Question 120
A risk analysis highlights a threat in this workflow: Two topics claim nearly identical scheduling intents. Exposure: Overlapping scheduling intents might invoke a change before confirmation. Which enforceable safeguard addresses it at the relevant boundary?
- Disable write tools and label generated material for review
- Require confirmation when intent remains ambiguous
- Preserve conversation context and avoid infinite re-prompting
- Apply tenant policies and scoped content permissions
Explanation
Correct Answer
B:When overlapping scheduling intents might invoke a change before confirmation, the use case two topics claim nearly identical scheduling intents requires a guardrail aligned with the behavior that distinct descriptions and inputs improve topic selection.
Incorrect Answers
A:This restriction is associated with marketing specialists need tailored descriptions from existing briefings and addresses the exposure that a text-only drafting assistant might be given unnecessary record-writing powers; it does not specifically protect the target workflow against overlapping scheduling intents might invoke a change before confirmation.
C:That protection is tied to the mechanism that confidence, retry ceilings and escalation routing enable resolution; it does not prevent the distinct risk where overlapping scheduling intents might invoke a change before confirmation in the system under review.
D:The safeguard for employees want curated assistance without owning a separate LLM runtime deals with a different threat: a declarative agent might be published beyond permitted tenant groups; even after it is enforced, the exposure that overlapping scheduling intents might invoke a change before confirmation remains possible in the target workflow where distinct descriptions and inputs improve topic selection.
Question 121
After deployment, stakeholders need an interpretable measure for this use case: Users use novel requests outside the supported topic inventory. The pilot showed that the agent repeats an unhelpful default message. Which metric captures the needed result?
- Unsupported requests resolved through useful escalation
- External knowledge retrievals respecting access permissions
- Role-specific tasks completed in supported Copilot surfaces
- Correct topic selection for closely related user intents
Explanation
Correct Answer
A:This scorecard needs an outcome related to users use novel requests outside the supported topic inventory; unsupported requests resolved through useful escalation. That indicator can be checked against the earlier symptom that the agent repeats an unhelpful default message.
Incorrect Answers
B:For the workflow where workers need internal wiki articles from outside Microsoft 365, external knowledge retrievals respecting access permissions is an actionable measure; it does not establish whether the agent repeats an unhelpful default message in the situation under review.
C:For the workflow where employees want curated assistance without owning a separate LLM runtime, role-specific tasks completed in supported Copilot surfaces is an actionable measure; it does not establish whether the agent repeats an unhelpful default message in the situation under review.
D:For the situation where two topics claim nearly identical scheduling intents, correct topic selection for closely related user intents is a sensible indicator; the problem that the agent repeats an unhelpful default message requires measurement of unsupported requests resolved through useful escalation.
Question 122
An enterprise team is making a platform design decision. Business context: Policy documents contain draft, superseded, and currently effective revisions. Present failure: The agent cites a superseded rule despite a newer policy existing. Which approach resolves the mismatch?
- Add fallback routing for model unavailability and throttling
- Filter retrieval by effective dates and authoritative version
- Normalize identifiers and remove duplicated grounding records
- Ground conversational answers in SharePoint knowledge with permissions
- Use retrieval augmentation before tuning for changing facts
- Use quality-and-cost telemetry to tune routing thresholds
Explanation
Correct Answer
B:The architecture must reconcile two facts: policy documents contain draft, superseded, and currently effective revisions, yet the agent cites a superseded rule despite a newer policy existing; the selected pattern does so through the mechanism that version and effective-date metadata select authoritative content at retrieval time.
Incorrect Answers
A:When peak load can saturate a primary model endpoint, health-aware fallback uses a compatible approved model when needed; in the stated workflow, however, the agent cites a superseded rule despite a newer policy existing; the relevant architectural behavior is that version and effective-date metadata select authoritative content at retrieval time.
C:The technical behavior of this alternative is that canonical keys and deduplication prevent incorrect cross-record joins; the user case being assessed is policy documents contain draft, superseded, and currently effective revisions with a different failure: the agent cites a superseded rule despite a newer policy existing.
D:In the case where employee policies reside in governed Microsoft 365 documents, the alternate design depends on supported SharePoint grounding retrieves organizational content within access controls; it does not correct the target failure, the agent cites a superseded rule despite a newer policy existing.
E:For another type of architecture, retrieval separates volatile knowledge from model parameters is useful; this design must instead support policy documents contain draft, superseded, and currently effective revisions despite the problem where the agent cites a superseded rule despite a newer policy existing.
F:The technical behavior of this alternative is that feedback-driven thresholds rebalance accuracy, latency and spend; the user case being assessed is policy documents contain draft, superseded, and currently effective revisions with a different failure: the agent cites a superseded rule despite a newer policy existing.
Question 123
The architecture board has evidence that small-model outputs omit critical contractual exceptions. The affected situation is: Some cases need cross-document reasoning and detailed justification. What change most directly establishes the missing behavior?
