Microsoft AB-100 Agent Model and Prompt Acceptance Practice Test

Practice Exams:

View All

Microsoft AB-100 Agent Model and Prompt Acceptance Practice Test

Topic 11 covers Agent, Model and Prompt Acceptance for Microsoft AB-100, using the current skills measured as of July 22, 2026 and primary Microsoft documentation. For broader exam preparation, review the AB-100 Exam Dumps page. These are original practice questions, and every option includes a scenario-specific explanation.

Question 1

Before a wider rollout of agent, model and prompt acceptance, the model acceptance board reviews the current evidence. The current solution leaves a task-success oracle independent of generated wording unresolved, and the gap is now affecting the stated business or technical requirement. What should the architect recommend?

Define a task-success oracle independent of generated wording.
Treat training examples separately from validation examples and evaluate each with its own evidence.
Set acceptance thresholds before running the experiment.
Compare a custom model with a suitable baseline.
Require documented limitations before approving model use.

Correct Answer: A

Correct Answer

Answer A is correct because This is correct because it applies the reserved architecture decision—define a task-success oracle independent of generated wording—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘separate training examples from validation examples’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘set acceptance thresholds before running the experiment’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘compare a custom model with a suitable baseline’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘require documented limitations before approving model use’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 2

A pilot for agent, model and prompt acceptance reaches a decision point for the agent release council. The current solution leaves a representative evaluation set for business use unresolved, and the gap is now affecting the stated business or technical requirement. Which decision is most appropriate?

Select a representative evaluation set for business use.
Repeat stochastic trials where one run is unreliable.
Detect leakage from repeated customer records.
Compare prompts while holding model and data constant.
Include operational latency in acceptance criteria in the architecture or business-case boundary before comparing alternatives.

Correct Answer: A

Correct Answer

Answer A is correct because This is correct because it applies the reserved architecture decision—select a representative evaluation set for business use—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘repeat stochastic trials where one run is unreliable’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘detect leakage from repeated customer records’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘compare prompts while holding model and data constant’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘include operational latency in acceptance criteria’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 3

During a design workshop on agent, model and prompt acceptance, the model acceptance board identifies a constraint. The review has identified a specific issue: add rare high-impact failures to the test design. Which decision is most appropriate?

Select a business-relevant metric for asymmetric error costs.
Test required structured output against a schema.
Add rare high-impact failures to the test design.
Evaluate resource consumption under expected load.
Test degradation when a dependency is unavailable.

Correct Answer: C

Correct Answer

Answer C is correct because This is correct because it applies the reserved architecture decision—add rare high-impact failures to the test design—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘select a business-relevant metric for asymmetric error costs’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘test required structured output against a schema’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate resource consumption under expected load’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘test degradation when a dependency is unavailable’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 4

A production-readiness review of agent, model and prompt acceptance gives the prompt quality team new evidence. Existing evidence is insufficient to make a safe release or architecture decision about multi-turn behavior beyond isolated prompts. Which action should the team take next?

Run a controlled evaluation set with explicit acceptance thresholds for task completion, quality, safety, and tool behavior.
Evaluate minority cohorts hidden by aggregate performance.
Define rejection criteria for unsupported input types.
Evaluate escalation accuracy for unresolved cases.
Evaluate multi-turn behavior beyond isolated prompts.

Correct Answer: E

Correct Answer

Answer E is correct because This is correct because it applies the reserved architecture decision—evaluate multi-turn behavior beyond isolated prompts—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is a useful verification step after the primary architecture decision, but it does not by itself resolve the decision the scenario asks for. The design choice must be made first, and this evidence can then confirm that the chosen approach behaves as intended.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate minority cohorts hidden by aggregate performance’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘define rejection criteria for unsupported input types’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate escalation accuracy for unresolved cases’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 5

A pilot for agent, model and prompt acceptance reaches a decision point for the prompt quality team. The review has identified a specific issue: measure tool selection accuracy separately from final answer fluency. Which action should the team take next?

