Microsoft AI-102 (Designing and Implementing a Microsoft Azure AI Solution) Practice Test 5

Question 40

An analyzer returns fluent summaries but the workflow needs a typed claim amount for accounting. What should change?

  1. Route the multipage document to Document Intelligence rather than still-image captioning
  2. Add an explicit numeric claim-amount field to the analyzer extraction schema
  3. Preserve document IDs, source versions, extraction model identifiers, and relevant response metadata
  4. Poll or wait for the asynchronous video indexing job to complete before requesting final insights

Explanation

Correct Answer

B:

Structured field definitions constrain the expected output contract more directly than asking for an unconstrained narrative. In the claims intake with mixed media, confirm the fix by retesting the affected behavior after adding an explicit numeric claim-amount field to the analyzer extraction schema.

Incorrect Answers

A:

Although useful elsewhere in the claims intake with mixed media, document Intelligence has document and table extraction workflows; a single-image caption request is not a multipage parser; the present case is different and calls for the team to add an explicit numeric claim-amount field to the analyzer extraction schema.

C:

That change addresses another failure mode: model and source provenance allow analysts to compare historic results to the original document; choosing a different prebuilt model is not incident traceability; it does not address this case, for which the relevant intervention is to add an explicit numeric claim-amount field to the analyzer extraction schema.

D:

Video indexing is asynchronous for many workflows; a pending job is not an empty transcript and should not be treated as finished in its own workflow, but that is a different correction from the present requirement to add an explicit numeric claim-amount field to the analyzer extraction schema. The reported issue would remain.

Question 41

An application cannot read a connected storage account despite the connection existing. What additional requirement should be checked?

  1. Grant the configured identity or credential the required permissions on the storage resource
  2. Review the model's licensing terms and allowed deployment offer
  3. Review project-scoped access assignments for that identity
  4. Verify regional support for the exact model version and deployment category

Explanation

Correct Answer

A:

A connection records how to reach a service but does not replace its authorization checks. In the enterprise generative-AI development environment, confirm the fix by retesting the affected behavior after granting the configured identity or credential the required permissions on the storage resource.

Incorrect Answers

B:

Model catalogs include different publishers and offers; capability alone does not establish permitted commercial use in its own workflow, but that is a different correction from the present requirement to grant the configured identity or credential the required permissions on the storage resource. The reported issue would remain.

C:

Although useful elsewhere in the enterprise generative-AI development environment, environment access does not necessarily authorize every project action; permissions must match the target project scope; the present case is different and calls for the team to grant the configured identity or credential the required permissions on the storage resource.

D:

That change addresses another failure mode: availability can depend on model, version, and offer type; previous success in another region is not a universal guarantee; it does not address this case, for which the relevant intervention is to grant the configured identity or credential the required permissions on the storage resource.

Question 42

The project is handed to a second team, which has read access to shared model assets but no record of who approved deployment. What governance asset is missing?

  1. Confirm that the API operation and payload are supported by the newly deployed model
  2. Maintain project-linked release provenance and approval records for promoted deployments
  3. Resolve the correct project connection for each application's retrieval workflow
  4. Provision the underlying resource and deployment in an approved supported region

Explanation

Correct Answer

B:

A project organizes technical assets, while a separate approval trail records who authorized production changes and which version was promoted. In the enterprise generative-AI development environment, confirm the fix by retesting the affected behavior after maintaining project-linked release provenance and approval records for promoted deployments.

Incorrect Answers

A:

Although useful elsewhere in the enterprise generative-AI development environment, serving a different model does not imply every earlier API capability or parameter remains compatible; the present case is different and calls for the team to maintain project-linked release provenance and approval records for promoted deployments.

C:

Using the intended connection isolates integration targets; a generic project name is not proof of the selected external index in its own workflow, but that is a different correction from the present requirement to maintain project-linked release provenance and approval records for promoted deployments. The reported issue would remain.