- Use retrieval augmentation before tuning for changing facts
- Add fallback routing for model unavailability and throttling
- Route high-stakes reasoning to a stronger validated model
- Replace an unsuitable prebuilt design with a custom-engine agent
Explanation
Correct Answer
C:The mechanism that complexity classification selects a capable model for difficult tasks fits the scenario where some cases need cross-document reasoning and detailed justification; it addresses the defect that small-model outputs omit critical contractual exceptions.
Incorrect Answers
A:Retrieval separates volatile knowledge from model parameters; its main purpose is answers depend on policies revised every week, not correction of the reported defect where small-model outputs omit critical contractual exceptions in the target process.
B:Requests fail despite another approved deployment being available; the alternate design can address that situation because health-aware fallback uses a compatible approved model when needed, whereas this question needs a solution to small-model outputs omit critical contractual exceptions.
D:There is a valid place for the mechanism that owning the engine provides control over models, routing and execution, namely a workflow needs proprietary scheduling, model choice and runtime control; it does not remove the specific defect that small-model outputs omit critical contractual exceptions from this case.
Question 124
A business stakeholder describes this operating situation: Employees express several intents and need interactive clarification. Testing instead shows that a rigid menu fails whenever a request contains two intents. What design approach is most appropriate?
- Use a deterministic Power Automate flow for stable approvals
- Use explicit handoff contracts for asynchronous specialist work
- Use MCP for an agent's access to external tools and data
- Use a Copilot Studio conversational agent for variable natural-language requests
Explanation
Correct Answer
D:Because generative orchestration interprets intent and can combine authorized tools and knowledge, this approach corrects the reported breakdown where a rigid menu fails whenever a request contains two intents; the affected situation is that employees express several intents and need interactive clarification.
Incorrect Answers
A:The alternative addresses employees receive inconsistent approvals when prose drives the decision by a design in which explicit conditions execute known business rules predictably and leave an auditable branch history; it does not supply the missing relationship for employees express several intents and need interactive clarification, where a rigid menu fails whenever a request contains two intents.
B:For another type of architecture, task IDs, status transitions, deadlines and durable handoff state support coordination is useful; this design must instead support employees express several intents and need interactive clarification despite the problem where a rigid menu fails whenever a request contains two intents.
C:The alternative addresses every agent reimplements the same proprietary tool signatures by a design in which MCP presents supported tools and resources through a common integration contract; it does not supply the missing relationship for employees express several intents and need interactive clarification, where a rigid menu fails whenever a request contains two intents.
Question 125
The architecture board has evidence that unvetted labeled data poisons a retrained classifier. The affected situation is: Business analysts can submit examples but not change production behavior. What change most directly establishes the missing behavior?
- Enforce tenant and entity boundaries before vector search results are used
- Restrict who can supply tuning data and approve model updates
- Apply output filtering and usage restrictions to model interfaces
- Register each trained model with dataset and evaluation provenance
Explanation
Correct Answer
B:The mechanism that separated data-curation and release permissions preserve integrity fits the scenario where business analysts can submit examples but not change production behavior; it addresses the defect that unvetted labeled data poisons a retrained classifier.
Incorrect Answers
A:There is a valid place for the mechanism that partitioning and authorized filters constrain relevant candidates, namely two customers store similarly named support incidents; it does not remove the specific defect that unvetted labeled data poisons a retrained classifier from this case.
C:Safety controls and authorization checks limit misuse and disclosure; its main purpose is an internal model sometimes returns confidential text verbatim, not correction of the reported defect where unvetted labeled data poisons a retrained classifier in the target process.
D:A model registry ties versions to lineage, evaluations and approval; its main purpose is data scientists cannot explain why a newer classifier behaves differently, not correction of the reported defect where unvetted labeled data poisons a retrained classifier in the target process.
Question 126
An agent-assisted service does not behave as specified: A release uses stale prompts against a new tool signature. It is used in this situation: Model and tool changes are promoted independently. Which observable diagnostic procedure is best?
- Trace model version to source manifest and transformation hash
- Inspect solution dependencies and common entity schema
- Audit quarantined document reasons and promotion events
- Compare deployed agent configuration to its approved manifest
- Compare ticket assignment confusion matrices across versions
Explanation
Correct Answer
D:A release uses stale prompts against a new tool signature; to attribute the defect rather than guess at a fix, comparing deployed agent configuration to its approved manifest examines the dependency where a complete agent release manifest pins related runtime artifacts.
Incorrect Answers
A:For the failure that no one can identify which dataset produced a model, tracing model version to source manifest and transformation hash is relevant; this question instead reports that a release uses stale prompts against a new tool signature, so different evidence is needed.
B:Independent releases cause contradictory customer statuses; examining that other path does not determine whether a complete agent release manifest pins related runtime artifacts, which is what distinguishes the issue in the stated use case.
C:For the failure that unreviewed confidential drafts appear in agent answers, auditing quarantined document reasons and promotion events is relevant; this question instead reports that a release uses stale prompts against a new tool signature, so different evidence is needed.
E:The evidence this method produces applies to a workflow where a support agent changes how it classifies priority intents; it cannot by itself establish why a release uses stale prompts against a new tool signature in the current scenario.
Question 127
Implementation alternatives must be compared for a workflow where average accuracy hides failures on complex finance questions. The pilot shows this gap: Lower-cost routing degrades a business-critical subgroup. Which alternative should lead the solution?
- Tune model routing based on segmented evaluation outcomes
- Restrict who can supply tuning data and approve model updates
- Cluster similar agent failures using embeddings or AI-assisted analysis
- Track throughput and throttling to plan capacity
Explanation
Correct Answer
A:For the need that average accuracy hides failures on complex finance questions, the relevant mechanism is that stratified evaluation guides thresholds for difficult requests; that architectural dependency addresses the observed failure where lower-cost routing degrades a business-critical subgroup.
Incorrect Answers
B:Separated data-curation and release permissions preserve integrity; its main purpose is business analysts can submit examples but not change production behavior, not correction of the reported defect where lower-cost routing degrades a business-critical subgroup in the target process.
C:There is a valid place for the mechanism that semantic clustering groups related failure patterns for review, namely thousands of transcripts contain recurring misunderstandings; it does not remove the specific defect that lower-cost routing degrades a business-critical subgroup from this case.
D: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 stratified evaluation guides thresholds for difficult requests because lower-cost routing degrades a business-critical subgroup.
Question 128
The architect wants evidence before altering the deployment. In a workflow where the task follows fixed approval branches with no interpretation needed, employees receive inconsistent approvals when prose drives the decision. What should the team examine first?
- Inspect task envelopes and agent capability declarations
- Inspect approvals against the approved decision table
- Inspect traces for instruction-following from retrieved content
- Review available tool inventory versus approved actions
Explanation
Correct Answer
B:For the situation where the task follows fixed approval branches with no interpretation needed, the defect is that employees receive inconsistent approvals when prose drives the decision; inspecting approvals against the approved decision table helps verify the cause-and-effect path: explicit conditions execute known business rules predictably and leave an auditable branch history.
Incorrect Answers
A:The evidence this method produces applies to a workflow where independently hosted agents must exchange tasks and results; it cannot by itself establish why employees receive inconsistent approvals when prose drives the decision in the current scenario.
C:In the process where third-party documents include sentences telling the model to ignore rules, the diagnostic procedure could explain answers adopt an instruction embedded in a source document; this particular incident instead turns on whether explicit conditions execute known business rules predictably and leave an auditable branch history and why employees receive inconsistent approvals when prose drives the decision.
D:The proposed probe concerns agent instructions mention many APIs but only two are permitted, where allowlisting and server-enforced identity restrict actual executable capability; this incident instead requires evidence about explicit conditions execute known business rules predictably and leave an auditable branch history and the failure that employees receive inconsistent approvals when prose drives the decision.
Question 129
Troubleshooting starts from a business complaint: Human testers overlook values just above approval limits. The affected process is: Agent handles threshold-based return decisions. Which investigation is most directly tied to the reported defect?
- Compare query, filters and top retrieved source identifiers
- Compare quality by task complexity and selected model
- Inspect tests for each threshold and alternate branch
- Inspect overlapping tool definitions and erroneous call paths
Explanation
Correct Answer
C:In the setting where agent handles threshold-based return decisions, inspecting tests for each threshold and alternate branch can isolate the failure in which human testers overlook values just above approval limits, because assisted test generation explores stated boundary conditions.
Incorrect Answers
A: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 assisted test generation explores stated boundary conditions instead.
B:The evidence this method produces applies to a workflow where average accuracy hides failures on complex finance questions; it cannot by itself establish why human testers overlook values just above approval limits in the current scenario.
D:The evidence this method produces applies to a workflow where planner calls the wrong tool among overlapping capabilities; it cannot by itself establish why human testers overlook values just above approval limits in the current scenario.
Question 130
A design adds autonomous behavior to a process where users use novel requests outside the supported topic inventory. This creates a documented risk: An unmatched fallback might accidentally bypass required identity verification. Which enforced control is the best fit?
- Honor source access when syncing or federating content
- Validate types and authorization at the called service
- Stop and compensate when a required approval fails
- Prevent fallback from bypassing authentication or policies
Explanation
Correct Answer
D:In the context where users use novel requests outside the supported topic inventory, the exposure is that an unmatched fallback might accidentally bypass required identity verification; containing it requires a guardrail in the operation where fallback handles unmatched intent and can ask for clarification or escalate.
Incorrect Answers
A:This restriction is associated with workers need internal wiki articles from outside Microsoft 365 and addresses the exposure that an external connector might index documents without honoring original ACLs; it does not specifically protect the target workflow against an unmatched fallback might accidentally bypass required identity verification.
B:Another process could need this guardrail because an API might accept a valid-looking but unauthorized generated identifier; the design being assessed supports users use novel requests outside the supported topic inventory and has the explicit exposure that an unmatched fallback might accidentally bypass required identity verification.
C:Where approval processing requires sequential verification and notification, the guard can address a flow might notify the customer even though a required approval failed; the exposure that an unmatched fallback might accidentally bypass required identity verification remains relevant to the present assessment.