Question 1
An insurer wants to automate claim letters but cannot define what successful automation means. The sponsor only quotes model accuracy. What should the solution architect prioritize?
- Classify use cases and data sensitivity before choosing ingestion, deployment, and retention controls.
- Separate low-risk drafting from a gated closure workflow with explicit authorization and human sign-off.
- Define a business KPI, baseline process cost, and acceptable error budget before selecting an architecture.
- Name content owners and freshness SLAs as prerequisites for an answer architecture.
Explanation
Correct Answer
C:Outcome and risk targets determine whether assisted drafting or automation produces value; a generic model score cannot measure cycle-time savings or review cost.
Incorrect Answers
A:The relevant trade-off is not primarily data classification driving architecture. A successful fix needs measurable business acceptance criteria; otherwise measurable business acceptance criteria remains unverified even after this change.
B:The control would be justified if a safe boundary around irreversible actions were the limiting factor. In this scenario the decisive target is measurable business acceptance criteria; success requires an architecture that will define a business KPI, baseline process cost, and acceptable error budget before selecting an architecture.
D:This shifts attention toward ownership and freshness of source knowledge when the relevant decision is about measurable business acceptance criteria. Unless the architecture can define a business KPI, baseline process cost, and acceptable error budget before selecting an architecture, the failure described in the stem remains unresolved.
Question 2
A bank proposes an agent to approve loans even though approval policy is a fixed threshold table. Which design choice best matches the stated constraint?
- Add appealability, explainability and accountability to the use-case acceptance requirements.
- Collect representative locale-specific samples and acceptance criteria during discovery.
- Keep deterministic eligibility decisions in rules software and use Claude for explaining exceptions to reviewers.
- Reject unnecessary LLM use and enforce catalog constraints with conventional validation.
Explanation
Correct Answer
C:Rules provide auditable, repeatable policy decisions; language modeling can support narrative tasks without taking authority over credit adjudication.
Incorrect Answers
A:The design concern is real but different: accountable decisions and appeal processes. It does not impose the specific requirement for deterministic decisions for policy thresholds; a suitable solution must keep deterministic eligibility decisions in rules software and use Claude for explaining exceptions to reviewers and demonstrate the result.
B:This addresses language-specific requirements and coverage; however, the problem asks for deterministic decisions for policy thresholds. Without the safeguard to keep deterministic eligibility decisions in rules software and use Claude for explaining exceptions to reviewers, the original failure mode is still possible.
D:There is a legitimate use case for correct use of deterministic validation, but it does not supply the test or enforcement needed for deterministic decisions for policy thresholds. The design needs to keep deterministic eligibility decisions in rules software and use Claude for explaining exceptions to reviewers.
Question 3
A service desk processes repetitive emails, but one in twenty messages requests account closure with irreversible effects. Which architecture adjustment is most appropriate?
- Define a business KPI, baseline process cost, and acceptable error budget before selecting an architecture.
- Separate low-risk drafting from a gated closure workflow with explicit authorization and human sign-off.
- Add appealability, explainability and accountability to the use-case acceptance requirements.
- Distinguish informational output from authenticated actions in the business capability map.
Explanation
Correct Answer
B:Business-process mapping identifies irreversible steps; the model can prepare context while an authorized actor owns the state change.
Incorrect Answers
A:This recommendation targets measurable business acceptance criteria at a different point in the workflow. It leaves a safe boundary around irreversible actions outside the design contract and does not require the team to separate low-risk drafting from a gated closure workflow with explicit authorization and human sign-off.
C:This would make sense as an additional project activity for accountable decisions and appeal processes. The priority required here is a safe boundary around irreversible actions, so the workflow still needs a control that will separate low-risk drafting from a gated closure workflow with explicit authorization and human sign-off.
D:By itself, this improves transaction authorization requirements, but the case is not asking the architect to optimize that dimension. The requirement is a safe boundary around irreversible actions, supported by an approach to separate low-risk drafting from a gated closure workflow with explicit authorization and human sign-off.
Question 4
The executive team wants an enterprise assistant, but departments disagree about whether data is public, confidential, or regulated. Which next step is most defensible?
- Name content owners and freshness SLAs as prerequisites for an answer architecture.
- Classify use cases and data sensitivity before choosing ingestion, deployment, and retention controls.
- Add appealability, explainability and accountability to the use-case acceptance requirements.
- Define a business KPI, baseline process cost, and acceptable error budget before selecting an architecture.
Explanation
Correct Answer
B:A solution cannot promise compliant handling before knowing which categories of records enter prompts, indexes, logs and tools.
Incorrect Answers
A:This shifts attention toward ownership and freshness of source knowledge when the relevant decision is about data classification driving architecture. Unless the architecture can classify use cases and data sensitivity before choosing ingestion, deployment, and retention controls, the failure described in the stem remains unresolved.
C:Although accountable decisions and appeal processes can be valuable, it is not the acceptance condition in the question. The missing provision is data classification driving architecture, addressed by the control to classify use cases and data sensitivity before choosing ingestion, deployment, and retention controls.
D:It is aimed at a different outcome, measurable business acceptance criteria. No evidence is given that it provides data classification driving architecture; the needed mechanism is to classify use cases and data sensitivity before choosing ingestion, deployment, and retention controls at the appropriate boundary.
Question 5
A retailer plans customer-facing personalized advice with no evidence that customers want the suggested feature. Which approach should be recommended?
- Instrument time-on-task, correction burden and review latency against current workflow.
- Run a small, instrumented pilot against a non-AI baseline and agreed conversion metric.
- Define a business KPI, baseline process cost, and acceptable error budget before selecting an architecture.
- Name content owners and freshness SLAs as prerequisites for an answer architecture.
Explanation
Correct Answer
B:A feasibility trial measures whether the proposed user journey improves an actual outcome before enterprise integration spend.
Incorrect Answers
A:There is a legitimate use case for a defensible workflow productivity metric, but it does not supply the test or enforcement needed for evidence of user value before investment. The design needs to run a small, instrumented pilot against a non-AI baseline and agreed conversion metric.
C:The alternative could reduce risk associated with measurable business acceptance criteria, yet the described behavior requires evidence of user value before investment. That outcome depends on the capability to run a small, instrumented pilot against a non-AI baseline and agreed conversion metric.
D:The proposed measure is aimed at ownership and freshness of source knowledge, not evidence of user value before investment. The architecture must instead run a small, instrumented pilot against a non-AI baseline and agreed conversion metric; a separate change is still needed for evidence of user value before investment.
Question 6
The sales team proposes an LLM for overnight catalog validation whose rules are fully expressible as database constraints. What is the best production design response?
- Add appealability, explainability and accountability to the use-case acceptance requirements.
- Keep deterministic eligibility decisions in rules software and use Claude for explaining exceptions to reviewers.
- Classify use cases and data sensitivity before choosing ingestion, deployment, and retention controls.
- Reject unnecessary LLM use and enforce catalog constraints with conventional validation.
Explanation
Correct Answer
D:Schema constraints give exact outcomes and predictable costs for deterministic checks; generation adds uncertainty without meaningful benefit.
Incorrect Answers
A:By itself, this improves accountable decisions and appeal processes, but the case is not asking the architect to optimize that dimension. The requirement is correct use of deterministic validation, supported by an approach to reject unnecessary LLM use and enforce catalog constraints with conventional validation.
B:Its main effect concerns deterministic decisions for policy thresholds. The case explicitly hinges on correct use of deterministic validation; to resolve it reliably, the implementation has to reject unnecessary LLM use and enforce catalog constraints with conventional validation, which this choice does not establish.
C:The option might improve data classification driving architecture; that remains insufficient for correct use of deterministic validation. The case is resolved by correct use of deterministic validation, which depends on a design that will reject unnecessary LLM use and enforce catalog constraints with conventional validation.
Question 7
A legal research assistant drafts persuasive summaries but stakeholders require each statement linked to a current policy source. Which change most directly satisfies this requirement?
- Add appealability, explainability and accountability to the use-case acceptance requirements.
- Distinguish informational output from authenticated actions in the business capability map.
- Name content owners and freshness SLAs as prerequisites for an answer architecture.
- Treat grounded evidence and reviewability as first-class product requirements.
- Collect representative locale-specific samples and acceptance criteria during discovery.
Explanation
Correct Answer
D:Source-level evidence must be available in retrieval and UI design, not retrofitted by asking a model to sound confident.
Incorrect Answers
A:It is aimed at a different outcome, accountable decisions and appeal processes. No evidence is given that it provides traceable support for generated claims; the needed mechanism is to treat grounded evidence and reviewability as first-class product requirements at the appropriate boundary.