D:

That change addresses another failure mode: altering a client URL or routing label does not relocate service-side model execution; it does not address this case, for which the relevant intervention is to maintain project-linked release provenance and approval records for promoted deployments.

Question 43

A release for an enterprise generative-AI development environment is blocked until the application can satisfy residency and model availability constraints before provisioning. The tester will verify that the provisioned resource resides in the approved region and supports the required model. What should be implemented?

  1. Create a model deployment
  2. Configure a project connection
  3. Create a Microsoft Foundry project
  4. Choose a supported deployment region

Explanation

Correct Answer

D:

Regional support and deployment availability must be checked for the intended resource and model offering. For this enterprise generative-AI development environment, the intended evidence is that the provisioned resource resides in the approved region and supports the required model. A deployment name or prompt template cannot relocate underlying service processing.

Incorrect Answers

A:

A model deployment makes a compatible model available for inference under its deployment name and supported endpoint; it may be used to expose a selected model through a usable inference deployment in a enterprise generative-AI development environment, but the current requirement is instead to satisfy residency and model availability constraints before provisioning. That separate capability does not prove that the provisioned resource resides in the approved region and supports the required model; A deployment alone is not a dataset or a governing policy document, so selecting that alternative would not meet the stated acceptance check.

B:

Project connections hold resource references and associated authorization configuration for integrated services; it may be used to authorize a project application to use an attached resource such as search or storage in a enterprise generative-AI development environment, but the current requirement is instead to satisfy residency and model availability constraints before provisioning. That separate capability does not prove that the provisioned resource resides in the approved region and supports the required model; Connecting a resource does not automatically create an embedding index inside it, so selecting that alternative would not meet the stated acceptance check.

C:

A Foundry project organizes application-oriented resources and connections for a team within the applicable Foundry environment; it may be used to isolate application assets, evaluations, and development work for one AI solution in a enterprise generative-AI development environment, but the current requirement is instead to satisfy residency and model availability constraints before provisioning. That separate capability does not prove that the provisioned resource resides in the approved region and supports the required model; A project is not the model endpoint serving inference requests, so selecting that alternative would not meet the stated acceptance check.

Question 44

A new model release is tested in staging while production must remain stable. What deployment strategy supports this?

  1. Grant the configured identity or credential the required permissions on the storage resource
  2. Review project-scoped access assignments for that identity
  3. Create a separately named deployment for the candidate model version
  4. Verify regional support for the exact model version and deployment category

Explanation

Correct Answer

C:

Distinct deployments let clients target staged and production versions intentionally without silently swapping the endpoint target. In the enterprise generative-AI development environment, confirm the fix by retesting the affected behavior after creating a separately named deployment for the candidate model version.

Incorrect Answers

A:

That change addresses another failure mode: a connection records how to reach a service but does not replace its authorization checks; it does not address this case, for which the relevant intervention is to create a separately named deployment for the candidate model version.

B:

Environment access does not necessarily authorize every project action; permissions must match the target project scope in its own workflow, but that is a different correction from the present requirement to create a separately named deployment for the candidate model version. The reported issue would remain.

D:

Although useful elsewhere in the enterprise generative-AI development environment, availability can depend on model, version, and offer type; previous success in another region is not a universal guarantee; the present case is different and calls for the team to create a separately named deployment for the candidate model version.

Question 45

A department stores regulated inputs in an approved geography and wants to provision a model elsewhere for lower latency. What requirement takes precedence?

  1. Update the project's Search connection to reference the intended resource
  2. Use distinct Foundry projects aligned with the two application workspaces
  3. Select a supported deployment region that satisfies the approved data-residency policy
  4. Verify that the deployment named `helpdesk-prod` exists and matches the request

Explanation

Correct Answer

C:

Performance preference cannot override a stated processing-residency requirement; support and governance must both be met. In the enterprise generative-AI development environment, confirm the fix by retesting the affected behavior after selecting a supported deployment region that satisfies the approved data-residency policy.

