a workflow where a newly deployed Foundry agent has no detailed diagnostics; the reported failure concerns the path for which expert-checked evaluation examples quantify evidence fidelity instead.
Question 774
A test plan focuses on a previously missed failure: A large context prompt quietly grows every release. It occurred when model bills increase although request volume is stable. Which check directly validates that behavior?
- Test a conversation that ends politely without resolving the case
- Re-run a sampled historical release on its frozen corpus
- Re-run the same multi-intent corpus on both configurations
- Replay a stable request set on both prompt versions
- Replay complex and simple cases through revised thresholds
Explanation
Correct Answer
D:The scenario is that model bills increase although request volume is stable; the defect to rule out is that a large context prompt quietly grows every release; testing via replaying a stable request set on both prompt versions challenges the design dependency where token telemetry isolates prompt, retrieval and generation cost changes.
Incorrect Answers
A:For a service bot handles many chats but few cases are resolved, the alternate test is about whether outcome measures connect agent use with completed business work; it cannot prove the outcome needed after the reported regression where a large context prompt quietly grows every release.
B:For an updated agent performs differently months after release, the alternate test is about whether versioned evaluation sets, model IDs and thresholds document release claims; it cannot prove the outcome needed after the reported regression where a large context prompt quietly grows every release.
C:The verification concerning a new topic description changes which agent is invoked checks trace comparisons reveal shifts in plan selection and step order against regression appears only for compound user requests; its result says little about the observed issue where a large context prompt quietly grows every release.
E:In a solution where average accuracy hides failures on complex finance questions, replaying complex and simple cases through revised thresholds could catch the defect that lower-cost routing degrades a business-critical subgroup; the current release gate must challenge the failure where a large context prompt quietly grows every release.
Question 775
Before production promotion, the team needs evidence for this setting: Employees want curated assistance without owning a separate LLM runtime. A prior regression caused the following: A custom service would duplicate built-in host orchestration. Which validation is most persuasive?
- Create an escalation from a test case and inspect audit records
- Update an approved prompt and validate dependent flows in test
- Install the agent to a pilot Microsoft 365 audience
- Attempt authorized and denied inventory calls through the client
Explanation
Correct Answer
C:The check is targeted rather than generic: declarative agents define instructions, knowledge and actions in the supported host; its pass/fail evidence applies to the workflow where employees want curated assistance without owning a separate LLM runtime and to the prior defect that a custom service would duplicate built-in host orchestration.
Incorrect Answers
A:The suggested test serves a representative must initiate an exception workflow from the case, in which AI suggestions are never converted into traceable escalation records; the present scenario instead concerns the reported problem where a custom service would duplicate built-in host orchestration.
B:A pass on that verification would speak to multiple low-code apps need a common extraction prompt; it would not show that declarative agents define instructions, knowledge and actions in the supported host, which is needed to avoid the observed defect a custom service would duplicate built-in host orchestration.
D:A pass on that verification would speak to several agents need the same inventory query operation; it would not show that declarative agents define instructions, knowledge and actions in the supported host, which is needed to avoid the observed defect a custom service would duplicate built-in host orchestration.
Question 776
A release must prove that the agent lacks knowledge of non-Microsoft content will not recur. The application is used in this situation: Teams want approved external records searchable in Copilot. Which acceptance test most directly checks that failure mode?
- Query an external article using users with different permissions
- Verify task handoff and result exchange across two supported runtimes
- Prototype the distinguishing behavior under production constraints
- Validate critical user journeys in a supported product trial
Explanation
Correct Answer
A:Querying an external article using users with different permissions challenges the path where connectors expose external enterprise content to supported Microsoft search/Copilot experiences for the business setting teams want approved external records searchable in Copilot; the specific defect to exclude is that the agent lacks knowledge of non-Microsoft content.
Incorrect Answers
B:Verifying task handoff and result exchange across two supported runtimes verifies an adjacent objective where independently hosted agents must exchange tasks and results; it leaves untested the causal path in which connectors expose external enterprise content to supported Microsoft search/Copilot experiences for this case.
C:For competitive requirements depend on unique multi-model control, the alternate test is about whether custom engineering trades more control for more operating responsibility; it cannot prove the outcome needed after the reported regression where the agent lacks knowledge of non-Microsoft content.
D:The suggested test serves the requirement is common and product coverage is adequate, in which a bespoke solution creates needless infrastructure overhead; the present scenario instead concerns the reported problem where the agent lacks knowledge of non-Microsoft content.
Question 777
An agent-assisted service does not behave as specified: A general large model exceeds per-document unit economics. It is used in this situation: High-volume narrow extraction needs predictable latency and cost. Which observable diagnostic procedure is best?
