Question 20
Operations cannot tell whether low answer quality began in document parsing, retrieval, or the model. Which design choice best matches the stated constraint?
- Use stable request identifiers and idempotent writes at the posting boundary.
- Capture per-stage evidence, latencies, model identifiers, retrieved document IDs and validation outcomes.
- Propagate retrieval status and enforce a fail-closed or explicit uncertainty response.
- Define explicit ingestion, extraction, retrieval, inference, validation, human review and case-posting stages with correlation IDs.
Explanation
Correct Answer
B:End-to-end instrumentation separates ingestion defects from reasoning defects and makes root-cause analysis possible.
Incorrect Answers
A:This is a valid architectural concern when optimizing duplicate prevention for retried writes. Here it leaves observability across pipeline stages unaddressed, whereas the intended change must capture per-stage evidence, latencies, model identifiers, retrieved document IDs and validation outcomes before the case can be considered resolved.
C:By itself, this improves safe degradation when evidence retrieval fails, but the case is not asking the architect to optimize that dimension. The requirement is observability across pipeline stages, supported by an approach to capture per-stage evidence, latencies, model identifiers, retrieved document IDs and validation outcomes.
D:Its main effect concerns traceable stage-by-stage execution. The case explicitly hinges on observability across pipeline stages; to resolve it reliably, the implementation has to capture per-stage evidence, latencies, model identifiers, retrieved document IDs and validation outcomes, which this choice does not establish.
Question 21
A marketing generator is scheduled nightly but sometimes writes outputs after the compliance approval window has closed. Which architecture adjustment is most appropriate?
- Define explicit ingestion, extraction, retrieval, inference, validation, human review and case-posting stages with correlation IDs.
- Use consented, minimal preference state separate from expiring conversational context.
- Add stateful approval gates and time-bounded publishing authorization.
- Propagate retrieval status and enforce a fail-closed or explicit uncertainty response.
- Use stable request identifiers and idempotent writes at the posting boundary.
Explanation
Correct Answer
C:Workflow state prevents late completions from bypassing a review decision or posting with expired permission. Place a verifiable contract or checkpoint at the stage responsible for time-sensitive workflow approvals, then test recovery without silently repeating external writes.
Incorrect Answers
A:This addresses traceable stage-by-stage execution; however, the problem asks for time-sensitive workflow approvals. Without the safeguard to add stateful approval gates and time-bounded publishing authorization, the original failure mode is still possible.
B:Although minimal persistent memory boundaries can be valuable, it is not the acceptance condition in the question. The missing provision is time-sensitive workflow approvals, addressed by the control to add stateful approval gates and time-bounded publishing authorization.
D:It emphasizes safe degradation when evidence retrieval fails without closing the stated gap in time-sensitive workflow approvals. By contrast, the remedy to add stateful approval gates and time-bounded publishing authorization acts on the condition that makes this scenario fail.
E:That is primarily a technique for duplicate prevention for retried writes. It lacks the explicit measure for time-sensitive workflow approvals that is necessary here, namely a system that will add stateful approval gates and time-bounded publishing authorization.
Question 22
A product team changes document format weekly and the prompt must not break every downstream consumer. Which approach should be recommended?
- Use stable request identifiers and idempotent writes at the posting boundary.
- Propagate retrieval status and enforce a fail-closed or explicit uncertainty response.
- Use consented, minimal preference state separate from expiring conversational context.
- Use versioned interfaces between parsing, retrieval, generation and response rendering.
Explanation
Correct Answer
D:Stable contracts isolate changing upstream formats and permit tested component upgrades without cascading rewrites.
Incorrect Answers
A:Consider its actual operating scope: duplicate prevention for retried writes. The requested assurance is modular component contracts, which calls for an implementation that can use versioned interfaces between parsing, retrieval, generation and response rendering, not merely for a related improvement.
B:The alternative could reduce risk associated with safe degradation when evidence retrieval fails, yet the described behavior requires modular component contracts. That outcome depends on the capability to use versioned interfaces between parsing, retrieval, generation and response rendering.
C:The proposed measure is aimed at minimal persistent memory boundaries, not modular component contracts. The architecture must instead use versioned interfaces between parsing, retrieval, generation and response rendering; a separate change is still needed for modular component contracts.
Question 23
A personalized FAQ service must remember user preference but may not preserve raw conversation transcripts indefinitely. What is the best production design response?
- Add stateful approval gates and time-bounded publishing authorization.
- Persist workflow checkpoints and deduplicate external actions using transaction IDs.
- Use versioned interfaces between parsing, retrieval, generation and response rendering.
- Use consented, minimal preference state separate from expiring conversational context.
Explanation
Correct Answer
D:Persisting only necessary structured state supports continuity without turning transcript retention into a privacy liability.
