Practice Exams:

Microsoft AI-103 Dumps

EXAM: Developing AI Apps and Agents on Azure
Verified By Experts
Premium Bundle
$49.99

AI-103 Premium ETE File


  • Premium File 187 Questions & Answers
Whats Included:
  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates

Microsoft AI-103 Practice Test 50

Question 490

During a architecture migration, a quality assurance lead must accept a change for rag-based agent citation. The issue to be addressed is: An agent answers from retrieved references without linking the supporting passage. Which concrete verification action demonstrates that the relevant behavior works?

  1. Return source metadata tied to each important retrieved claim
  2. Attach a plausible but unrelated URL to make an answer appear verified
  3. Seed an expiring session and verify deletion across dependent stores
  4. Sample high-impact claims and inspect corresponding source spans

Explanation

Correct Answer

D:

To verify this change, the education technology company should sample high-impact claims and inspect corresponding source spans. Operators still need the implementation to return source metadata tied to each important retrieved claim and the preventive requirement to enforce claim-to-source checks in agent evaluations.

Incorrect Answers

A:

The action to return source metadata tied to each important retrieved claim can contribute to rag-based agent citation as a implementation step; here, an agent answers from retrieved references without linking the supporting passage. In the education technology company acceptance review, trying to return source metadata tied to each important retrieved claim does not test the exact rag-based agent citation condition where an agent answers from retrieved references without linking the supporting passage; the relevant test is to sample high-impact claims and inspect corresponding source spans.

B:

In the education technology company acceptance review, trying to attach a plausible but unrelated URL to make an answer appear verified does not test the exact rag-based agent citation condition where an agent answers from retrieved references without linking the supporting passage; the relevant test is to sample high-impact claims and inspect corresponding source spans. Acceptance should instead demonstrate that the service can sample high-impact claims and inspect corresponding source spans, using the described failure as the test criterion.

C:

For memory retention for agents, it can be sensible to seed an expiring session and verify deletion across dependent stores if an agent stores user conversations indefinitely despite retention obligations. Without the actual change to sample high-impact claims and inspect corresponding source spans, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.

Question 491

At a public library network, the AI platform engineer is preparing a source update. The concrete requirement is: A planner agent passes incomplete instructions to an executor agent. Which design action BEST addresses the stated need?

  1. Define typed handoff payloads with task limits and verified evidence
  2. Submit a missing-authorization handoff and confirm it is rejected
  3. Send an unstructured free-text plan and assume the executor knows permissions
  4. Declare strict parameter names types and required fields for callable functions

Explanation

Correct Answer

A:

The proposed implementation is to define typed handoff payloads with task limits and verified evidence; this matches the operational requirement that a planner agent passes incomplete instructions to an executor agent. The team can submit a missing-authorization handoff and confirm it is rejected. The team should also require independent validation at each agent boundary.

Incorrect Answers

B:

The action to submit a missing-authorization handoff and confirm it is rejected can contribute to multi-agent handoff schema as a targeted acceptance check step; here, a planner agent passes incomplete instructions to an executor agent. Choosing to submit a missing-authorization handoff and confirm it is rejected cannot directly replace the specified action to define typed handoff payloads with task limits and verified evidence when a planner agent passes incomplete instructions to an executor agent.

C:

Choosing to send an unstructured free-text plan and assume the executor knows permissions cannot directly replace the specified action to define typed handoff payloads with task limits and verified evidence when a planner agent passes incomplete instructions to an executor agent. The actual implementation must define typed handoff payloads with task limits and verified evidence, because the stated service condition is that a planner agent passes incomplete instructions to an executor agent.

D:

Choosing to declare strict parameter names types and required fields for callable functions would direct effort toward typed tool function, whose characteristic issue is that an agent invents API parameters while calling an order-management tool. The relevant corrective or preventive action here remains to define typed handoff payloads with task limits and verified evidence, with its own validation criteria.