Assess robustness to a plausible input shift.
Test prompt robustness to irrelevant appended content.
Require documented limitations before approving model use.
Calibrate automated judgments with reviewed examples.
Measure tool selection accuracy separately from final answer fluency.

Correct Answer: E

Correct Answer

Answer E is correct because This is correct because it applies the reserved architecture decision—measure tool selection accuracy separately from final answer fluency—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘assess robustness to a plausible input shift’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘test prompt robustness to irrelevant appended content’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘require documented limitations before approving model use’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘calibrate automated judgments with reviewed examples’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 6

An AI validation group is reviewing agent, model and prompt acceptance. Existing evidence is insufficient to make a safe release or architecture decision about refusal when the requested action exceeds permitted scope. Which action should the team take next?

Compare prompts while holding model and data constant.
Validate examples against unintended output bias.
Test refusal when the requested action exceeds permitted scope.
Define a model acceptance threshold for the actual task.
Validate calibration before treating scores as confidence.

Correct Answer: C

Correct Answer

Answer C is correct because This is correct because it applies the reserved architecture decision—test refusal when the requested action exceeds permitted scope—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘compare prompts while holding model and data constant’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘validate examples against unintended output bias’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘define a model acceptance threshold for the actual task’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘validate calibration before treating scores as confidence’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 7

A pilot for agent, model and prompt acceptance reaches a decision point for the prompt quality team. Existing evidence is insufficient to make a safe release or architecture decision about results against a non-agent process baseline. Which action should the team take next?

Compare results against a non-agent process baseline.
Compare a custom model with a suitable baseline.
Check instructions on conflicting business evidence.
Treat training examples separately from validation examples and evaluate each with its own evidence.
Test required structured output against a schema.

Correct Answer: A

Correct Answer

Answer A is correct because This is correct because it applies the reserved architecture decision—compare results against a non-agent process baseline—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘compare a custom model with a suitable baseline’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘check instructions on conflicting business evidence’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘separate training examples from validation examples’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘test required structured output against a schema’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 8

Before a wider rollout of agent, model and prompt acceptance, the model acceptance board reviews the current evidence. The next stage is running the experiment, but acceptance thresholds has not yet been made explicit or testable. Which decision is most appropriate?

Evaluate behavior with the stated constraint—essential context is missing—before committing the design.
Run a controlled evaluation set with explicit acceptance thresholds for task completion, quality, safety, and tool behavior.
Set acceptance thresholds before running the experiment.
Include operational latency in acceptance criteria in the architecture or business-case boundary before comparing alternatives.
Detect leakage from repeated customer records.

Correct Answer: C

Correct Answer

Answer C is correct because This is correct because it applies the reserved architecture decision—set acceptance thresholds before running the experiment—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate behavior when essential context is missing’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is a useful verification step after the primary architecture decision, but it does not by itself resolve the decision the scenario asks for. The design choice must be made first, and this evidence can then confirm that the chosen approach behaves as intended.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘include operational latency in acceptance criteria’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘detect leakage from repeated customer records’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 9

A production-readiness review of agent, model and prompt acceptance gives the prompt quality team new evidence. The review has identified a specific issue: repeat stochastic trials where one run is unreliable. What should the architect recommend?

Evaluate resource consumption under expected load.
Select a business-relevant metric for asymmetric error costs.
Test prompt robustness to irrelevant appended content.
Repeat stochastic trials where one run is unreliable.
Evaluate translation without losing critical business terms.

Correct Answer: D

Correct Answer

Answer D is correct because This is correct because it applies the reserved architecture decision—repeat stochastic trials where one run is unreliable—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate resource consumption under expected load’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘select a business-relevant metric for asymmetric error costs’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘test prompt robustness to irrelevant appended content’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate translation without losing critical business terms’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 10

Before a wider rollout of agent, model and prompt acceptance, the safety evaluation program reviews the current evidence. Existing evidence is insufficient to make a safe release or architecture decision about degradation when a dependency is unavailable. Which approach best fits the requirement?