- Benchmark accuracy, throughput and resource demands together
- Compare host limitations against the requirements matrix
- Compare orchestration requirements with host extensibility limits
- Compare gap size to supported extension interfaces
Explanation
Correct Answer
A:Where a general large model exceeds per-document unit economics, the useful evidence comes from benchmarking accuracy, throughput and resource demands together; its focus matches the architectural mechanism that smaller models can be evaluated for narrow tasks with lower cost in the process where high-volume narrow extraction needs predictable latency and cost.
Incorrect Answers
B:The proposed inspection is associated with a workflow needs proprietary scheduling, model choice and runtime control, where owning the engine provides control over models, routing and execution; it would not expose the failure mode that a general large model exceeds per-document unit economics with the required specificity.
C:The solution demands its own model selection and hosting policies; its matching diagnostic check is reasonable for host-managed orchestration cannot satisfy a unique execution plan; the present investigation needs evidence for the different failure where a general large model exceeds per-document unit economics.
D:In the process where current users need one governed custom action inside their workflow, the diagnostic procedure could explain full replacement increases retraining and migration risk; this particular incident instead turns on whether smaller models can be evaluated for narrow tasks with lower cost and why a general large model exceeds per-document unit economics.
Question 778
A quality engineer is writing a regression case after the incident where a monolingual analysis misses non-English escalations. The underlying business scenario is: Support documents arrive in several customer languages. What should the test exercise?
- Test edge cases with conflicting documents and incomplete facts
- Send representative messages through translation and human review
- Load-test realistic concurrent requests and test rollback
- Run representative notes and validate schema and exclusions
Explanation
Correct Answer
B:For the required workflow where support documents arrive in several customer languages, the test needs to reproduce or prevent the failure where a monolingual analysis misses non-English escalations; sending representative messages through translation and human review exercises the path where translation provides standardized text input while preserving original evidence.
Incorrect Answers
A:This check exercises a different system behavior: reasoning can plan through evidence and tools within enforced boundaries; a pass there cannot establish that a monolingual analysis misses non-English escalations no longer occurs in the required workflow.
C:The tested behavior is that a managed serving endpoint needs scaling, monitoring and rollback design; the necessary evidence here concerns translation provides standardized text input while preserving original evidence and the previously observed a monolingual analysis misses non-English escalations.
D:The tested behavior is that prompt actions encapsulate reusable inputs and expected outputs; the necessary evidence here concerns translation provides standardized text input while preserving original evidence and the previously observed a monolingual analysis misses non-English escalations.
Question 779
Business leaders reject an activity-only metric for this process: A third-party AI tool changes where submitted documents are processed. They need to detect the issue where approved data crosses an unreviewed processor. Which performance indicator fits?
- Third-party processing paths with verified compliance records
- Sensitive requests crossing unauthorized residency boundaries
- Release candidates satisfying critical evaluation thresholds
- Model changes with complete actor and approval records
Explanation
Correct Answer
A:This scorecard needs an outcome related to a third-party AI tool changes where submitted documents are processed; third-party processing paths with verified compliance records. That indicator can be checked against the earlier symptom that approved data crosses an unreviewed processor.
Incorrect Answers
B:For the situation where a regulated workload must keep data in permitted locations, sensitive requests crossing unauthorized residency boundaries is a sensible indicator; the problem that approved data crosses an unreviewed processor requires measurement of third-party processing paths with verified compliance records.
C:The premise of this indicator is evaluation suites validate quality and safety before publication; the architecture being evaluated instead serves a third-party AI tool changes where submitted documents are processed, and its success is measured by third-party processing paths with verified compliance records.
D:A model is replaced without a traceable change ticket; its intended measurement is model changes with complete actor and approval records; substituting that result here would not capture whether approved data crosses an unreviewed processor has been corrected.
Question 780
The architect wants evidence before altering the deployment. In a workflow where a request crosses Copilot Studio, a flow and the CRM backend, logs cannot reconstruct which API mutation followed an answer. What should the team examine first?
- Trace source exception through target state and error queue
- Inspect solution layers for unmanaged production changes
- Compare request IDs to backend transaction counts
- Trace one business request through each dependent component
Explanation
Correct Answer
D:The investigation directly tests the causal connection: correlation identifiers link the distributed operational path; it collects evidence for a workflow where a request crosses Copilot Studio, a flow and the CRM backend but the outcome is instead that logs cannot reconstruct which API mutation followed an answer.
Incorrect Answers
A:Tracing source exception through target state and error queue explores a separate connection, where failure-injection tests verify retries and compensation between systems; the diagnosed requirement is a request crosses Copilot Studio, a flow and the CRM backend, and the actual defect is logs cannot reconstruct which API mutation followed an answer.
B:The evidence this method produces applies to a workflow where production components should not be edited casually; it cannot by itself establish why logs cannot reconstruct which API mutation followed an answer in the current scenario.
C:Agents modify records through external APIs; its matching diagnostic check is reasonable for successful chat sessions conceal repeated backend writes; the present investigation needs evidence for the different failure where logs cannot reconstruct which API mutation followed an answer.