Question 492

After a cost review at a software consultancy, the team observes this failure: A planner agent passes incomplete instructions to an executor agent. Which flawed assumption or implementation choice would MOST DIRECTLY explain why the intended outcome was not achieved?

  1. Submit a missing-authorization handoff and confirm it is rejected
  2. Execute the payment tool while merely displaying an approval request
  3. Send an unstructured free-text plan and assume the executor knows permissions
  4. Define typed handoff payloads with task limits and verified evidence

Explanation

Correct Answer

C:

This design mistake is to send an unstructured free-text plan and assume the executor knows permissions; it conflicts with the actual multi-agent handoff schema behavior needed when a planner agent passes incomplete instructions to an executor agent. The software consultancy should instead define typed handoff payloads with task limits and verified evidence, followed by a check to submit a missing-authorization handoff and confirm it is rejected.

Incorrect Answers

A:

The alternative to submit a missing-authorization handoff and confirm it is rejected is a targeted acceptance check step rather than the faulty assumption being diagnosed. This software consultancy failure-analysis alternative of submit a missing-authorization handoff and confirm it is rejected does not diagnose the particular multi-agent handoff schema defect where a planner agent passes incomplete instructions to an executor agent; the keyed mistake is send an unstructured free-text plan and assume the executor knows permissions.

B:

The option to execute the payment tool while merely displaying an approval request would normally serve a different sequential approval workflow requirement, where a procurement agent must validate a request before seeking human purchase approval. The alternative may be useful in its own context, but it does not directly supply send an unstructured free-text plan and assume the executor knows permissions for the stated decision.

D:

The alternative to define typed handoff payloads with task limits and verified evidence is a implementation step rather than the faulty assumption being diagnosed. This software consultancy failure-analysis alternative of define typed handoff payloads with task limits and verified evidence does not diagnose the particular multi-agent handoff schema defect where a planner agent passes incomplete instructions to an executor agent; the keyed mistake is send an unstructured free-text plan and assume the executor knows permissions.

Question 493

The regional hospital is responding to a reliability exercise. A product support lead discovers this issue: A planner agent passes incomplete instructions to an executor agent. What corrective step should the team take to restore the intended functionality most directly?

  1. Submit a missing-authorization handoff and confirm it is rejected
  2. Repair handoff fields and revalidate the execution scope
  3. Require independent validation at each agent boundary
  4. Send an unstructured free-text plan and assume the executor knows permissions
  5. Define typed handoff payloads with task limits and verified evidence
  6. Repair workflow prompts and tool permissions from traced failed cases

Explanation

Correct Answer

B:

The direct repair is to repair handoff fields and revalidate the execution scope. Once the condition is corrected, the regional hospital should submit a missing-authorization handoff and confirm it is rejected. To reduce recurrence, it should also require independent validation at each agent boundary; those steps provide different recovery and prevention functions.

Incorrect Answers

A:

The action to submit a missing-authorization handoff and confirm it is rejected can contribute to multi-agent handoff schema as a targeted acceptance check step; here, a planner agent passes incomplete instructions to an executor agent. During regional hospital remediation, the proposed step to submit a missing-authorization handoff and confirm it is rejected is not the direct restoration action for multi-agent handoff schema when a planner agent passes incomplete instructions to an executor agent; the keyed repair is repair handoff fields and revalidate the execution scope.

C:

The action to require independent validation at each agent boundary can contribute to multi-agent handoff schema as a ongoing preventive rule step; here, a planner agent passes incomplete instructions to an executor agent. During regional hospital remediation, the proposed step to require independent validation at each agent boundary is not the direct restoration action for multi-agent handoff schema when a planner agent passes incomplete instructions to an executor agent; the keyed repair is repair handoff fields and revalidate the execution scope.

D:

During regional hospital remediation, the proposed step to send an unstructured free-text plan and assume the executor knows permissions is not the direct restoration action for multi-agent handoff schema when a planner agent passes incomplete instructions to an executor agent; the keyed repair is repair handoff fields and revalidate the execution scope. After remediation the team must submit a missing-authorization handoff and confirm it is rejected; the action in this option does not itself establish that target recovery outcome.

E:

The action to define typed handoff payloads with task limits and verified evidence can contribute to multi-agent handoff schema as a implementation step; here, a planner agent passes incomplete instructions to an executor agent. During regional hospital remediation, the proposed step to define typed handoff payloads with task limits and verified evidence is not the direct restoration action for multi-agent handoff schema when a planner agent passes incomplete instructions to an executor agent; the keyed repair is repair handoff fields and revalidate the execution scope.

F:

For agent evaluation data, it can be sensible to repair workflow prompts and tool permissions from traced failed cases if an agent seems helpful but its actions frequently violate workflow constraints. Without the actual change to repair handoff fields and revalidate the execution scope, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.

Question 494

A data engineer at a insurance platform wants the following issue to be prevented in later deployments: A planner agent passes incomplete instructions to an executor agent. Which enduring control or engineering guardrail BEST prevents recurrence?

  1. Send an unstructured free-text plan and assume the executor knows permissions
  2. Require independent validation at each agent boundary
  3. Require operation-level access checks independent of language model text
  4. Repair handoff fields and revalidate the execution scope

Explanation

Correct Answer

B:

Standardizing the boundary to require independent validation at each agent boundary is preventive because the same multi-agent handoff schema failure could otherwise return: a planner agent passes incomplete instructions to an executor agent. For this insurance platform, the service should also define typed handoff payloads with task limits and verified evidence, and tests should submit a missing-authorization handoff and confirm it is rejected so that the ongoing requirement is measurable.

Incorrect Answers

A:

For prevention at the insurance platform, the measure to send an unstructured free-text plan and assume the executor knows permissions is weaker or aimed at another issue than the multi-agent handoff schema guardrail to require independent validation at each agent boundary needed when a planner agent passes incomplete instructions to an executor agent. The lasting guardrail for this specific threat is to require independent validation at each agent boundary, not simply to complete an adjacent service task.

C:

The action to require operation-level access checks independent of language model text addresses agent function authorization rather than the decisive multi-agent handoff schema condition. The insurance platform should not infer that fixing one adjacent AI service validates its unrelated workflow; the requested action is to require independent validation at each agent boundary.

D:

The action to repair handoff fields and revalidate the execution scope can contribute to multi-agent handoff schema as a corrective remediation step; here, a planner agent passes incomplete instructions to an executor agent. For prevention at the insurance platform, the measure to repair handoff fields and revalidate the execution scope is weaker or aimed at another issue than the multi-agent handoff schema guardrail to require independent validation at each agent boundary needed when a planner agent passes incomplete instructions to an executor agent.

Question 495

During a cost review, a service administrator must accept a change for multi-agent handoff schema. The issue to be addressed is: A planner agent passes incomplete instructions to an executor agent. Which concrete verification action demonstrates that the relevant behavior works?

  1. Simulate a dropped response and confirm exactly one committed purchase
  2. Send an unstructured free-text plan and assume the executor knows permissions
  3. Submit a missing-authorization handoff and confirm it is rejected
  4. Require independent validation at each agent boundary
  5. Define typed handoff payloads with task limits and verified evidence

Explanation

Correct Answer

C:

The targeted acceptance test is to submit a missing-authorization handoff and confirm it is rejected. The design should first define typed handoff payloads with task limits and verified evidence. A robust media platform deployment also needs to require independent validation at each agent boundary; passing this check provides relevant evidence but does not certify every possible input or security risk.

Incorrect Answers

A:

Choosing to simulate a dropped response and confirm exactly one committed purchase would direct effort toward idempotent agent actions, whose characteristic issue is that a payment agent retries after a network disconnect and creates a duplicate purchase. The relevant corrective or preventive action here remains to submit a missing-authorization handoff and confirm it is rejected, with its own validation criteria.

B:

In the media platform acceptance review, trying to send an unstructured free-text plan and assume the executor knows permissions does not test the exact multi-agent handoff schema condition where a planner agent passes incomplete instructions to an executor agent; the relevant test is to submit a missing-authorization handoff and confirm it is rejected. Acceptance should instead demonstrate that the service can submit a missing-authorization handoff and confirm it is rejected, using the described failure as the test criterion.

D:

The action to require independent validation at each agent boundary can contribute to multi-agent handoff schema as a ongoing preventive rule step; here, a planner agent passes incomplete instructions to an executor agent. In the media platform acceptance review, trying to require independent validation at each agent boundary does not test the exact multi-agent handoff schema condition where a planner agent passes incomplete instructions to an executor agent; the relevant test is to submit a missing-authorization handoff and confirm it is rejected.

E:

The action to define typed handoff payloads with task limits and verified evidence can contribute to multi-agent handoff schema as a implementation step; here, a planner agent passes incomplete instructions to an executor agent. In the media platform acceptance review, trying to define typed handoff payloads with task limits and verified evidence does not test the exact multi-agent handoff schema condition where a planner agent passes incomplete instructions to an executor agent; the relevant test is to submit a missing-authorization handoff and confirm it is rejected.

Question 496

A cloud developer at a education technology company is conducting a support escalation concerning specialist agent selection. The observed problem is: A finance question is routed to a generic vision agent with unrelated tools. Which TWO actions together represent the needed implementation AND a durable preventive safeguard? (Select TWO answers.)

  1. Use the agent with the most available tool permissions for every task
  2. Restrict specialist tool inventories and evaluate routing precision
  3. Ask a tagged finance question and check selected specialist and actions
  4. Route specialized tasks to agents with relevant tools and bounded expertise
  5. Record correlated traces for model calls tool invocations and handoffs

Explanation

Correct Answers

B:

Standardizing the boundary to restrict specialist tool inventories and evaluate routing precision is preventive because the same specialist agent selection failure could otherwise return: a finance question is routed to a generic vision agent with unrelated tools. For this education technology company, the service should also route specialized tasks to agents with relevant tools and bounded expertise, and tests should ask a tagged finance question and check selected specialist and actions so that the ongoing requirement is measurable.

D:

Choosing to route specialized tasks to agents with relevant tools and bounded expertise directly targets the described specialist agent selection gap: a finance question is routed to a generic vision agent with unrelated tools. As a lasting boundary, they also need to restrict specialist tool inventories and evaluate routing precision, because an implementation by itself does not prove safe operation.

Incorrect Answers

A:

Choosing to use the agent with the most available tool permissions for every task cannot directly replace the specified action to restrict specialist tool inventories and evaluate routing precision when a finance question is routed to a generic vision agent with unrelated tools. The actual implementation must route specialized tasks to agents with relevant tools and bounded expertise, because the stated service condition is that a finance question is routed to a generic vision agent with unrelated tools.

C:

The action to ask a tagged finance question and check selected specialist and actions can contribute to specialist agent selection as a targeted acceptance check step; here, a finance question is routed to a generic vision agent with unrelated tools. Choosing to ask a tagged finance question and check selected specialist and actions cannot directly replace the specified action to restrict specialist tool inventories and evaluate routing precision when a finance question is routed to a generic vision agent with unrelated tools.

E:

The option to record correlated traces for model calls tool invocations and handoffs would normally serve a different agent tracing across tools requirement, where a complex agent fails after several model calls but operators lack the failure step. The alternative may be useful in its own context, but it does not directly supply route specialized tasks to agents with relevant tools and bounded expertise for the stated decision.

Question 497

After a service audit at a pharmaceutical research group, the team observes this failure: A finance question is routed to a generic vision agent with unrelated tools. Which flawed assumption or implementation choice would MOST DIRECTLY explain why the intended outcome was not achieved?

  1. Send an unstructured free-text plan and assume the executor knows permissions
  2. Route specialized tasks to agents with relevant tools and bounded expertise
  3. Ask a tagged finance question and check selected specialist and actions
  4. Use the agent with the most available tool permissions for every task

Explanation

Correct Answer

D:

The faulty assumption is use the agent with the most available tool permissions for every task. A functioning alternative would be to route specialized tasks to agents with relevant tools and bounded expertise, and the relevant validation would be to ask a tagged finance question and check selected specialist and actions.

Incorrect Answers

A:

The option to send an unstructured free-text plan and assume the executor knows permissions would normally serve a different multi-agent handoff schema requirement, where a planner agent passes incomplete instructions to an executor agent. The alternative may be useful in its own context, but it does not directly supply use the agent with the most available tool permissions for every task for the stated decision.

B:

The alternative to route specialized tasks to agents with relevant tools and bounded expertise is a implementation step rather than the faulty assumption being diagnosed. This pharmaceutical research group failure-analysis alternative of route specialized tasks to agents with relevant tools and bounded expertise does not diagnose the particular specialist agent selection defect where a finance question is routed to a generic vision agent with unrelated tools; the keyed mistake is use the agent with the most available tool permissions for every task.

C:

The alternative to ask a tagged finance question and check selected specialist and actions is a targeted acceptance check step rather than the faulty assumption being diagnosed. This pharmaceutical research group failure-analysis alternative of ask a tagged finance question and check selected specialist and actions does not diagnose the particular specialist agent selection defect where a finance question is routed to a generic vision agent with unrelated tools; the keyed mistake is use the agent with the most available tool permissions for every task.

Question 498

During a security tabletop exercise at a travel support provider, an AI team must address specialist agent selection. The requirement is: A finance question is routed to a generic vision agent with unrelated tools. Which THREE actions together implement the needed solution, its preventive guardrail, and a targeted acceptance test? (Select THREE answers.)

  1. Validate audience expiry and least privilege on tokens
  2. Ask a tagged finance question and check selected specialist and actions
  3. Reroute the request to the correct domain specialist and review outputs
  4. Use the agent with the most available tool permissions for every task
  5. Route specialized tasks to agents with relevant tools and bounded expertise
  6. Restrict specialist tool inventories and evaluate routing precision

Explanation

Correct Answers

B:

To verify this change, the travel support provider should ask a tagged finance question and check selected specialist and actions. Operators still need the implementation to route specialized tasks to agents with relevant tools and bounded expertise and the preventive requirement to restrict specialist tool inventories and evaluate routing precision.

E:

Choosing to route specialized tasks to agents with relevant tools and bounded expertise directly targets the described specialist agent selection gap: a finance question is routed to a generic vision agent with unrelated tools. As a lasting boundary, they also need to restrict specialist tool inventories and evaluate routing precision, because an implementation by itself does not prove safe operation.

F:

Standardizing the boundary to restrict specialist tool inventories and evaluate routing precision is preventive because the same specialist agent selection failure could otherwise return: a finance question is routed to a generic vision agent with unrelated tools. For this travel support provider, the service should also route specialized tasks to agents with relevant tools and bounded expertise, and tests should ask a tagged finance question and check selected specialist and actions so that the ongoing requirement is measurable.

Incorrect Answers

A:

For external api auth for agent, it can be sensible to validate audience expiry and least privilege on tokens if an agent must access a customer API using delegated user permissions. Without the actual change to route specialized tasks to agents with relevant tools and bounded expertise, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.

C:

The action to reroute the request to the correct domain specialist and review outputs can contribute to specialist agent selection as a corrective remediation step; here, a finance question is routed to a generic vision agent with unrelated tools. During travel support provider remediation, the proposed step to reroute the request to the correct domain specialist and review outputs is not the direct restoration action for specialist agent selection when a finance question is routed to a generic vision agent with unrelated tools; the keyed repair is ask a tagged finance question and check selected specialist and actions.

D:

During travel support provider remediation, the proposed step to use the agent with the most available tool permissions for every task is not the direct restoration action for specialist agent selection when a finance question is routed to a generic vision agent with unrelated tools; the keyed repair is ask a tagged finance question and check selected specialist and actions. After remediation the team must ask a tagged finance question and check selected specialist and actions; the action in this option does not itself establish that target recovery outcome.

Question 499

A agent designer at a scientific publisher wants the following issue to be prevented in later deployments: A finance question is routed to a generic vision agent with unrelated tools. Which enduring control or engineering guardrail BEST prevents recurrence?

  1. Reroute the request to the correct domain specialist and review outputs
  2. Validate tool input before any side effects
  3. Use the agent with the most available tool permissions for every task
  4. Restrict specialist tool inventories and evaluate routing precision

Explanation

Correct Answer

D:

The sustained prevention measure is to restrict specialist tool inventories and evaluate routing precision. The implementation can still use the separate action to route specialized tasks to agents with relevant tools and bounded expertise, but a guardrail constrains future changes. The scientific publisher can measure whether the rule works by arranging to ask a tagged finance question and check selected specialist and actions.

Incorrect Answers

A:

The action to reroute the request to the correct domain specialist and review outputs can contribute to specialist agent selection as a corrective remediation step; here, a finance question is routed to a generic vision agent with unrelated tools. For prevention at the scientific publisher, the measure to reroute the request to the correct domain specialist and review outputs is weaker or aimed at another issue than the specialist agent selection guardrail to restrict specialist tool inventories and evaluate routing precision needed when a finance question is routed to a generic vision agent with unrelated tools.

B:

The action to validate tool input before any side effects addresses typed tool function rather than the decisive specialist agent selection condition. The scientific publisher should not infer that fixing one adjacent AI service validates its unrelated workflow; the requested action is to restrict specialist tool inventories and evaluate routing precision.

C:

For prevention at the scientific publisher, the measure to use the agent with the most available tool permissions for every task is weaker or aimed at another issue than the specialist agent selection guardrail to restrict specialist tool inventories and evaluate routing precision needed when a finance question is routed to a generic vision agent with unrelated tools. The lasting guardrail for this specific threat is to restrict specialist tool inventories and evaluate routing precision, not simply to complete an adjacent service task.

Question 500

During a service audit, a responsible-AI reviewer must accept a change for specialist agent selection. The issue to be addressed is: A finance question is routed to a generic vision agent with unrelated tools. Which concrete verification action demonstrates that the relevant behavior works?

  1. Route specialized tasks to agents with relevant tools and bounded expertise
  2. Use the agent with the most available tool permissions for every task
  3. Ask a tagged finance question and check selected specialist and actions
  4. Sample high-impact claims and inspect corresponding source spans

Explanation

Correct Answer

C:

To verify this change, the nonprofit organization should ask a tagged finance question and check selected specialist and actions. Operators still need the implementation to route specialized tasks to agents with relevant tools and bounded expertise and the preventive requirement to restrict specialist tool inventories and evaluate routing precision.

Incorrect Answers

A:

The action to route specialized tasks to agents with relevant tools and bounded expertise can contribute to specialist agent selection as a implementation step; here, a finance question is routed to a generic vision agent with unrelated tools. In the nonprofit organization acceptance review, trying to route specialized tasks to agents with relevant tools and bounded expertise does not test the exact specialist agent selection condition where a finance question is routed to a generic vision agent with unrelated tools; the relevant test is to ask a tagged finance question and check selected specialist and actions.

B:

In the nonprofit organization acceptance review, trying to use the agent with the most available tool permissions for every task does not test the exact specialist agent selection condition where a finance question is routed to a generic vision agent with unrelated tools; the relevant test is to ask a tagged finance question and check selected specialist and actions. Acceptance should instead demonstrate that the service can ask a tagged finance question and check selected specialist and actions, using the described failure as the test criterion.

D:

For rag-based agent citation, it can be sensible to sample high-impact claims and inspect corresponding source spans if an agent answers from retrieved references without linking the supporting passage. Without the actual change to ask a tagged finance question and check selected specialist and actions, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.

Developing AI Apps and Agents on Azure (AI-103) Practice Tests