B:It emphasizes transaction authorization requirements without closing the stated gap in traceable support for generated claims. By contrast, the remedy to treat grounded evidence and reviewability as first-class product requirements acts on the condition that makes this scenario fail.
C:That is primarily a technique for ownership and freshness of source knowledge. It lacks the explicit measure for traceable support for generated claims that is necessary here, namely a system that will treat grounded evidence and reviewability as first-class product requirements.
E:This recommendation targets language-specific requirements and coverage at a different point in the workflow. It leaves traceable support for generated claims outside the design contract and does not require the team to treat grounded evidence and reviewability as first-class product requirements.
Question 8
A multilingual support project is scoped only using English chat transcripts although launch includes five languages. What design decision would best fit these facts?
- Keep deterministic eligibility decisions in rules software and use Claude for explaining exceptions to reviewers.
- Collect representative locale-specific samples and acceptance criteria during discovery.
- Treat grounded evidence and reviewability as first-class product requirements.
- Add appealability, explainability and accountability to the use-case acceptance requirements.
Explanation
Correct Answer
B:Language coverage must be tested using local terminology, scripts and user intents; English success does not establish multilingual fitness.
Incorrect Answers
A:The proposed measure is aimed at deterministic decisions for policy thresholds, not language-specific requirements and coverage. The architecture must instead collect representative locale-specific samples and acceptance criteria during discovery; a separate change is still needed for language-specific requirements and coverage.
C:The control would be justified if traceable support for generated claims were the limiting factor. In this scenario the decisive target is language-specific requirements and coverage; success requires an architecture that will collect representative locale-specific samples and acceptance criteria during discovery.
D:The option belongs to the broader work on accountable decisions and appeal processes, but it cannot by itself deliver language-specific requirements and coverage. A decision on this case needs the team to collect representative locale-specific samples and acceptance criteria during discovery and verify that outcome.
Question 9
A hospital pilot has strong clinician enthusiasm but procurement requires a quantified reduction in documentation effort. Which intervention most directly addresses the failure?
- Separate low-risk drafting from a gated closure workflow with explicit authorization and human sign-off.
- Treat grounded evidence and reviewability as first-class product requirements.
- Run a small, instrumented pilot against a non-AI baseline and agreed conversion metric.
- Instrument time-on-task, correction burden and review latency against current workflow.
Explanation
Correct Answer
D:Clinician satisfaction is useful qualitative evidence; procurement needs an attributable net time-saving after human verification.
Incorrect Answers
A:The proposed action prioritizes a safe boundary around irreversible actions. That can coexist with the right solution, but it cannot replace a defensible workflow productivity metric: the application must still instrument time-on-task, correction burden and review latency against current workflow.
B:The design concern is real but different: traceable support for generated claims. It does not impose the specific requirement for a defensible workflow productivity metric; a suitable solution must instrument time-on-task, correction burden and review latency against current workflow and demonstrate the result.
C:This addresses evidence of user value before investment; however, the problem asks for a defensible workflow productivity metric. Without the safeguard to instrument time-on-task, correction burden and review latency against current workflow, the original failure mode is still possible.
Question 10
A team assumes a proof of concept can write directly into a billing system because reviewers liked the generated text. What should the solution architect prioritize?
- Treat grounded evidence and reviewability as first-class product requirements.
- Define a business KPI, baseline process cost, and acceptable error budget before selecting an architecture.
- Distinguish informational output from authenticated actions in the business capability map.
- Separate low-risk drafting from a gated closure workflow with explicit authorization and human sign-off.
Explanation
Correct Answer
C:Drafting and financial posting have different failure consequences; the architecture must identify who may commit each side effect.
Incorrect Answers
A:Improving traceable support for generated claims does not imply transaction authorization requirements. The selected design must distinguish informational output from authenticated actions in the business capability map, whereas this choice omits that direct protection or verification step.
B:This addresses measurable business acceptance criteria; however, the problem asks for transaction authorization requirements. Without the safeguard to distinguish informational output from authenticated actions in the business capability map, the original failure mode is still possible.
D:This would make sense as an additional project activity for a safe boundary around irreversible actions. The priority required here is transaction authorization requirements, so the workflow still needs a control that will distinguish informational output from authenticated actions in the business capability map.