How should I use these Plan and manage an Azure AI solution practice questions?

Read each scenario carefully and choose your answer before revealing the explanation. Then check why your choice was right or wrong. Repeat until the reasoning feels automatic.

Can I practise just Plan and manage an Azure AI solution questions in a focused session?

Yes — use the session launcher on this page to start a 10-, 20-, 30- or 50-question session drawn entirely from the Plan and manage an Azure AI solution domain.

AI-102 · topic practice

Plan and manage an Azure AI solution practice questions

This domain covers provisioning, securing, monitoring, and cost-managing Azure AI resources: Azure OpenAI, Azure AI Services multi-service accounts, Azure AI Search, and Language endpoints. Questions test choosing keys versus Microsoft Entra ID, private endpoints, managed identity, diagnostic logging, and scaling or tier decisions that affect latency and spend.

Courseiva uses original exam-style practice questions designed for learning and revision. The goal is to understand the concepts, recognise exam patterns, and improve through explanations — not memorise copied exam dumps.

Editorial oversight:Johnson Ajibi· MSc IT Security, IEEE Senior Member

20 questionsDomain: Plan and manage an Azure AI solution

Practice 10 questions Browse domain →

What the exam tests

What to know about Plan and manage an Azure AI solution

Be able to provision and secure Azure AI resources, choose between keys and Microsoft Entra ID, and tune Azure AI Search capacity. The critical skill is matching each scenario to the right service and access mechanism without over-provisioning.

Configuring Azure OpenAI and Azure AI Search with managed identity, private endpoints, and key rotation

Securing Azure AI Language endpoints using Microsoft Entra ID authentication and role-based access control

Improving Azure AI Search query latency via replicas, partitions, and query configuration choices

Selecting the correct Azure AI service for vision, language, speech, and document intelligence scenarios

Watch out for

Common Plan and manage an Azure AI solution exam traps

▸Assuming API keys alone secure an endpoint, when tenant-restricted access requires Microsoft Entra ID authentication and RBAC roles.
▸Scaling Azure AI Search partitions to cut latency, when replicas increase query throughput and partitions mainly affect index size.
▸Confusing multi-service Azure AI Services keys with individual service keys, causing endpoint or billing mismatches in scenarios.

Practice set

Plan and manage an Azure AI solution questions

20 questions · select your answer, then reveal the explanation

Question 1mediummultiple choice

Read the full Plan and manage an AI solution explanation →

A company plans to deploy an Azure AI solution that processes sensitive customer data. The solution must comply with GDPR and ensure data residency within the European Union. Which Azure resource configuration should be used?

Trap 1: Deploy the Azure AI services in multiple regions globally and use…

Multi-region deployment may replicate data outside EU.

Trap 2: Use the Azure AI services 'Data Residency' SKU.

No such SKU exists; data residency is managed by region selection.

Trap 3: Use the Free tier of Azure AI services.

The Free tier does not guarantee data residency.

Study all Plan and manage an Azure AI solution common traps →

A
Deploy the Azure AI services in multiple regions globally and use geo-replication.
Why it fails: Multi-region deployment may replicate data outside EU.
B
Use the Azure AI services 'Data Residency' SKU.
Why it fails: No such SKU exists; data residency is managed by region selection.
C
Use the Free tier of Azure AI services.
Why it fails: The Free tier does not guarantee data residency.
D
Create an Azure AI services resource in a specific EU region and set the data residency option to 'EU'.
Azure AI services allow you to choose a region to control data residency.

Full breakdown with real-world context →

Question 2hardmultiple choice

Read the full Plan and manage an AI solution explanation →

A company is deploying a custom vision model using Azure Custom Vision. The training data contains images with varying resolutions. The model must achieve high accuracy. Which pre-processing step should be applied to the images before training?

Trap 1: Convert images to grayscale.

Color information may be important for the model.

Trap 2: Normalize pixel values to a range of 0-1.

Custom Vision handles normalization internally.

Trap 3: Apply data augmentation techniques like random cropping.

Data augmentation is typically done during training, not as a mandatory pre-processing step.

Study all Plan and manage an Azure AI solution common traps →

A
Resize all images to the same dimensions (e.g., 224x224).
Custom Vision expects consistent image sizes for optimal performance.
B
Convert images to grayscale.
Why it fails: Color information may be important for the model.
C
Normalize pixel values to a range of 0-1.
Why it fails: Custom Vision handles normalization internally.
D
Apply data augmentation techniques like random cropping.
Why it fails: Data augmentation is typically done during training, not as a mandatory pre-processing step.

Full breakdown with real-world context →

Question 3hardmulti select

Read the full Plan and manage an AI solution explanation →

A company is planning to use Azure AI Document Intelligence to extract data from invoices. The solution must handle variations in invoice formats. Which TWO actions should be taken to improve accuracy?

Trap 1: Avoid using custom models; rely on prebuilt models only.

Prebuilt invoice models cover common layouts only; format variation across suppliers demands custom neural or composed models trained on your own labelled samples. Relying on prebuilt models alone is tempting because it avoids training effort, and would be correct for standardised invoices matching the prebuilt schema.

Trap 2: Use a large number of unlabeled samples.