Verify uncertainty wording under weak evidence.
Validate examples against unintended output bias.
Define rejection criteria for unsupported input types.
Evaluate minority cohorts hidden by aggregate performance.
Test degradation when a dependency is unavailable.

Correct Answer: E

Correct Answer

Answer E is correct because This is correct because it applies the reserved architecture decision—test degradation when a dependency is unavailable—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘verify uncertainty wording under weak evidence’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate minority cohorts hidden by aggregate performance’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 11

A production-readiness check identifies escalation accuracy for unresolved cases as the remaining risk. Which action best reduces that risk?

Require documented limitations before approving model use.
Check instructions on conflicting business evidence.
Check output stability for equivalent user formulations.
Evaluate escalation accuracy for unresolved cases.
Assess robustness to a plausible input shift.

Correct Answer: D

Correct Answer

Answer D is correct because This is correct because it applies the reserved architecture decision—evaluate escalation accuracy for unresolved cases—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘require documented limitations before approving model use’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘check instructions on conflicting business evidence’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘check output stability for equivalent user formulations’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘assess robustness to a plausible input shift’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 12

During a design workshop on agent, model and prompt acceptance, the model acceptance board identifies a constraint. The review has identified a specific issue: calibrate automated judgments with reviewed examples. What should the architect recommend?

Calibrate automated judgments with reviewed examples.
Validate calibration before treating scores as confidence.
Run a controlled evaluation set with explicit acceptance thresholds for task completion, quality, safety, and tool behavior.
Compare prompts while holding model and data constant.
Test a prompt on a deliberately out-of-scope task.

Correct Answer: A

Correct Answer

Answer A is correct because This is correct because it applies the reserved architecture decision—calibrate automated judgments with reviewed examples—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘validate calibration before treating scores as confidence’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is a useful verification step after the primary architecture decision, but it does not by itself resolve the decision the scenario asks for. The design choice must be made first, and this evidence can then confirm that the chosen approach behaves as intended.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘test a prompt on a deliberately out-of-scope task’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 13

An agent release council is reviewing agent, model and prompt acceptance. The current solution leaves a model acceptance threshold for the actual task unresolved, and the gap is now affecting the stated business or technical requirement. What should the architect recommend?

Test required structured output against a schema.
Compare a custom model with a suitable baseline.
Evaluate translation without losing critical business terms.
Define a model acceptance threshold for the actual task.
Define a task-success oracle independent of generated wording.

Correct Answer: D

Correct Answer

Answer D is correct because This is correct because it applies the reserved architecture decision—define a model acceptance threshold for the actual task—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘test required structured output against a schema’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate translation without losing critical business terms’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘define a task-success oracle independent of generated wording’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 14

An agent release council is reviewing agent, model and prompt acceptance. Stakeholders are treating training examples and validation examples as the same decision, which is obscuring the actual control boundary. Which approach best fits the requirement?

Include operational latency in acceptance criteria in the architecture or business-case boundary before comparing alternatives.
Separate training examples from validation examples.
Select a representative evaluation set for business use.
Verify uncertainty wording under weak evidence.
Evaluate behavior with the stated constraint—essential context is missing—before committing the design.

Correct Answer: B

Correct Answer

Answer B is correct because This is correct because it applies the reserved architecture decision—separate training examples from validation examples—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘include operational latency in acceptance criteria’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘select a representative evaluation set for business use’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘verify uncertainty wording under weak evidence’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate behavior when essential context is missing’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 15

The prompt quality team is preparing an architecture decision record for agent, model and prompt acceptance. The current solution leaves leakage from repeated customer records unresolved, and the gap is now affecting the stated business or technical requirement. What is the best design choice?

