Microsoft AI-103 (Developing AI Apps and Agents on Azure) Practice Test 10

Question 90

Engineers have implemented a change to adjust quota allocation for sustained traffic. What validation would expose a remaining defect?

  1. Test service configuration calls and document searches with each role independently
  2. Track 429 incidence, retry amplification, and successful completion time under load
  3. Compare sustained token throughput and throttling against the new allocation
  4. Measure input tokens per answered query alongside grounded-answer quality

Explanation

Correct Answer

C:

The accepted change is to adjust quota allocation for sustained traffic; this evaluation checks whether it can serve a stable higher request rate when the deployment repeatedly reaches its assigned token budget, since tokens-per-minute and request limits constrain throughput; allocation or capacity decisions must match the deployment's actual measured demand.

Incorrect Answers

A:

This test concerns the ability to let operators change infrastructure while limiting which identities can read sensitive index documents, for which management-plane roles and data-plane roles authorize different operations; control over deployment configuration does not imply every content read permission; it does not establish whether the change to adjust quota allocation for sustained traffic works for the current release.

B:

This test concerns the ability to recover from temporary HTTP 429 responses without amplifying load on the model deployment, for which exponential backoff with jitter and respect for retry hints reduces synchronized retries; extra retries cannot replace insufficient sustained capacity; it does not establish whether the change to adjust quota allocation for sustained traffic works for the current release.

D:

This test concerns the ability to reduce inference spend caused by oversized repeated document context, for which input tokens contribute to context limits and usage costs; pruning irrelevant evidence can improve efficiency without changing the model deployment; it does not establish whether the change to adjust quota allocation for sustained traffic works for the current release.

Question 91

A pilot has a functional gap: A model gives polished answers but includes unsupported policy exceptions not present in the retrieved pages. How should the team address that gap?

  1. Run source-grounded evaluations and inspect failing claim-to-passage alignments
  2. Collect and classify guardrail outcomes with privacy-aware security telemetry
  3. Add correlated request tracing across retrieval, tools, and model calls
  4. Instrument each tool call with duration, status, and bounded retry decisions

Explanation

Correct Answer

A:

The fault signal is that a model gives polished answers but includes unsupported policy exceptions not present in the retrieved pages; the targeted fix works because groundedness evaluations compare generated claims with supplied evidence; fluency or user satisfaction alone cannot establish factual support, restoring the ability to detect fluent answers that introduce claims unsupported by retrieved passages.

Incorrect Answers

B:

If the failing function were to identify spikes in unsafe or blocked prompts and unsafe generated material, this change could be relevant because guardrail and content-filter telemetry reveals policy-related events; it should be segmented without exposing sensitive payloads unnecessarily; the actual symptom is that a model gives polished answers but includes unsupported policy exceptions not present in the retrieved pages, which calls for a different repair.

C:

If the failing function were to locate whether a slow RAG response is delayed by search, tool calls, or model generation, this change could be relevant because distributed traces expose timing and relationships among retrieval, tool and inference spans; aggregate CPU graphs alone cannot isolate these steps; the actual symptom is that a model gives polished answers but includes unsupported policy exceptions not present in the retrieved pages, which calls for a different repair.

D:

If the failing function were to find why agents fail to complete workflows despite healthy model endpoints, this change could be relevant because tool telemetry records HTTP failures, permission denials and timeouts at action boundaries; model token metrics will not diagnose a downstream API error; the actual symptom is that a model gives polished answers but includes unsupported policy exceptions not present in the retrieved pages, which calls for a different repair.

Question 92

After implementing a workflow to restrict agent outbound tool access, the team needs a robust release gate. Which check should it choose?

  1. Attempt blocked destinations in a controlled red-team test and inspect egress logs
  2. Validate vector dimensions and semantic recall using a fixed representative query set
  3. Validate authorized queries succeed and index modification is still denied
  4. Confirm no requests use the retired credential and access continues during rotation

Explanation

Correct Answer

A:

The accepted change is to restrict agent outbound tool access; this evaluation checks whether it can prevent a compromised agent from calling unapproved external endpoints, since tool allowlists, network egress policies, and scoped identities limit what an agent can reach; prompt instructions alone are not a network control.