Unlabelled samples cannot train a custom extraction model; Document Intelligence custom models require labelled samples with tagged fields and key-value pairs. Large unlabelled volumes are tempting because they are cheap to collect, and would be correct for unsupervised clustering or pre-training rather than supervised field extraction.

Trap 3: Use a single prebuilt model for all invoices.

A single prebuilt model cannot accommodate differing supplier layouts, so extraction accuracy drops on invoices that deviate from its trained schema. Consolidating on one model is tempting because it simplifies deployment, and would be correct where all invoices share one consistent, standardised format.

Study all Plan and manage an Azure AI solution common traps →

A
Avoid using custom models; rely on prebuilt models only.
Why it fails: Prebuilt invoice models cover common layouts only; format variation across suppliers demands custom neural or composed models trained on your own labelled samples. Relying on prebuilt models alone is tempting because it avoids training effort, and would be correct for standardised invoices matching the prebuilt schema.
B
Include samples of different layouts in the training set.
Including samples of different layouts teaches the custom model to generalise across format variation rather than memorising one template. This satisfies the stem's requirement to handle variations in invoice formats, improving extraction accuracy on unseen layouts.
C
Use a large number of unlabeled samples.
Why it fails: Unlabelled samples cannot train a custom extraction model; Document Intelligence custom models require labelled samples with tagged fields and key-value pairs. Large unlabelled volumes are tempting because they are cheap to collect, and would be correct for unsupervised clustering or pre-training rather than supervised field extraction.
D
Use a single prebuilt model for all invoices.
Why it fails: A single prebuilt model cannot accommodate differing supplier layouts, so extraction accuracy drops on invoices that deviate from its trained schema. Consolidating on one model is tempting because it simplifies deployment, and would be correct where all invoices share one consistent, standardised format.
E
Train a custom model with labeled invoice samples.
A custom model trained on labelled invoice samples learns the specific field locations and terminology of the company's documents, outperforming the prebuilt invoice model when formats vary. This satisfies the stem's accuracy requirement across differing invoice layouts.

Full breakdown with real-world context →

Question 4hardmulti select

Read the full Plan and manage an AI solution explanation →

You are deploying an Azure AI solution that uses multiple services. You need to manage access keys securely. Which TWO methods should you use?

Trap 1: Embed the keys directly in application code.

Embedding keys in code is a security risk.

Trap 2: Store keys in Azure App Configuration.

App Configuration is for configuration, not secrets; Key Vault is preferred.

Trap 3: Use Key Vault's automatic key rotation feature.

Key Vault can rotate keys, but it's not the primary method for access management; the question asks for methods to manage keys securely.

Study all Plan and manage an Azure AI solution common traps →

A
Use Azure Key Vault to store the keys.
Key Vault securely stores secrets like access keys.
B
Use managed identities for Azure resources where possible.
Managed identities eliminate the need for keys by providing an automatically managed identity.
C
Embed the keys directly in application code.
Why it fails: Embedding keys in code is a security risk.
D
Store keys in Azure App Configuration.
Why it fails: App Configuration is for configuration, not secrets; Key Vault is preferred.
E
Use Key Vault's automatic key rotation feature.
Why it fails: Key Vault can rotate keys, but it's not the primary method for access management; the question asks for methods to manage keys securely.

Full breakdown with real-world context →

Question 5hardmultiple choice

Read the full Plan and manage an AI solution explanation →

You are an Azure AI engineer at Contoso Ltd. The company has an Azure AI solution that uses Azure AI Language to analyze customer feedback. The solution is deployed in the East US region and uses the S0 pricing tier. Recently, the volume of feedback has increased significantly, causing the service to throttle requests. The application logs show HTTP 429 (Too Many Requests) errors during peak hours. The development team has already implemented retry logic with exponential backoff, but the errors persist. You need to recommend a solution to handle the increased load without changing the application code. The solution must minimize cost. What should you do?

Trap 1: Use Azure Front Door to cache responses and reduce load on the…

Front Door caches static content, not dynamic API responses.

Trap 2: Switch the Azure AI Language resource to the Free tier to reduce…

Free tier has lower rate limits, increasing throttling.

Trap 3: Move the service to a different Azure region with higher capacity.

Moving regions does not change the capacity of a single resource.

Study all Plan and manage an Azure AI solution common traps →

A
Upgrade the Azure AI Language resource to a higher pricing tier (e.g., S1) or create additional resources and load balance.
Higher tiers provide higher rate limits and throughput.
B
Use Azure Front Door to cache responses and reduce load on the service.
Why it fails: Front Door caches static content, not dynamic API responses.
C
Switch the Azure AI Language resource to the Free tier to reduce costs.
Why it fails: Free tier has lower rate limits, increasing throttling.
D
Move the service to a different Azure region with higher capacity.
Why it fails: Moving regions does not change the capacity of a single resource.

Full breakdown with real-world context →

Question 6hardmultiple choice

Read the full Plan and manage an AI solution explanation →

A company plans to use Azure Cognitive Search to index a large number of PDF documents stored in Azure Blob Storage. The documents contain sensitive personally identifiable information (PII). The search results should only be accessible to authorized users based on their Azure Active Directory (Azure AD) group membership. Which combination of features should the company implement?

Trap 1: Generate shared access signatures (SAS) for each document and…

SAS tokens grant direct blob access, not search-result trimming, and embedding them in an index exposes credentials to every client. The correct option applies document-level security filters evaluated against the caller's Microsoft Entra ID group claims at query time. SAS suits time-limited external sharing of individual storage objects.

Trap 2: Use Azure RBAC to assign permissions to users for the search…

Azure RBAC on the search service grants control-plane and query access, but field-level security is not a native Azure Cognitive Search feature, so PII fields cannot be restricted per group this way. Document-level trimming via security filters is the correct mechanism for Entra ID group-based access.

Trap 3: Store Microsoft Entra ID group membership in a field in the search…

Index fields are not evaluated at query time against the caller's token, so storing group membership there enforces nothing; the indexer merely copies blob metadata. Security filters, which the correct option uses, apply document-level access based on the authenticated user's Microsoft Entra ID groups. Storing membership suits faceting or display, not authorisation.

Study all Plan and manage an Azure AI solution common traps →

A
Generate shared access signatures (SAS) for each document and include them in the search index.
Why it fails: SAS tokens grant direct blob access, not search-result trimming, and embedding them in an index exposes credentials to every client. The correct option applies document-level security filters evaluated against the caller's Microsoft Entra ID group claims at query time. SAS suits time-limited external sharing of individual storage objects.
B
Index the documents and use security filters with Microsoft Entra ID group-based access control.
Security filters in Azure Cognitive Search apply document-level trimming at query time, using the end user's Microsoft Entra ID group memberships passed via the query request. This satisfies the stem's constraint that results remain visible only to authorised users, since filtering happens before results return rather than relying on index-level permissions.
C
Use Azure RBAC to assign permissions to users for the search service and configure field-level security.
Why it fails: Azure RBAC on the search service grants control-plane and query access, but field-level security is not a native Azure Cognitive Search feature, so PII fields cannot be restricted per group this way. Document-level trimming via security filters is the correct mechanism for Entra ID group-based access.
D
Store Microsoft Entra ID group membership in a field in the search index and configure an indexer to map the field.
Why it fails: Index fields are not evaluated at query time against the caller's token, so storing group membership there enforces nothing; the indexer merely copies blob metadata. Security filters, which the correct option uses, apply document-level access based on the authenticated user's Microsoft Entra ID groups. Storing membership suits faceting or display, not authorisation.

Full breakdown with real-world context →

Question 7hardmulti select

Read the full Plan and manage an AI solution explanation →

A company is designing a solution that uses Azure Cognitive Services for text analytics. The solution must meet the following requirements: - Detect sentiment in customer feedback. - Extract key phrases from the feedback. - Identify the language of the feedback automatically. - Ensure that the solution can scale to handle thousands of requests per second. Which TWO actions should the company take?

Trap 1: Deploy a separate Language Detection service for language…

Text Analytics API includes language detection.

Trap 2: Use the Free pricing tier to reduce costs.

Free tier has low throughput limits.

Trap 3: Use Azure Functions to aggregate results from multiple API calls.

Not required; the API handles results per request.

Study all Plan and manage an Azure AI solution common traps →

A
Deploy a separate Language Detection service for language identification.
Why it fails: Text Analytics API includes language detection.
B
Use the Text Analytics API for sentiment analysis and key phrase extraction.
Text Analytics API provides these features.
C
Use the Free pricing tier to reduce costs.
Why it fails: Free tier has low throughput limits.
D
Deploy the Cognitive Services resource in multiple regions and use a load balancer.
Multi-region deployment with load balancing supports high throughput.
E
Use Azure Functions to aggregate results from multiple API calls.
Why it fails: Not required; the API handles results per request.

Full breakdown with real-world context →

Question 8hardmultiple choice

Read the full Plan and manage an AI solution explanation →

You are an AI engineer at a healthcare company. The company uses Azure Cognitive Services to process medical records. They have a Computer Vision resource deployed in the East US region. Recently, they implemented a custom vision model for detecting specific anomalies in X-ray images. The model was trained using the Custom Vision portal and exported as a TensorFlow model. They deployed the model to an Azure Kubernetes Service (AKS) cluster using a Docker container. The container runs the model and exposes a REST API endpoint for inference. The endpoint is used by a web application that is also hosted in the same AKS cluster. The web application is experiencing high latency when making inference requests. The latency spikes up to 10 seconds during peak hours. The AKS cluster has autoscaling enabled based on CPU metrics. The container's resource limits are set to 1 CPU core and 2 GB memory. The model's inference time on a single image is approximately 500 ms on the development machine. The team has not changed the model or the application code recently. The number of concurrent users has increased by 50% in the last month. What should you do to reduce inference latency?

Trap 1: Increase the memory limit for the container to 8 GB.

Memory is not the bottleneck; CPU is.

Trap 2: Deploy additional replicas of the container and use a load balancer.

Adding replicas with a load balancer distributes requests across multiple identical containers, which reduces per-container CPU load but does not address the root cause: the model’s 500 ms inference time per image is a fixed per-request latency that cannot be shortened by horizontal scaling. This option is tempting because it is the standard remedy for throughput bottlenecks under increased concurrent user load, and it would be correct if the latency were caused by queueing delays from insufficient processing capacity rather than by the model’s own single-image inference duration.

Trap 3: Optimize the model by quantizing it to reduce inference time.

Inference time per image is already low; the issue is concurrency.

Study all Plan and manage an Azure AI solution common traps →

A
Increase the memory limit for the container to 8 GB.
Why it fails: Memory is not the bottleneck; CPU is.
B
Deploy additional replicas of the container and use a load balancer.
Why it fails: Adding replicas with a load balancer distributes requests across multiple identical containers, which reduces per-container CPU load but does not address the root cause: the model’s 500 ms inference time per image is a fixed per-request latency that cannot be shortened by horizontal scaling. This option is tempting because it is the standard remedy for throughput bottlenecks under increased concurrent user load, and it would be correct if the latency were caused by queueing delays from insufficient processing capacity rather than by the model’s own single-image inference duration.
C
Optimize the model by quantizing it to reduce inference time.
Why it fails: Inference time per image is already low; the issue is concurrency.
D
Increase the CPU limit for the container to 4 cores.
More CPU allows handling more concurrent requests, reducing latency.

Full breakdown with real-world context →

Question 9mediummultiple choice

Read the full Plan and manage an AI solution explanation →

A company is building a custom question-answering solution using Azure AI Language. They need to ensure that the model can provide answers from a set of internal documents, but only to authenticated users from the company's Azure Active Directory tenant. The solution should minimize latency and cost. Which deployment option should the team choose?

Trap 1: Use the public endpoint of Azure AI Language with Microsoft Entra…

Public endpoint does not inherently restrict access to the company's tenant; it requires additional network controls.

Trap 2: Deploy a serverless endpoint with Azure Functions and Azure…

Serverless may have cold-start latency and is not ideal for frequent queries; also introduces more complexity.

Trap 3: Use the prebuilt question-answering model from Azure AI Language…

Prebuilt models cannot be customized with internal documents; they only answer from a predefined FAQ.

Study all Plan and manage an Azure AI solution common traps →

A
Deploy a custom question-answering model using a dedicated Azure AI Language resource with a private endpoint and managed identity.
A dedicated Azure AI Language resource with a private endpoint and managed identity keeps document data off public networks and authenticates via Microsoft Entra ID, satisfying the tenant-only access constraint while avoiding the added latency and cost of a multi-service or containerised deployment.
B
Use the public endpoint of Azure AI Language with Microsoft Entra ID token-based authentication.
Why it fails: Public endpoint does not inherently restrict access to the company's tenant; it requires additional network controls.
C
Deploy a serverless endpoint with Azure Functions and Azure Cognitive Search.
Why it fails: Serverless may have cold-start latency and is not ideal for frequent queries; also introduces more complexity.
D
Use the prebuilt question-answering model from Azure AI Language with a custom answer list.
Why it fails: Prebuilt models cannot be customized with internal documents; they only answer from a predefined FAQ.

Full breakdown with real-world context →

Question 10easymultiple choice

Read the full Plan and manage an AI solution explanation →

A developer is configuring an Azure AI Language resource for sentiment analysis. The solution must process social media posts in real-time with a throughput of 1000 requests per minute. After testing, the developer notices that the API returns a 429 (Too Many Requests) error when the load exceeds 500 requests per minute. What is the most likely cause and solution?

Trap 1: Scale out the resource by creating multiple Azure AI Language…

Rate limits are per resource; multiple instances would increase total throughput but not the limit per instance.

Trap 2: Implement retry logic with exponential backoff to handle 429 errors.

This is a workaround but does not address the root cause of insufficient capacity.

Trap 3: Use Azure API Management to cache responses and reduce calls.

Caching may not help with real-time processing and does not resolve the rate limit issue.

Study all Plan and manage an Azure AI solution common traps →

A
Scale out the resource by creating multiple Azure AI Language instances and load balancing requests.
Why it fails: Rate limits are per resource; multiple instances would increase total throughput but not the limit per instance.
B
Upgrade the Azure AI Language resource to a higher tier (e.g., Standard S) to increase the rate limit.
Upgrading the Azure AI Language resource tier directly raises the provisioned transactions-per-second quota, which is the constraint being hit at 500 requests per minute. The 429 responses indicate throttling against the tier's rate limit, so moving to a higher tier such as Standard S increases the permitted throughput to satisfy the 1000 requests per minute requirement.
C
Implement retry logic with exponential backoff to handle 429 errors.
Why it fails: This is a workaround but does not address the root cause of insufficient capacity.
D
Use Azure API Management to cache responses and reduce calls.
Why it fails: Caching may not help with real-time processing and does not resolve the rate limit issue.

Full breakdown with real-world context →

Question 11hardmultiple choice

Read the full Plan and manage an AI solution explanation →

A healthcare organization is deploying an Azure AI Document Intelligence solution to extract medical record data. The solution must comply with HIPAA and process sensitive patient data. The data must be encrypted at rest and in transit, and access must be audited. The solution will be used by multiple departments, each requiring different levels of access. Which combination of Azure services should the architect recommend?

Trap 1: Azure AI Document Intelligence with public endpoint, Microsoft…

Public endpoint exposes the service to the internet; private endpoints are required for HIPAA compliance.

Trap 2: Azure AI Document Intelligence with public endpoint, Microsoft…

Public endpoint is not secure for sensitive data; lacks private connectivity.

Trap 3: Azure AI Document Intelligence with service endpoints, shared…

SAS tokens are less secure than managed identities and service endpoints do not provide the same level of isolation as private endpoints.

Study all Plan and manage an Azure AI solution common traps →

A
Azure AI Document Intelligence with public endpoint, Microsoft Entra ID authentication, and Azure Storage with customer-managed keys (CMK).
Why it fails: Public endpoint exposes the service to the internet; private endpoints are required for HIPAA compliance.
B
Azure AI Document Intelligence with public endpoint, Microsoft Entra ID authentication, and Azure Storage with encryption at rest using Microsoft-managed keys.
Why it fails: Public endpoint is not secure for sensitive data; lacks private connectivity.
C
Azure AI Document Intelligence with private endpoints, managed identity authentication, Azure Storage with customer-managed keys (CMK), and Azure RBAC for access control.
Private endpoints remove public exposure, satisfying the in-transit encryption and audit constraints, while managed identity eliminates stored credentials. Customer-managed keys encrypt storage at rest under organisational control, and Azure RBAC delivers the per-department access tiers the scenario demands.
D
Azure AI Document Intelligence with service endpoints, shared access signature (SAS) tokens, and Azure Storage with encryption at rest using Microsoft-managed keys.
Why it fails: SAS tokens are less secure than managed identities and service endpoints do not provide the same level of isolation as private endpoints.

Full breakdown with real-world context →

Question 12mediumdrag order

Read the full Plan and manage an AI solution explanation →

Drag and drop the steps to create a custom Azure AI Translator model into the correct order.

Drag or tap steps into the slots.

Steps

Order

1Step 1

2Step 2

3Step 3

4Step 4

5Step 5

Trap 1: Create Custom Translator resource, Prepare parallel data, Upload…

This order is incorrect because while creating the resource early is possible, it is more logical to prepare the parallel data first to ensure you have the data before creating the resource. The official workflow typically starts with data preparation.

Trap 2: Upload parallel data, Create Custom Translator resource, Train…

This order is incorrect because you cannot upload parallel data before creating the Custom Translator resource, as the resource is required to provide the storage and model management capabilities.

Trap 3: Prepare parallel data, Create Custom Translator resource, Train…

This order is incorrect because you must upload the parallel data before training the model; the training process requires the data to be available in the model workspace.

Study all Plan and manage an Azure AI solution common traps →

A
Prepare parallel data, Create Custom Translator resource, Upload parallel data, Train model, Deploy model
This is the correct order because you must first have the parallel data ready, then create the resource to host the model, upload the data, train the model using that data, and finally deploy the trained model for use.
B
Create Custom Translator resource, Prepare parallel data, Upload parallel data, Train model, Deploy model
Why it fails: This order is incorrect because while creating the resource early is possible, it is more logical to prepare the parallel data first to ensure you have the data before creating the resource. The official workflow typically starts with data preparation.
C
Upload parallel data, Create Custom Translator resource, Train model, Deploy model
Why it fails: This order is incorrect because you cannot upload parallel data before creating the Custom Translator resource, as the resource is required to provide the storage and model management capabilities.
D
Prepare parallel data, Create Custom Translator resource, Train model, Upload parallel data, Deploy model
Why it fails: This order is incorrect because you must upload the parallel data before training the model; the training process requires the data to be available in the model workspace.

Full breakdown with real-world context →

Question 13mediumdrag order

Read the full Plan and manage an AI solution explanation →

Drag and drop the steps to troubleshoot a failed Azure AI Search indexer execution into the correct order.

Drag or tap steps into the slots.

Steps

Order

1Step 1

2Step 2

3Step 3

4Step 4

5Step 5

Trap 1: Step 1: Verify data source connectivity, Step 2: Check indexer…

This is incorrect because verifying the data source without first checking the execution history may miss context about the failure and waste time if the issue is elsewhere.

Trap 2: Step 1: Review error messages, Step 2: Check indexer execution…

This is incorrect because reviewing error messages before checking the execution history is premature; the history provides context for understanding the errors.

Trap 3: Step 1: Check skillset execution, Step 2: Verify data source…

This is incorrect because checking the skillset before verifying the data source and history overlooks more common failure points.

Study all Plan and manage an Azure AI solution common traps →

A
Step 1: Check indexer execution history, Step 2: Review error messages, Step 3: Verify data source connectivity, Step 4: Check skillset execution, Step 5: Re-run the indexer
This is the correct order because you should first examine the execution history to understand the failure context, then review specific error messages, verify the data source as the most common cause, check the skillset for transformation issues, and finally re-run after resolving issues.
B
Step 1: Verify data source connectivity, Step 2: Check indexer execution history, Step 3: Review error messages, Step 4: Check skillset execution, Step 5: Re-run the indexer
Why it fails: This is incorrect because verifying the data source without first checking the execution history may miss context about the failure and waste time if the issue is elsewhere.
C
Step 1: Review error messages, Step 2: Check indexer execution history, Step 3: Check skillset execution, Step 4: Verify data source connectivity, Step 5: Re-run the indexer
Why it fails: This is incorrect because reviewing error messages before checking the execution history is premature; the history provides context for understanding the errors.
D
Step 1: Check skillset execution, Step 2: Verify data source connectivity, Step 3: Check indexer execution history, Step 4: Review error messages, Step 5: Re-run the indexer
Why it fails: This is incorrect because checking the skillset before verifying the data source and history overlooks more common failure points.

Full breakdown with real-world context →

Question 14mediummatching

Read the full Plan and manage an AI solution explanation →

Match each Azure AI concept to its description.

Drag a concept onto its matching description — or click a concept then click the description.

Concepts

Matches

URL to access a cognitive service

Authentication credential for API access

Azure datacenter location for the service

Container for related Azure resources

Billing and access management container

Trap 1: Azure AI Perception: Detects and monitors offensive or risky content

Incorrect — this describes Content Safety, not Perception.

Trap 2: Azure AI Decision: Creates conversational agents and speech-based…

Incorrect — this describes Communication, not Decision.

Study all Plan and manage an Azure AI solution common traps →

A
Azure AI Perception: Processes visual and auditory data to understand the world
Correct because Perception services like Computer Vision and Speech to text handle visual and auditory data.
B
Azure AI Communication: Creates conversational agents and speech-based interactions
Correct because Communication services include Bot Service and Speech.
C
Azure AI Decision: Provides tools for building decision-making and recommendation systems
Correct because Decision services include Personalizer and Anomaly Detector.
D
Azure AI Content Safety: Detects and monitors offensive or risky content
Correct because Content Safety detects inappropriate content.
E
Azure AI Perception: Detects and monitors offensive or risky content
Why it fails: Incorrect — this describes Content Safety, not Perception.
F
Azure AI Decision: Creates conversational agents and speech-based interactions
Why it fails: Incorrect — this describes Communication, not Decision.

Full breakdown with real-world context →

Question 15mediummultiple choice

Read the full Plan and manage an AI solution explanation →

You are deploying an Azure AI Search solution that indexes customer support tickets. The solution must support fuzzy search for misspelled keywords and boost results from premium customers. Which two features should you configure?

Trap 1: Synonym maps

Synonym maps expand search terms but do not handle misspellings or boosting.

Trap 2: Suggesters

Suggesters enable autocomplete and did-you-mean, not boosting.

Trap 3: Custom analyzers

Custom analyzers control tokenization but do not provide boosting.

Study all Plan and manage an Azure AI solution common traps →

A
Scoring profiles
Scoring profiles allow boosting based on fields like customer tier.
B
Synonym maps
Why it fails: Synonym maps expand search terms but do not handle misspellings or boosting.
C
Suggesters
Why it fails: Suggesters enable autocomplete and did-you-mean, not boosting.
D
Custom analyzers
Why it fails: Custom analyzers control tokenization but do not provide boosting.
E
Fuzzy search parameters
Fuzzy search handles misspellings by allowing approximate matches.

Full breakdown with real-world context →

Question 16mediummultiple choice

Read the full Plan and manage an AI solution explanation →

Refer to the exhibit. You are configuring an Azure AI Foundry agent for customer support. The agent uses Azure AI Search for retrieval and Azure OpenAI for generation. Users report that the agent provides correct answers but sometimes includes inappropriate language. What is the most likely cause?

Exhibit

{
  "features": {
    "search": {
      "type": "azure_ai_search",
      "index_name": "support-index",
      "semantic_configuration": "default"
    },
    "chat": {
      "type": "azure_openai",
      "deployment": "gpt-4o",
      "system_message": "You are a helpful assistant."
    },
    "content_safety": {
      "type": "azure_content_safety",
      "blocklist": "custom-blocklist",
      "threshold": "high"
    }
  }
}

Trap 1: The OpenAI deployment 'gpt-4o' does not support content filtering

Azure OpenAI deployments support content filtering regardless of model.

Trap 2: The content safety threshold should be set to 'low' to block more…

'High' is the most restrictive; 'low' would allow more content.

Trap 3: The semantic configuration is not optimized for safety

Semantic configuration affects relevance, not content safety.

Study all Plan and manage an Azure AI solution common traps →

A
The content safety blocklist is not applied to the chat output
Content safety filters applied to prompts do not automatically govern model completions; the output filter must be configured separately. Without a blocklist or category filter on the chat response, inappropriate language passes through even when retrieval and grounding are correct.
B
The OpenAI deployment 'gpt-4o' does not support content filtering
Why it fails: Azure OpenAI deployments support content filtering regardless of model.
C
The content safety threshold should be set to 'low' to block more content
Why it fails: 'High' is the most restrictive; 'low' would allow more content.
D
The semantic configuration is not optimized for safety
Why it fails: Semantic configuration affects relevance, not content safety.

Full breakdown with real-world context →

Question 17mediummultiple choice

Read the full Plan and manage an AI solution explanation →

A company uses Azure AI Search to index product catalogs. The search must support multilingual queries and return results in the user's language. What should you configure?

Trap 1: Create synonym maps for each language

Synonym maps expand terms but do not provide language-specific analysis.

Trap 2: Configure scoring profiles based on language

Scoring profiles affect ranking, not language processing.

Trap 3: Enable semantic search

Semantic search improves understanding but does not handle multiple languages natively.

Study all Plan and manage an Azure AI solution common traps →

A
Assign appropriate language analyzers to fields
Language analyzers apply language-specific tokenisation, stemming and stop-word rules to each field, so queries and indexed content are processed consistently per language. This directly satisfies the multilingual query and same-language result requirement, which generic analyzers cannot achieve for inflected or non-space-delimited languages.
B
Create synonym maps for each language
Why it fails: Synonym maps expand terms but do not provide language-specific analysis.
C
Configure scoring profiles based on language
Why it fails: Scoring profiles affect ranking, not language processing.
D
Enable semantic search
Why it fails: Semantic search improves understanding but does not handle multiple languages natively.

Full breakdown with real-world context →

Question 18mediummultiple choice

Read the full Plan and manage an AI solution explanation →

Refer to the exhibit. You are using Azure OpenAI to build a trivia bot. The bot responds correctly to simple questions but sometimes gives overly verbose or off-topic answers. Which parameter should you adjust to make responses more focused?

Exhibit

{
  "completions": [
    {
      "role": "user",
      "content": "What is the capital of France?"
    },
    {
      "role": "assistant",
      "content": "The capital of France is Paris."
    }
  ],
  "max_tokens": 50,
  "temperature": 0.7,
  "top_p": 0.95
}

Trap 1: Increase temperature

Increasing temperature makes responses more random and less focused.

Trap 2: Decrease top_p

Decreasing top_p can reduce randomness but temperature is more straightforward.

Trap 3: Increase max_tokens

Increasing max_tokens allows longer responses but does not improve focus.

Study all Plan and manage an Azure AI solution common traps →

A
Increase temperature
Why it fails: Increasing temperature makes responses more random and less focused.
B
Decrease top_p
Why it fails: Decreasing top_p can reduce randomness but temperature is more straightforward.
C
Increase max_tokens
Why it fails: Increasing max_tokens allows longer responses but does not improve focus.
D
Decrease max_tokens
Why it fails: Decreasing max_tokens cuts off responses but does not improve focus on the topic.
E
Decrease temperature
Temperature controls sampling randomness; decreasing it makes token selection more deterministic, so the model favours high-probability continuations and produces tighter, more focused responses. This reduces the verbose, off-topic drift without altering the prompt or model choice.

Full breakdown with real-world context →

Question 19easymultiple choice

Read the full Plan and manage an AI solution explanation →

A healthcare organization uses Azure AI Language to extract medical entities from clinical notes. The solution must comply with HIPAA. What should you configure?

Trap 1: Configure data residency to a specific region

Data residency addresses location requirements, not HIPAA-specific controls.

Trap 2: Enable content filtering

Content filtering blocks offensive content, not for HIPAA compliance.

Trap 3: Isolate the service in a virtual network

Network isolation is a security measure but not sufficient alone for HIPAA.

Study all Plan and manage an Azure AI solution common traps →

A
Use Azure AI Language with HIPAA-enabled service and enable encryption
Azure AI Language offers HIPAA-compliant tiers with data encryption.
B
Configure data residency to a specific region
Why it fails: Data residency addresses location requirements, not HIPAA-specific controls.
C
Enable content filtering
Why it fails: Content filtering blocks offensive content, not for HIPAA compliance.
D
Isolate the service in a virtual network
Why it fails: Network isolation is a security measure but not sufficient alone for HIPAA.

Full breakdown with real-world context →

Question 20mediummultiple choice

Read the full Plan and manage an AI solution explanation →

Refer to the exhibit. You deploy an Azure AI Services multi-service account using this ARM template. After deployment, developers cannot access the service from their local machines. What is the most likely reason?

Exhibit

{
  "$schema": "https://schema.management.azure.com/schemas/2019-04-01/deploymentTemplate.json#",
  "contentVersion": "1.0.0.0",
  "resources": [
    {
      "type": "Microsoft.CognitiveServices/accounts",
      "apiVersion": "2023-05-01",
      "name": "myaiservices",
      "location": "[resourceGroup().location]",
      "sku": {
        "name": "S0"
      },
      "kind": "AIServices",
      "properties": {
        "customSubDomainName": "myaiservices",
        "networkAcls": {
          "defaultAction": "Deny"
        }
      }
    }
  ]
}

Trap 1: The location is not specified correctly

An incorrect location typically causes deployment failure or regional latency, not blocked local access. Developers failing from local machines points to network access restrictions, such as the default action denying public traffic, which the correct option addresses.

Trap 2: The pricing tier S0 does not support network restrictions

S0 supports network restrictions; the failure stems from the template's default network ACL action denying public access, so local developer traffic is blocked. The option is tempting because S0 genuinely governs scale and feature availability, and tier selection would be the answer if the stem described quota or capability limits rather than connectivity.

Trap 3: The custom subdomain name is not globally unique

A non-unique custom subdomain would fail deployment outright, yet the stem states deployment succeeded, so it cannot explain the access failure. Custom subdomains matter because token-based authentication against Microsoft Entra ID requires one; had the template omitted it entirely, that would be the genuine cause.

Study all Plan and manage an Azure AI solution common traps →

A
The location is not specified correctly
Why it fails: An incorrect location typically causes deployment failure or regional latency, not blocked local access. Developers failing from local machines points to network access restrictions, such as the default action denying public traffic, which the correct option addresses.
B
The pricing tier S0 does not support network restrictions
Why it fails: S0 supports network restrictions; the failure stems from the template's default network ACL action denying public access, so local developer traffic is blocked. The option is tempting because S0 genuinely governs scale and feature availability, and tier selection would be the answer if the stem described quota or capability limits rather than connectivity.
C
The custom subdomain name is not globally unique
Why it fails: A non-unique custom subdomain would fail deployment outright, yet the stem states deployment succeeded, so it cannot explain the access failure. Custom subdomains matter because token-based authentication against Microsoft Entra ID requires one; had the template omitted it entirely, that would be the genuine cause.
D
The network ACLs block all traffic by default
Azure AI Services accounts default to a 'Deny' network ACL, so unless the template explicitly sets defaultAction to Allow or lists permitted IPs, all inbound traffic is refused. That default denial explains why developers' local machines cannot reach the service endpoint after deployment.

Full breakdown with real-world context →

Continue with 20-question session →

Free account

Track your progress over time

Create a free account to save your results and see which topics improve across sessions.

Focused Plan and manage an Azure AI solution sessions

Start a Plan and manage an Azure AI solution only practice session

Every question in these sessions is drawn from the Plan and manage an Azure AI solution domain — nothing else.

10 questions 20 questions 30 questions 50 questions

Browse all Plan and manage an Azure AI solution questions →Mixed AI-102 session

Frequently asked questions

What does the AI-102 exam test about Plan and manage an Azure AI solution?: Be able to provision and secure Azure AI resources, choose between keys and Microsoft Entra ID, and tune Azure AI Search capacity. The critical skill is matching each scenario to the right service and access mechanism without over-provisioning.
How should I use these practice questions?: Select your answer before revealing the explanation. Then read why each option is right or wrong — this active recall approach builds retention far faster than re-reading notes.
Can I practise just Plan and manage an Azure AI solution questions in a focused session?: Yes — the session launcher on this page draws every question from the Plan and manage an Azure AI solution domain. Use a 10-question session first to gauge your baseline, then move to 20 or 30 once the weak spots are clear.
Where can I practise other AI-102 topics?: Use the topic links above to move to related areas, or go back to the AI-102 question bank to see all topics.
Are these real exam questions or dumps?: These are original practice questions written to test the same concepts the AI-102 exam covers. They are not copied from any real exam or dump site.

Plan and manage an Azure AI solution only

10 questions 20 questions 30 questions 50 questions

Mixed AI-102 session

Track your progress

A free account saves results across sessions and highlights which topics need work.

Study resources

All AI-102 questions Plan and manage an Azure AI solution domain overview AI-102 exam guide

Exam traps to avoid

▸Assuming API keys alone secure an endpoint, when tenant-restricted access requires Microsoft Entra ID authentication and RBAC roles.
▸Scaling Azure AI Search partitions to cut latency, when replicas increase query throughput and partitions mainly affect index size.
▸Confusing multi-service Azure AI Services keys with individual service keys, causing endpoint or billing mismatches in scenarios.

Plan and manage an Azure AI solution practice questions

What to know about Plan and manage an Azure AI solution

Common Plan and manage an Azure AI solution exam traps

Plan and manage an Azure AI solution questions

A company plans to deploy an Azure AI solution that processes sensitive customer data. The solution must comply with GDPR and ensure data residency within the European Union. Which Azure resource configuration should be used?

A company is deploying a custom vision model using Azure Custom Vision. The training data contains images with varying resolutions. The model must achieve high accuracy. Which pre-processing step should be applied to the images before training?

A company is planning to use Azure AI Document Intelligence to extract data from invoices. The solution must handle variations in invoice formats. Which TWO actions should be taken to improve accuracy?

You are deploying an Azure AI solution that uses multiple services. You need to manage access keys securely. Which TWO methods should you use?

Drag and drop the steps to create a custom Azure AI Translator model into the correct order.

Drag and drop the steps to troubleshoot a failed Azure AI Search indexer execution into the correct order.

Match each Azure AI concept to its description.

You are deploying an Azure AI Search solution that indexes customer support tickets. The solution must support fuzzy search for misspelled keywords and boost results from premium customers. Which two features should you configure?

Refer to the exhibit. You are configuring an Azure AI Foundry agent for customer support. The agent uses Azure AI Search for retrieval and Azure OpenAI for generation. Users report that the agent provides correct answers but sometimes includes inappropriate language. What is the most likely cause?

Exhibit

A company uses Azure AI Search to index product catalogs. The search must support multilingual queries and return results in the user's language. What should you configure?

Refer to the exhibit. You are using Azure OpenAI to build a trivia bot. The bot responds correctly to simple questions but sometimes gives overly verbose or off-topic answers. Which parameter should you adjust to make responses more focused?

Exhibit

A healthcare organization uses Azure AI Language to extract medical entities from clinical notes. The solution must comply with HIPAA. What should you configure?

Refer to the exhibit. You deploy an Azure AI Services multi-service account using this ARM template. After deployment, developers cannot access the service from their local machines. What is the most likely reason?

Exhibit

Track your progress over time

Start a Plan and manage an Azure AI solution only practice session

Related AI-102 topic practice pages

Implement computer vision solutions practice questions

Implement knowledge mining and information extraction solutions practice questions

Implement generative AI solutions practice questions