Add rare high-impact failures to the test design.
Test prompt robustness to irrelevant appended content.
Evaluate resource consumption under expected load.
Detect leakage from repeated customer records.
Check output stability for equivalent user formulations.

Correct Answer: D

Correct Answer

Answer D is correct because This is correct because it applies the reserved architecture decision—detect leakage from repeated customer records—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘add rare high-impact failures to the test design’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate resource consumption under expected load’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘check output stability for equivalent user formulations’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 16

During a design workshop on agent, model and prompt acceptance, the model acceptance board identifies a constraint. The current solution leaves a business-relevant metric for asymmetric error costs unresolved, and the gap is now affecting the stated business or technical requirement. Which decision is most appropriate?

Select a business-relevant metric for asymmetric error costs.
Run a controlled evaluation set with explicit acceptance thresholds for task completion, quality, safety, and tool behavior.
Reject a prompt improvement that harms a critical scenario.
Validate examples against unintended output bias.
Define rejection criteria for unsupported input types.

Correct Answer: A

Correct Answer

Answer A is correct because This is correct because it applies the reserved architecture decision—select a business-relevant metric for asymmetric error costs—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘reject a prompt improvement that harms a critical scenario’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘validate examples against unintended output bias’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘define rejection criteria for unsupported input types’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 17

The agent release council is preparing an architecture decision record for agent, model and prompt acceptance. Existing evidence is insufficient to make a safe release or architecture decision about minority cohorts hidden by aggregate performance. What should the architect recommend?

Evaluate minority cohorts hidden by aggregate performance.
Require documented limitations before approving model use.
Define a task-success oracle independent of generated wording.
Check instructions on conflicting business evidence.
Measure tool selection accuracy separately from final answer fluency.

Correct Answer: A

Correct Answer

Answer A is correct because This is correct because it applies the reserved architecture decision—evaluate minority cohorts hidden by aggregate performance—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘require documented limitations before approving model use’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘define a task-success oracle independent of generated wording’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘check instructions on conflicting business evidence’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘measure tool selection accuracy separately from final answer fluency’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 18

A safety evaluation program is reviewing agent, model and prompt acceptance. Existing evidence is insufficient to make a safe release or architecture decision about robustness to a plausible input shift. What should the architect recommend?

Select a representative evaluation set for business use.
Test a prompt on a deliberately out-of-scope task.
Assess robustness to a plausible input shift.
Test refusal when the requested action exceeds permitted scope.
Compare prompts while holding model and data constant.

Correct Answer: C

Correct Answer

Answer C is correct because This is correct because it applies the reserved architecture decision—assess robustness to a plausible input shift—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘select a representative evaluation set for business use’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘test a prompt on a deliberately out-of-scope task’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘test refusal when the requested action exceeds permitted scope’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘compare prompts while holding model and data constant’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 19

A pilot for agent, model and prompt acceptance reaches a decision point for the agent release council. Existing evidence is insufficient to make a safe release or architecture decision about calibration before treating scores as confidence. What should the architect recommend?

Add rare high-impact failures to the test design.
Validate calibration before treating scores as confidence.
Test required structured output against a schema.
Evaluate translation without losing critical business terms.
Compare results against a non-agent process baseline.

Correct Answer: B

Correct Answer

Answer B is correct because This is correct because it applies the reserved architecture decision—validate calibration before treating scores as confidence—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘test required structured output against a schema’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate translation without losing critical business terms’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘compare results against a non-agent process baseline’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 20

The AI validation group is preparing an architecture decision record for agent, model and prompt acceptance. Existing evidence is insufficient to make a safe release or architecture decision about a custom model with a suitable baseline. Which action should the team take next?

Evaluate behavior with the stated constraint—essential context is missing—before committing the design.
Compare a custom model with a suitable baseline.
Run a controlled evaluation set with explicit acceptance thresholds for task completion, quality, safety, and tool behavior.
Verify uncertainty wording under weak evidence.
Do not promote a model, agent, or prompt merely because average quality improved; critical acceptance thresholds must still pass.