Incorrect Answers

B:

This test concerns the ability to catch incompatible document and query vectors when retrieval quality suddenly collapses, for which embedding dimensionality and model compatibility must be maintained between indexed passages and query vectors; vector space changes require a coordinated migration; it does not establish whether the change to restrict agent outbound tool access works for the current release.

C:

This test concerns the ability to allow a retrieval application to query only the approved search resource without subscription administration rights, for which Azure RBAC can grant scoped data-plane permissions to the workload principal; broad management roles are unnecessary for query-only service calls; it does not establish whether the change to restrict agent outbound tool access works for the current release.

D:

This test concerns the ability to replace an exposed secret while keeping a production inference endpoint available, for which credential rollover with overlapping validity and centralized secret distribution enables staged updates; longer-term managed identity can reduce secret use; it does not establish whether the change to restrict agent outbound tool access works for the current release.

Question 93

Engineers have implemented a change to implement backoff based on rate-limit signals. What validation would expose a remaining defect?

  1. Track 429 incidence, retry amplification, and successful completion time under load
  2. Measure cache-hit ratio, stale-answer rate, and token savings after invalidation tests
  3. Measure input tokens per answered query alongside grounded-answer quality
  4. Graph cost per completed task and token usage across dialogue length bands

Explanation

Correct Answer

A:

The accepted change is to implement backoff based on rate-limit signals; this evaluation checks whether it can recover from temporary HTTP 429 responses without amplifying load on the model deployment, since exponential backoff with jitter and respect for retry hints reduces synchronized retries; extra retries cannot replace insufficient sustained capacity.

Incorrect Answers

B:

This test concerns the ability to avoid repeating the same approved reference answer for unchanged public content, for which an application cache can reduce repeated inference requests when answers and policy versions are safe to reuse; invalidation and access scope remain essential; it does not establish whether the change to implement backoff based on rate-limit signals works for the current release.

C:

This test concerns the ability to reduce inference spend caused by oversized repeated document context, for which input tokens contribute to context limits and usage costs; pruning irrelevant evidence can improve efficiency without changing the model deployment; it does not establish whether the change to implement backoff based on rate-limit signals works for the current release.

D:

This test concerns the ability to identify waste when conversation history and tool outputs grow over successive exchanges, for which per-turn token accounting reveals input-context inflation; controlled summarization or trimming can reduce cost while preserving vital state; it does not establish whether the change to implement backoff based on rate-limit signals works for the current release.

Question 94

An Azure AI delivery has 3 independently testable requirements: (1) avoid repeating the same approved reference answer for unchanged public content; (2) recover from temporary HTTP 429 responses without amplifying load on the model deployment; and (3) reduce inference spend caused by oversized repeated document context. Select the most directly specialized implementation for each requirement. Choose THREE.

  1. Trace each retrieval and generation span
  2. Implement backoff based on rate-limit signals
  3. Use caching for reusable deterministic outputs
  4. Adjust quota allocation for sustained traffic
  5. Inspect embedding pipeline health after model changes
  6. Limit prompt and retrieved context length

Explanation

Correct Answers

B:

For the stated requirement to recover from temporary HTTP 429 responses without amplifying load on the model deployment, exponential backoff with jitter and respect for retry hints reduces synchronized retries; extra retries cannot replace insufficient sustained capacity; this is the mechanism that the system must implement. This choice satisfies the specific deliverable to recover from temporary HTTP 429 responses without amplifying load on the model deployment, while the other deliverables require separate capabilities.

C:

For the stated requirement to avoid repeating the same approved reference answer for unchanged public content, an application cache can reduce repeated inference requests when answers and policy versions are safe to reuse; invalidation and access scope remain essential; this is the mechanism that the system must implement. This choice satisfies the specific deliverable to avoid repeating the same approved reference answer for unchanged public content, while the other deliverables require separate capabilities.

F:

For the stated requirement to reduce inference spend caused by oversized repeated document context, input tokens contribute to context limits and usage costs; pruning irrelevant evidence can improve efficiency without changing the model deployment; this is the mechanism that the system must implement. This choice satisfies the specific deliverable to reduce inference spend caused by oversized repeated document context, while the other deliverables require separate capabilities.