Incorrect Answers

A:

Project connections associate the application with external services; a valid connection must point to the correct resource in its own workflow, but that is a different correction from the present requirement to select a supported deployment region that satisfies the approved data-residency policy. The reported issue would remain.

B:

Although useful elsewhere in the enterprise generative-AI development environment, projects provide application-level organization; sharing every asset in one project unnecessarily couples collaboration and resource discovery; the present case is different and calls for the team to select a supported deployment region that satisfies the approved data-residency policy.

D:

That change addresses another failure mode: inference requests address a deployed model using a valid deployment identifier; a resource alone does not create every named deployment; it does not address this case, for which the relevant intervention is to select a supported deployment region that satisfies the approved data-residency policy.

Question 46

The customer requires image inputs and text outputs; the first selected catalog model supports text only. What should be evaluated before any deployment?

  1. Update the project's Search connection to reference the intended resource
  2. Select a supported deployment region that satisfies the approved data-residency policy
  3. Use distinct Foundry projects aligned with the two application workspaces
  4. Compare the candidate model's documented input and output modalities

Explanation

Correct Answer

D:

A text-only model cannot gain vision input merely by changing its deployment name; model capabilities define feasible workloads. In the enterprise generative-AI development environment, confirm the fix by retesting the affected behavior after comparing the candidate model's documented input and output modalities.

Incorrect Answers

A:

That change addresses another failure mode: project connections associate the application with external services; a valid connection must point to the correct resource; it does not address this case, for which the relevant intervention is to compare the candidate model's documented input and output modalities.

B:

Performance preference cannot override a stated processing-residency requirement; support and governance must both be met in its own workflow, but that is a different correction from the present requirement to compare the candidate model's documented input and output modalities. The reported issue would remain.

C:

Although useful elsewhere in the enterprise generative-AI development environment, projects provide application-level organization; sharing every asset in one project unnecessarily couples collaboration and resource discovery; the present case is different and calls for the team to compare the candidate model's documented input and output modalities.

Question 47

An engineer changes an application endpoint hostname and assumes that model inference moves to a new geography. What correction is needed?

  1. Maintain project-linked release provenance and approval records for promoted deployments
  2. Resolve the correct project connection for each application's retrieval workflow
  3. Provision the underlying resource and deployment in an approved supported region
  4. Confirm that the API operation and payload are supported by the newly deployed model

Explanation

Correct Answer

C:

Altering a client URL or routing label does not relocate service-side model execution. In the enterprise generative-AI development environment, confirm the fix by retesting the affected behavior after provisioning the underlying resource and deployment in an approved supported region.

Incorrect Answers

A:

That change addresses another failure mode: a project organizes technical assets, while a separate approval trail records who authorized production changes and which version was promoted; it does not address this case, for which the relevant intervention is to provision the underlying resource and deployment in an approved supported region.

B:

Using the intended connection isolates integration targets; a generic project name is not proof of the selected external index in its own workflow, but that is a different correction from the present requirement to provision the underlying resource and deployment in an approved supported region. The reported issue would remain.

D:

Although useful elsewhere in the enterprise generative-AI development environment, serving a different model does not imply every earlier API capability or parameter remains compatible; the present case is different and calls for the team to provision the underlying resource and deployment in an approved supported region.

Question 48

Two independent business applications share one Foundry environment but need different datasets and evaluation assets. How should the team organize them?

  1. Select a supported deployment region that satisfies the approved data-residency policy
  2. Use distinct Foundry projects aligned with the two application workspaces
  3. Compare the candidate model's documented input and output modalities
  4. Verify that the deployment named `helpdesk-prod` exists and matches the request

Explanation

Correct Answer

B:

Projects provide application-level organization; sharing every asset in one project unnecessarily couples collaboration and resource discovery. In the enterprise generative-AI development environment, confirm the fix by retesting the affected behavior after using distinct Foundry projects aligned with the two application workspaces.

Incorrect Answers

A:

Performance preference cannot override a stated processing-residency requirement; support and governance must both be met in its own workflow, but that is a different correction from the present requirement to use distinct Foundry projects aligned with the two application workspaces. The reported issue would remain.