Correct Answer: B

Correct Answer

Answer B is correct because This is correct because it applies the reserved architecture decision—compare a custom model with a suitable baseline—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer E is incorrect because This safeguard can be important in a broader production design, but it is a supporting control rather than the primary decision required by this scenario. Selecting it alone would leave the core architecture choice unresolved.

Question 21

Before a wider rollout of agent, model and prompt acceptance, the agent release council reviews the current evidence. The review has identified a specific issue: include operational latency in acceptance criteria. What should the architect recommend?

Repeat stochastic trials where one run is unreliable.
Check output stability for equivalent user formulations.
Include operational latency in acceptance criteria.
Test prompt robustness to irrelevant appended content.
Measure tool selection accuracy separately from final answer fluency.

Correct Answer: C

Correct Answer

Answer C is correct because This is correct because it applies the reserved architecture decision—include operational latency in acceptance criteria—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘repeat stochastic trials where one run is unreliable’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘check output stability for equivalent user formulations’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘test prompt robustness to irrelevant appended content’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 22

An earlier design assumed the wrong thing about resource consumption under expected load. Which decision should replace that assumption?

Reject a prompt improvement that harms a critical scenario.
Evaluate resource consumption under expected load.
Test degradation when a dependency is unavailable.
Validate examples against unintended output bias.
Test refusal when the requested action exceeds permitted scope.

Correct Answer: B

Correct Answer

Answer B is correct because This is correct because it applies the reserved architecture decision—evaluate resource consumption under expected load—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘reject a prompt improvement that harms a critical scenario’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘test degradation when a dependency is unavailable’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘test refusal when the requested action exceeds permitted scope’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 23

A safety evaluation program is reviewing agent, model and prompt acceptance. The current solution leaves rejection criteria for unsupported input types unresolved, and the gap is now affecting the stated business or technical requirement. Which approach best fits the requirement?

Evaluate escalation accuracy for unresolved cases.
Define a task-success oracle independent of generated wording.
Compare results against a non-agent process baseline.
Check instructions on conflicting business evidence.
Define rejection criteria for unsupported input types.

Correct Answer: E

Correct Answer

Answer E is correct because This is correct because it applies the reserved architecture decision—define rejection criteria for unsupported input types—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate escalation accuracy for unresolved cases’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘define a task-success oracle independent of generated wording’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘compare results against a non-agent process baseline’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 24

During a design workshop on agent, model and prompt acceptance, the AI validation group identifies a constraint. The current solution leaves documented limitations before approving model use unresolved, and the gap is now affecting the stated business or technical requirement. Which decision is most appropriate?

Set acceptance thresholds before running the experiment.
Test a prompt on a deliberately out-of-scope task.
Run a controlled evaluation set with explicit acceptance thresholds for task completion, quality, safety, and tool behavior.
Select a representative evaluation set for business use.
Require documented limitations before approving model use.

Correct Answer: E

Correct Answer

Answer E is correct because This is correct because it applies the reserved architecture decision—require documented limitations before approving model use—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘set acceptance thresholds before running the experiment’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘select a representative evaluation set for business use’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 25

Two prompt variants are being compared for the same agent. The model deployment, grounding dataset, evaluation set, and scoring rubric are unchanged so the team can isolate whether the prompt itself improves task completion. Which evaluation approach should the architect use?

Add rare high-impact failures to the test design.
Compare prompts while holding model and data constant.
Define a model acceptance threshold for the actual task.
Evaluate translation without losing critical business terms.
Repeat stochastic trials where one run is unreliable.

Correct Answer: B

Correct Answer

Answer B is correct because This is correct because it applies the reserved architecture decision—compare prompts while holding model and data constant—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘define a model acceptance threshold for the actual task’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘repeat stochastic trials where one run is unreliable’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 26

A production-readiness check identifies required structured output against a schema as the remaining risk. Which action best reduces that risk?