Incorrect Answers

A:

For a system whose objective is to locate whether a slow RAG response is delayed by search, tool calls, or model generation, distributed traces expose timing and relationships among retrieval, tool and inference spans; aggregate CPU graphs alone cannot isolate these steps; this does not directly meet the required outcome to reduce inference spend caused by oversized repeated document context.

D:

For a system whose objective is to serve a stable higher request rate when the deployment repeatedly reaches its assigned token budget, tokens-per-minute and request limits constrain throughput; allocation or capacity decisions must match the deployment's actual measured demand; this does not directly meet the required outcome to reduce inference spend caused by oversized repeated document context.

E:

For a system whose objective is to catch incompatible document and query vectors when retrieval quality suddenly collapses, embedding dimensionality and model compatibility must be maintained between indexed passages and query vectors; vector space changes require a coordinated migration; this does not directly meet the required outcome to reduce inference spend caused by oversized repeated document context.

Question 95

A cloud engineering team needs its AI workflow to prevent a compromised agent from calling unapproved external endpoints. What technical approach most directly enables that?

  1. Assign least-privilege data access to a workload identity
  2. Measure groundedness with cited source evidence
  3. Restrict agent outbound tool access
  4. Separate data-plane use from resource management

Explanation

Correct Answer

C:

For the stated requirement to prevent a compromised agent from calling unapproved external endpoints, tool allowlists, network egress policies, and scoped identities limit what an agent can reach; prompt instructions alone are not a network control; this is the mechanism that the system must implement.

Incorrect Answers

A:

For a system whose objective is to allow a retrieval application to query only the approved search resource without subscription administration rights, Azure RBAC can grant scoped data-plane permissions to the workload principal; broad management roles are unnecessary for query-only service calls; this does not directly meet the required outcome to prevent a compromised agent from calling unapproved external endpoints.

B:

For a system whose objective is to detect fluent answers that introduce claims unsupported by retrieved passages, groundedness evaluations compare generated claims with supplied evidence; fluency or user satisfaction alone cannot establish factual support; this does not directly meet the required outcome to prevent a compromised agent from calling unapproved external endpoints.

D:

For a system whose objective is to let operators change infrastructure while limiting which identities can read sensitive index documents, management-plane roles and data-plane roles authorize different operations; control over deployment configuration does not imply every content read permission; this does not directly meet the required outcome to prevent a compromised agent from calling unapproved external endpoints.

Question 96

An Azure AI delivery has 2 independently testable requirements: (1) identify waste when conversation history and tool outputs grow over successive exchanges; and (2) catch incompatible document and query vectors when retrieval quality suddenly collapses. Select the most directly specialized implementation for each requirement. Choose TWO.

  1. Inspect embedding pipeline health after model changes
  2. Monitor token utilization across agent turns
  3. Track tool invocation errors and timeouts
  4. Use caching for reusable deterministic outputs
  5. Assign least-privilege data access to a workload identity
  6. Compare behavior across model or prompt versions

Explanation

Correct Answers

A:

For the stated requirement to catch incompatible document and query vectors when retrieval quality suddenly collapses, embedding dimensionality and model compatibility must be maintained between indexed passages and query vectors; vector space changes require a coordinated migration; this is the mechanism that the system must implement. This choice satisfies the specific deliverable to catch incompatible document and query vectors when retrieval quality suddenly collapses, while the other deliverables require separate capabilities.

B:

For the stated requirement to identify waste when conversation history and tool outputs grow over successive exchanges, per-turn token accounting reveals input-context inflation; controlled summarization or trimming can reduce cost while preserving vital state; this is the mechanism that the system must implement. This choice satisfies the specific deliverable to identify waste when conversation history and tool outputs grow over successive exchanges, while the other deliverables require separate capabilities.

Incorrect Answers

C:

For a system whose objective is to find why agents fail to complete workflows despite healthy model endpoints, tool telemetry records HTTP failures, permission denials and timeouts at action boundaries; model token metrics will not diagnose a downstream API error; this does not directly meet the required outcome to catch incompatible document and query vectors when retrieval quality suddenly collapses.