C:

Although useful elsewhere in the enterprise generative-AI development environment, a text-only model cannot gain vision input merely by changing its deployment name; model capabilities define feasible workloads; the present case is different and calls for the team to use distinct Foundry projects aligned with the two application workspaces.

D:

That change addresses another failure mode: inference requests address a deployed model using a valid deployment identifier; a resource alone does not create every named deployment; it does not address this case, for which the relevant intervention is to use distinct Foundry projects aligned with the two application workspaces.

Question 49

A legal team requires restrictions on model licensing before proof-of-concept use. What catalog metadata must be reviewed?

  1. Verify regional support for the exact model version and deployment category
  2. Review project-scoped access assignments for that identity
  3. Review the model's licensing terms and allowed deployment offer
  4. Create a separately named deployment for the candidate model version

Explanation

Correct Answer

C:

Model catalogs include different publishers and offers; capability alone does not establish permitted commercial use. In the enterprise generative-AI development environment, confirm the fix by retesting the affected behavior after reviewing the model's licensing terms and allowed deployment offer.

Incorrect Answers

A:

Although useful elsewhere in the enterprise generative-AI development environment, availability can depend on model, version, and offer type; previous success in another region is not a universal guarantee; the present case is different and calls for the team to review the model's licensing terms and allowed deployment offer.

B:

That change addresses another failure mode: environment access does not necessarily authorize every project action; permissions must match the target project scope; it does not address this case, for which the relevant intervention is to review the model's licensing terms and allowed deployment offer.

D:

Distinct deployments let clients target staged and production versions intentionally without silently swapping the endpoint target in its own workflow, but that is a different correction from the present requirement to review the model's licensing terms and allowed deployment offer. The reported issue would remain.

Question 50

An enterprise generative-AI development environment must assess available models against language, modality, deployment, and licensing requirements. Reviewers will accept the implementation when the team has selected a model family whose capabilities fit its evaluation criteria. Which approach best fulfills that requirement?

  1. Configure a project connection
  2. Choose a model from the Foundry model catalog
  3. Create a Microsoft Foundry project
  4. Choose a supported deployment region

Explanation

Correct Answer

B:

The model catalog provides discoverable model choices and model information before deployment. For this enterprise generative-AI development environment, the intended evidence is that the team has selected a model family whose capabilities fit its evaluation criteria. Browsing a catalog does not itself reserve serving capacity or return model responses.

Incorrect Answers

A:

Project connections hold resource references and associated authorization configuration for integrated services; it may be used to authorize a project application to use an attached resource such as search or storage in a enterprise generative-AI development environment, but the current requirement is instead to assess available models against language, modality, deployment, and licensing requirements. That separate capability does not prove that the team has selected a model family whose capabilities fit its evaluation criteria; Connecting a resource does not automatically create an embedding index inside it, so selecting that alternative would not meet the stated acceptance check.

C:

A Foundry project organizes application-oriented resources and connections for a team within the applicable Foundry environment; it may be used to isolate application assets, evaluations, and development work for one AI solution in a enterprise generative-AI development environment, but the current requirement is instead to assess available models against language, modality, deployment, and licensing requirements. That separate capability does not prove that the team has selected a model family whose capabilities fit its evaluation criteria; A project is not the model endpoint serving inference requests, so selecting that alternative would not meet the stated acceptance check.

D:

Regional support and deployment availability must be checked for the intended resource and model offering; it may be used to satisfy residency and model availability constraints before provisioning in a enterprise generative-AI development environment, but the current requirement is instead to assess available models against language, modality, deployment, and licensing requirements. That separate capability does not prove that the team has selected a model family whose capabilities fit its evaluation criteria; A deployment name or prompt template cannot relocate underlying service processing, so selecting that alternative would not meet the stated acceptance check.

Designing and Implementing a Microsoft Azure AI Solution (AI-102) Practice Tests