Evaluate multi-turn behavior beyond isolated prompts.
Test degradation when a dependency is unavailable.
Test required structured output against a schema.
Verify uncertainty wording under weak evidence.
Treat training examples separately from validation examples and evaluate each with its own evidence.

Correct Answer: C

Correct Answer

Answer C is correct because This is correct because it applies the reserved architecture decision—test required structured output against a schema—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate multi-turn behavior beyond isolated prompts’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘test degradation when a dependency is unavailable’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘separate training examples from validation examples’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 27

A production-readiness review of agent, model and prompt acceptance gives the AI validation group new evidence. The team has confirmed that essential context is missing. It must now decide how to proceed. Which approach best fits the requirement?

Check output stability for equivalent user formulations.
Evaluate behavior when essential context is missing.
Measure tool selection accuracy separately from final answer fluency.
Detect leakage from repeated customer records.
Evaluate escalation accuracy for unresolved cases.

Correct Answer: B

Correct Answer

Answer B is correct because This is correct because it applies the reserved architecture decision—evaluate behavior when essential context is missing—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘check output stability for equivalent user formulations’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘measure tool selection accuracy separately from final answer fluency’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘detect leakage from repeated customer records’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate escalation accuracy for unresolved cases’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 28

A prompt passes normal test cases, but reviewers are concerned that harmless extra text appended to a user request might cause the agent to ignore key instructions. Which validation should be added before release?

Test refusal when the requested action exceeds permitted scope.
Calibrate automated judgments with reviewed examples.
Reject a prompt improvement that harms a critical scenario.
Run a controlled evaluation set with explicit acceptance thresholds for task completion, quality, safety, and tool behavior.
Test prompt robustness to irrelevant appended content.

Correct Answer: E

Correct Answer

Answer E is correct because This is correct because it applies the reserved architecture decision—test prompt robustness to irrelevant appended content—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘test refusal when the requested action exceeds permitted scope’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘calibrate automated judgments with reviewed examples’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is a useful verification step after the primary architecture decision, but it does not by itself resolve the decision the scenario asks for. The design choice must be made first, and this evidence can then confirm that the chosen approach behaves as intended.

Question 29

The operating model needs an explicit rule for examples against unintended output bias. Which design decision should be recorded?

Compare results against a non-agent process baseline.
Validate examples against unintended output bias.
Evaluate minority cohorts hidden by aggregate performance.
Define a task-success oracle independent of generated wording.
Define a model acceptance threshold for the actual task.

Correct Answer: B

Correct Answer

Answer B is correct because This is correct because it applies the reserved architecture decision—validate examples against unintended output bias—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘compare results against a non-agent process baseline’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate minority cohorts hidden by aggregate performance’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘define a task-success oracle independent of generated wording’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘define a model acceptance threshold for the actual task’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 30

Before a wider rollout of agent, model and prompt acceptance, the AI validation group reviews the current evidence. The review has identified a specific issue: check instructions on conflicting business evidence. Which approach best fits the requirement?

Check instructions on conflicting business evidence.
Treat training examples separately from validation examples and evaluate each with its own evidence.
Set acceptance thresholds before running the experiment.
Assess robustness to a plausible input shift.
Select a representative evaluation set for business use.

Correct Answer: A

Correct Answer

Answer A is correct because This is correct because it applies the reserved architecture decision—check instructions on conflicting business evidence—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘assess robustness to a plausible input shift’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘select a representative evaluation set for business use’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 31

During a design workshop on agent, model and prompt acceptance, the agent release council identifies a constraint. Existing evidence is insufficient to make a safe release or architecture decision about a prompt on a deliberately out-of-scope task. What should the architect recommend?