D:

For a system whose objective is to avoid repeating the same approved reference answer for unchanged public content, an application cache can reduce repeated inference requests when answers and policy versions are safe to reuse; invalidation and access scope remain essential; this does not directly meet the required outcome to catch incompatible document and query vectors when retrieval quality suddenly collapses.

E:

For a system whose objective is to allow a retrieval application to query only the approved search resource without subscription administration rights, Azure RBAC can grant scoped data-plane permissions to the workload principal; broad management roles are unnecessary for query-only service calls; this does not directly meet the required outcome to catch incompatible document and query vectors when retrieval quality suddenly collapses.

F:

For a system whose objective is to detect regression when a newly deployed model changes how it answers the same business questions, version-tagged test results separate model changes from request-mix changes and allow a reliable rollback decision; this does not directly meet the required outcome to catch incompatible document and query vectors when retrieval quality suddenly collapses.

Question 97

The planned deployment includes a mechanism to inspect embedding pipeline health after model changes. Which outcome should the QA team measure?

  1. Validate vector dimensions and semantic recall using a fixed representative query set
  2. Measure input tokens per answered query alongside grounded-answer quality
  3. Graph cost per completed task and token usage across dialogue length bands
  4. Compare task accuracy, groundedness, and safety scores by deployment revision

Explanation

Correct Answer

A:

The accepted change is to inspect embedding pipeline health after model changes; this evaluation checks whether it can catch incompatible document and query vectors when retrieval quality suddenly collapses, since embedding dimensionality and model compatibility must be maintained between indexed passages and query vectors; vector space changes require a coordinated migration.

Incorrect Answers

B:

This test concerns the ability to reduce inference spend caused by oversized repeated document context, for which input tokens contribute to context limits and usage costs; pruning irrelevant evidence can improve efficiency without changing the model deployment; it does not establish whether the change to inspect embedding pipeline health after model changes works for the current release.

C:

This test concerns the ability to identify waste when conversation history and tool outputs grow over successive exchanges, for which per-turn token accounting reveals input-context inflation; controlled summarization or trimming can reduce cost while preserving vital state; it does not establish whether the change to inspect embedding pipeline health after model changes works for the current release.

D:

This test concerns the ability to detect regression when a newly deployed model changes how it answers the same business questions, for which version-tagged test results separate model changes from request-mix changes and allow a reliable rollback decision; it does not establish whether the change to inspect embedding pipeline health after model changes works for the current release.

Question 98

An engineer receives this incident report: The agent repeatedly cites an old handbook because the scheduled source refresh silently fails. What should be changed?

  1. Add correlated request tracing across retrieval, tools, and model calls
  2. Collect and classify guardrail outcomes with privacy-aware security telemetry
  3. Alert on ingestion failures and index freshness drift, then repair the source pipeline
  4. Tag outputs by model and prompt version and run side-by-side regression cases

Explanation

Correct Answer

C:

The fault signal is that the agent repeatedly cites an old handbook because the scheduled source refresh silently fails; the targeted fix works because indexer execution status, document counts, and source timestamps show whether retrieval data is current; a good model response cannot compensate for missing content, restoring the ability to detect when newly approved source documents are not available to the grounding pipeline.

Incorrect Answers

A:

If the failing function were to locate whether a slow RAG response is delayed by search, tool calls, or model generation, this change could be relevant because distributed traces expose timing and relationships among retrieval, tool and inference spans; aggregate CPU graphs alone cannot isolate these steps; the actual symptom is that the agent repeatedly cites an old handbook because the scheduled source refresh silently fails, which calls for a different repair.

B:

If the failing function were to identify spikes in unsafe or blocked prompts and unsafe generated material, this change could be relevant because guardrail and content-filter telemetry reveals policy-related events; it should be segmented without exposing sensitive payloads unnecessarily; the actual symptom is that the agent repeatedly cites an old handbook because the scheduled source refresh silently fails, which calls for a different repair.

D:

If the failing function were to detect regression when a newly deployed model changes how it answers the same business questions, this change could be relevant because version-tagged test results separate model changes from request-mix changes and allow a reliable rollback decision; the actual symptom is that the agent repeatedly cites an old handbook because the scheduled source refresh silently fails, which calls for a different repair.

Question 99

After deployment, the team observes: The application receives authorization failures and an operator proposes giving it subscription Owner. Which remediation addresses the likely cause?

  1. Introduce a new credential through a secret store, roll replicas, then revoke the exposed credential
  2. Grant the specific Search index data-reading permission at the appropriate resource scope
  3. Cache validated FAQ outputs by normalized question and content version with proper expiration
  4. Align query embeddings with index embeddings and re-embed stored content when migrating models
  5. Assign the required data-plane reader role separately from the management role

Explanation

Correct Answer

B:

The fault signal is that the application receives authorization failures and an operator proposes giving it subscription Owner; the targeted fix works because Azure RBAC can grant scoped data-plane permissions to the workload principal; broad management roles are unnecessary for query-only service calls, restoring the ability to allow a retrieval application to query only the approved search resource without subscription administration rights.

Incorrect Answers

A:

If the failing function were to replace an exposed secret while keeping a production inference endpoint available, this change could be relevant because credential rollover with overlapping validity and centralized secret distribution enables staged updates; longer-term managed identity can reduce secret use; the actual symptom is that the application receives authorization failures and an operator proposes giving it subscription Owner, which calls for a different repair.

C:

If the failing function were to avoid repeating the same approved reference answer for unchanged public content, this change could be relevant because an application cache can reduce repeated inference requests when answers and policy versions are safe to reuse; invalidation and access scope remain essential; the actual symptom is that the application receives authorization failures and an operator proposes giving it subscription Owner, which calls for a different repair.

D:

If the failing function were to catch incompatible document and query vectors when retrieval quality suddenly collapses, this change could be relevant because embedding dimensionality and model compatibility must be maintained between indexed passages and query vectors; vector space changes require a coordinated migration; the actual symptom is that the application receives authorization failures and an operator proposes giving it subscription Owner, which calls for a different repair.

E:

If the failing function were to let operators change infrastructure while limiting which identities can read sensitive index documents, this change could be relevant because management-plane roles and data-plane roles authorize different operations; control over deployment configuration does not imply every content read permission; the actual symptom is that the application receives authorization failures and an operator proposes giving it subscription Owner, which calls for a different repair.

Question 100

A user-acceptance test fails with the following observation: Customer complaints rise after a model refresh but traces do not identify which deployment revision produced each answer. What is the best correction?

  1. Align query embeddings with index embeddings and re-embed stored content when migrating models
  2. Tag outputs by model and prompt version and run side-by-side regression cases
  3. Cache validated FAQ outputs by normalized question and content version with proper expiration
  4. Instrument each tool call with duration, status, and bounded retry decisions

Explanation

Correct Answer

B:

The fault signal is that customer complaints rise after a model refresh but traces do not identify which deployment revision produced each answer; the targeted fix works because version-tagged test results separate model changes from request-mix changes and allow a reliable rollback decision, restoring the ability to detect regression when a newly deployed model changes how it answers the same business questions.

Incorrect Answers

A:

If the failing function were to catch incompatible document and query vectors when retrieval quality suddenly collapses, this change could be relevant because embedding dimensionality and model compatibility must be maintained between indexed passages and query vectors; vector space changes require a coordinated migration; the actual symptom is that customer complaints rise after a model refresh but traces do not identify which deployment revision produced each answer, which calls for a different repair.

C:

If the failing function were to avoid repeating the same approved reference answer for unchanged public content, this change could be relevant because an application cache can reduce repeated inference requests when answers and policy versions are safe to reuse; invalidation and access scope remain essential; the actual symptom is that customer complaints rise after a model refresh but traces do not identify which deployment revision produced each answer, which calls for a different repair.

D:

If the failing function were to find why agents fail to complete workflows despite healthy model endpoints, this change could be relevant because tool telemetry records HTTP failures, permission denials and timeouts at action boundaries; model token metrics will not diagnose a downstream API error; the actual symptom is that customer complaints rise after a model refresh but traces do not identify which deployment revision produced each answer, which calls for a different repair.

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