Incorrect Answers
A:Its main effect concerns time-sensitive workflow approvals. The case explicitly hinges on minimal persistent memory boundaries; to resolve it reliably, the implementation has to use consented, minimal preference state separate from expiring conversational context, which this choice does not establish.
B:The option might improve recovery without repeating committed actions; that remains insufficient for minimal persistent memory boundaries. The case is resolved by minimal persistent memory boundaries, which depends on a design that will use consented, minimal preference state separate from expiring conversational context.
C:The proposed action prioritizes modular component contracts. That can coexist with the right solution, but it cannot replace minimal persistent memory boundaries: the application must still use consented, minimal preference state separate from expiring conversational context.
Question 24
A policy question-answering interface only needs to retrieve relevant text and produce cited summaries; it never takes actions. Which change most directly satisfies this requirement?
- Prefer a single model step with a narrow retrieval call over a multi-agent team.
- Use an evaluator-optimizer workflow with a measurable completion criterion.
- Use a retrieval-augmented single-turn LLM design rather than autonomous agents.
- Eliminate the redundant model pass and retain targeted validation where risk warrants it.
Explanation
Correct Answer
C:Bounded question answering has no need for open-ended planning or delegated side effects, keeping evaluation and operations simpler.
Incorrect Answers
A:It emphasizes low-overhead design for simple tasks without closing the stated gap in bounded grounded question answering. By contrast, the remedy to use a retrieval-augmented single-turn LLM design rather than autonomous agents acts on the condition that makes this scenario fail.
B:That is primarily a technique for bounded iterative quality improvement. It lacks the explicit measure for bounded grounded question answering that is necessary here, namely a system that will use a retrieval-augmented single-turn LLM design rather than autonomous agents.
D:This recommendation targets removing unjustified multi-step complexity at a different point in the workflow. It leaves bounded grounded question answering outside the design contract and does not require the team to use a retrieval-augmented single-turn LLM design rather than autonomous agents.
Question 25
A manufacturing report always follows extraction, validation, enrichment and a fixed approval order. Which solution best resolves this operational gap?
- Implement a deterministic prompt-and-tool workflow with explicit stage transitions.
- Use a bounded agent loop with explicit tool budgets and stop conditions.
- Replace open-ended planning with a chained workflow for the known path.
- Keep classification in a versioned deterministic decision engine; use Claude only for explanation.
Explanation
Correct Answer
A:Known dependencies are easier to test and recover as fixed stages than as an agent choosing arbitrary next steps.
Incorrect Answers
B:The control would be justified if adaptive research with bounded autonomy were the limiting factor. In this scenario the decisive target is predictable ordered processing; success requires an architecture that will implement a deterministic prompt-and-tool workflow with explicit stage transitions.
C:The option belongs to the broader work on preventing needless agent autonomy, but it cannot by itself deliver predictable ordered processing. A decision on this case needs the team to implement a deterministic prompt-and-tool workflow with explicit stage transitions and verify that outcome.
D:This shifts attention toward reproducible classification under fixed rules when the relevant decision is about predictable ordered processing. Unless the architecture can implement a deterministic prompt-and-tool workflow with explicit stage transitions, the failure described in the stem remains unresolved.
Question 26
A research task may require searching, revising queries and selecting different sources depending on evidence found. What design decision would best fit these facts?
- Implement a deterministic prompt-and-tool workflow with explicit stage transitions.
- Replace open-ended planning with a chained workflow for the known path.
- Combine agentic evidence collection with a deterministic transaction gate.
- Use a bounded agent loop with explicit tool budgets and stop conditions.
Explanation
Correct Answer
D:Dynamic information gathering benefits from adaptive planning, but caps prevent unbounded latency and spend. Test whether the selected level of autonomy actually achieves adaptive research with bounded autonomy; extra agency is justified only by measurable gains in adaptive research with bounded autonomy.
Incorrect Answers
A:The design concern is real but different: predictable ordered processing. It does not impose the specific requirement for adaptive research with bounded autonomy; a suitable solution must use a bounded agent loop with explicit tool budgets and stop conditions and demonstrate the result.
B:This addresses preventing needless agent autonomy; however, the problem asks for adaptive research with bounded autonomy. Without the safeguard to use a bounded agent loop with explicit tool budgets and stop conditions, the original failure mode is still possible.
C:That choice would be responsive to hybrid agent-and-rules boundaries, but the reported defect is different. It leaves adaptive research with bounded autonomy unproved because there is no commitment to use a bounded agent loop with explicit tool budgets and stop conditions.
Question 27
A help desk receives billing, technical and policy requests that require distinct prompts and tool permissions. Which intervention most directly addresses the failure?
- Use a bounded agent loop with explicit tool budgets and stop conditions.
- Use an evaluator-optimizer workflow with a measurable completion criterion.
- Route inputs to specialized workflows based on validated intent and confidence.
- Run independent validations in parallel and aggregate the verified results.
Explanation
Correct Answer
C:Routing reduces irrelevant tool exposure and applies task-specific instructions to each request category. Test whether the selected level of autonomy actually achieves task-specific routing and permissions; extra agency is justified only by measurable gains in task-specific routing and permissions.
Incorrect Answers
A:This recommendation targets adaptive research with bounded autonomy at a different point in the workflow. It leaves task-specific routing and permissions outside the design contract and does not require the team to route inputs to specialized workflows based on validated intent and confidence.
B:This is a valid architectural concern when optimizing bounded iterative quality improvement. Here it leaves task-specific routing and permissions unaddressed, whereas the intended change must route inputs to specialized workflows based on validated intent and confidence before the case can be considered resolved.
D:By itself, this improves parallel work on independent units, but the case is not asking the architect to optimize that dimension. The requirement is task-specific routing and permissions, supported by an approach to route inputs to specialized workflows based on validated intent and confidence.
Question 28
An answer must meet a strict rubric and can be improved by one or two review passes before publishing. What should the solution architect prioritize?
- Use an evaluator-optimizer workflow with a measurable completion criterion.
- Eliminate the redundant model pass and retain targeted validation where risk warrants it.
- Combine agentic evidence collection with a deterministic transaction gate.
- Run independent validations in parallel and aggregate the verified results.
- Prefer a single model step with a narrow retrieval call over a multi-agent team.
Explanation
Correct Answer
A:A separate quality check can request targeted revisions; unlimited self-critique loops have diminishing value.
Incorrect Answers
B:For a workflow whose primary weakness was removing unjustified multi-step complexity, that proposal could be reasonable. Here the unmet obligation is bounded iterative quality improvement, and that requires the system to use an evaluator-optimizer workflow with a measurable completion criterion.
C:Although hybrid agent-and-rules boundaries can be valuable, it is not the acceptance condition in the question. The missing provision is bounded iterative quality improvement, addressed by the control to use an evaluator-optimizer workflow with a measurable completion criterion.
D:It is aimed at a different outcome, parallel work on independent units. No evidence is given that it provides bounded iterative quality improvement; the needed mechanism is to use an evaluator-optimizer workflow with a measurable completion criterion at the appropriate boundary.
E:It emphasizes low-overhead design for simple tasks without closing the stated gap in bounded iterative quality improvement. By contrast, the remedy to use an evaluator-optimizer workflow with a measurable completion criterion acts on the condition that makes this scenario fail.
Question 29
Ten independent product descriptions must be checked for prohibited claims by an afternoon deadline. Which architecture adjustment is most appropriate?
- Use an evaluator-optimizer workflow with a measurable completion criterion.
- Combine agentic evidence collection with a deterministic transaction gate.
- Prefer a single model step with a narrow retrieval call over a multi-agent team.
- Run independent validations in parallel and aggregate the verified results.
Explanation
Correct Answer
D:Parallelization reduces wall-clock latency when tasks do not depend on one another; aggregation must preserve per-item status.
Incorrect Answers
A:There is a legitimate use case for bounded iterative quality improvement, but it does not supply the test or enforcement needed for parallel work on independent units. The design needs to run independent validations in parallel and aggregate the verified results.
B:Consider its actual operating scope: hybrid agent-and-rules boundaries. The requested assurance is parallel work on independent units, which calls for an implementation that can run independent validations in parallel and aggregate the verified results, not merely for a related improvement.
C:The alternative could reduce risk associated with low-overhead design for simple tasks, yet the described behavior requires parallel work on independent units. That outcome depends on the capability to run independent validations in parallel and aggregate the verified results.
Question 30
A purchase approval agent occasionally chooses unnecessary tools although a fixed three-step sequence exists. Which next step is most defensible?
- Use a retrieval-augmented single-turn LLM design rather than autonomous agents.
- Implement a deterministic prompt-and-tool workflow with explicit stage transitions.
- Replace open-ended planning with a chained workflow for the known path.
- Use a bounded agent loop with explicit tool budgets and stop conditions.
Explanation
Correct Answer
C:Removing avoidable degrees of freedom makes the process easier to constrain, test and audit. Test whether the selected level of autonomy actually achieves preventing needless agent autonomy; extra agency is justified only by measurable gains in preventing needless agent autonomy.
Incorrect Answers
A:By itself, this improves bounded grounded question answering, but the case is not asking the architect to optimize that dimension. The requirement is preventing needless agent autonomy, supported by an approach to replace open-ended planning with a chained workflow for the known path.
B:Its main effect concerns predictable ordered processing. The case explicitly hinges on preventing needless agent autonomy; to resolve it reliably, the implementation has to replace open-ended planning with a chained workflow for the known path, which this choice does not establish.
D:The proposed action prioritizes adaptive research with bounded autonomy. That can coexist with the right solution, but it cannot replace preventing needless agent autonomy: the application must still replace open-ended planning with a chained workflow for the known path.