Repeat stochastic trials where one run is unreliable.
Validate calibration before treating scores as confidence.
Add rare high-impact failures to the test design.
Detect leakage from repeated customer records.
Test a prompt on a deliberately out-of-scope task.

Correct Answer: E

Correct Answer

Answer E is correct because This is correct because it applies the reserved architecture decision—test a prompt on a deliberately out-of-scope task—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘add rare high-impact failures to the test design’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 32

Before a wider rollout of agent, model and prompt acceptance, the agent release council reviews the current evidence. Existing evidence is insufficient to make a safe release or architecture decision about translation without losing critical business terms. What should the architect recommend?

Evaluate translation without losing critical business terms.
Run a controlled evaluation set with explicit acceptance thresholds for task completion, quality, safety, and tool behavior.
Select a business-relevant metric for asymmetric error costs.
Evaluate multi-turn behavior beyond isolated prompts.
Test degradation when a dependency is unavailable.

Correct Answer: A

Correct Answer

Answer A is correct because This is correct because it applies the reserved architecture decision—evaluate translation without losing critical business terms—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘select a business-relevant metric for asymmetric error costs’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate multi-turn behavior beyond isolated prompts’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 33

The model acceptance board is preparing an architecture decision record for agent, model and prompt acceptance. The review has identified a specific issue: verify uncertainty wording under weak evidence. What should the architect recommend?

Evaluate minority cohorts hidden by aggregate performance.
Include operational latency in acceptance criteria in the architecture or business-case boundary before comparing alternatives.
Measure tool selection accuracy separately from final answer fluency.
Evaluate escalation accuracy for unresolved cases.
Verify uncertainty wording under weak evidence.

Correct Answer: E

Correct Answer

Answer E is correct because This is correct because it applies the reserved architecture decision—verify uncertainty wording under weak evidence—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate minority cohorts hidden by aggregate performance’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘include operational latency in acceptance criteria’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 34

A production-readiness review of agent, model and prompt acceptance gives the AI validation group new evidence. The review has identified a specific issue: check output stability for equivalent user formulations. Which decision is most appropriate?

Assess robustness to a plausible input shift.
Test refusal when the requested action exceeds permitted scope.
Calibrate automated judgments with reviewed examples.
Check output stability for equivalent user formulations.
Evaluate resource consumption under expected load.

Correct Answer: D

Correct Answer

Answer D is correct because This is correct because it applies the reserved architecture decision—check output stability for equivalent user formulations—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer B is incorrect because This is not the best choice because it addresses the adjacent decision ‘test refusal when the requested action exceeds permitted scope’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer C is incorrect because This is not the best choice because it addresses the adjacent decision ‘calibrate automated judgments with reviewed examples’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer E is incorrect because This is not the best choice because it addresses the adjacent decision ‘evaluate resource consumption under expected load’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Question 35

An AI validation group is reviewing agent, model and prompt acceptance. The review has identified a specific issue: reject a prompt improvement that harms a critical scenario. What is the best design choice?

Validate calibration before treating scores as confidence.
Reject a prompt improvement that harms a critical scenario.
Define a model acceptance threshold for the actual task.
Define rejection criteria for unsupported input types.
Compare results against a non-agent process baseline.

Correct Answer: B

Correct Answer

Answer B is correct because This is correct because it applies the reserved architecture decision—reject a prompt improvement that harms a critical scenario—to the decisive constraint in the scenario. It keeps the action at the appropriate agent, model and prompt acceptance boundary and produces a result that can be verified before wider deployment.

Incorrect Answers

Answer A is incorrect because This is not the best choice because it addresses the adjacent decision ‘validate calibration before treating scores as confidence’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Answer D is incorrect because This is not the best choice because it addresses the adjacent decision ‘define rejection criteria for unsupported input types’ rather than the scenario’s decisive requirement. Although that action can be valid within agent, model and prompt acceptance, selecting it here would leave the stated constraint unresolved or move the design to the wrong stage.

Related posts: