10 free Microsoft AI-103 practice questions with the correct answer and a full explanation for each, taken from the CertStash pack of 135 questions. Work through them, then open each answer to check your reasoning.
Get all 135 questions (US$39) · Download these 10 as a PDF
Question 1
Show the case study this question is based on
CASE STUDY
This is a case study. Case studies are not timed separately from other exam sections. You can use as much exam time as you would like to complete each case study. However, there might be additional case studies or other exam sections. Manage your time to ensure that you can complete all the exam sections in the time provided. Pay attention to the Exam Progress at the top of the screen so you have sufficient time to complete any exam sections that follow this case study.
To answer the case study questions, you will bed to reference information that is provided in the case. Case studies and associated questions might contain exhibits or other resources that provide more information about the scenario described in the case. Information provided in an individual question does not apply to the other questions in the case study.
A Review Screen will appear at the end of this case study. From the Review Screen, you can review and change your answers before you move to the next exam section. After you leave this case study, you will NOT be able to return to it.
To start the case study -To display the first question in this case study, select the “Next” button. To the left of the question, a menu provides links to information such as business requirements, the existing environment, and problem statements. Please read through all this information before answering any questions.
When you are ready to answer a question, select the “Question” button to return to the question.
Overview
Company Information
Contoso, Ltd is a multinational retail company that builds, deploys, and manages generative AI and agent-based solutions by using Microsoft Foundry.
Existing Environment
Identity Environment
Contoso uses Microsoft Entra ID for identity management, authentication, and authorization capabilities that enable agents to access organizational resources and services.
Contoso recently formed a new AI engineering team named Agent1Dev Team to optimize and maintain existing AI solutions.
The team collaborates with solution architects, DevOps engineers, and security engineers to design, implement. monitor, and secure AI applications.
Contoso also has a team named Agent1Test Team that is responsible for validating AI solutions before the solution deployments.
Generative Environment
Contoso has a Microsoft Foundry deployment that contains two projects named Project1 and Project2.
Project1
Project1 contains a customer support agent named Agent1 that assists customers with product inquiries and troubleshooting requests.
Agent1 has the following configurations:
Agent1 uses a base model deployment.
A safety evaluation pipeline is NOT enabled.
Tool invocation approval workflows are NOT enabled.
Conversation memory constraints are NOT configured.
Agent1 interacts with customers by using digital support channels and answers general questions about Contoso products.
Project1 is deployed to an Azure region located in the European Union (EU).
Agent1Dev Team will use Project1 to optimize and maintain Agent1.
Project2
Project2 contains a deployed video generation model. The marketing department at Contoso has access to Project2 and plans to use the model to develop a video creation solution.
Development of the solution is incomplete.
Data Environment
Contoso stores product-related information in Azure resources that support AI applications.
The Azure environment contains an Azure Blob Storage account named storage1 that stores product detail sheets for all the Contoso products.
The product sheets include specifications, feature descriptions, and product support information that Agent1 can use to answer customer questions. The product sheets are stored in the PDF format.
Problem Statements
Contoso identifies the following issues:
Agent1 has only general knowledge of the Contoso products.
A recent chat interaction with Agent1 was analyzed for sentiment. The results of the analysis have NOT been processed yet.
Agent1 does NOT use the detailed product information in the product sheets stored in storage1 when responding to customer questions.
The finance department at Contoso reports that vendor invoices must be reviewed manually to ensure that the invoices match the terms defined in the vendor contracts. The invoices contain tables, logos, and varied layouts that make the documents difficult to process consistently.
Requirements
Planned Changes
Contoso plans to implement the following changes:
Implement a solution for Project1 that analyzes the vendor invoices by evaluating both the visual layout and the textual content of the invoices, so that the invoice details can be verified against the vendor contract terms.
Update the base model deployment used by Agent1 and standardize the model version to ensure continuity and consistent responses.
Enable Agent1 to retrieve and use the detailed product information from the product sheets stored in storage1.
Implement an indexing solution for the product sheets that Agent1 can use to answer customer questions.
Complete the development of the video creation solution.
Technical Requirements
Contoso identifies the following technical requirements:
The model deployment used by Agent1 must support scalable, high-throughput generative AI workloads and dynamically scale to handle variable customer support traffic, without requiring reserved throughput capacity.
The product sheets must be processed by using an indexing pipeline that enables semantic and vector search, so that Agent1 can retrieve the relevant product information.
Responses generated by using the product sheet information must be relevant, complete, and accurate.
Agent1 must be able to use the product sheets to answer natural language questions about product details.
The model version used by Agent1 must remain consistent to ensure stable responses.
The data processed by the model must remain within the EU.
Security and Compliance Requirements
Contoso identifies the following security and compliance requirements:
API keys must NOT be used to access Foundry-deployed models.
Access to the Azure resources must follow the principle of least privilege.
The developers at Contoso must authenticate to Microsoft Foundry resources by using Microsoft Entra authentication.
Access to Project1 must be assigned to the members of Agent1Dev Team by using a security group named SC_Agent1_Dev.
Access to Project1 must be assigned to the members of Agent1Test Team by using a security group named SC_Agent1_Test.
Agent1 must never reveal customer information, even if a document that contains customer data is added erroneously to the product sheet repository in storage1.
The product sheets might contain images that include embedded text. Agent1 must be protected from malicious instructions potentially hidden within the images.
Business Requirements
Contoso identifies the following business requirements:
Users that interact with Agent1 must have a personalized experience in future interactions, including the ability for Agent1 to retain conversation context and recall relevant information from previous interactions.
Agent1 must answer questions only about the products sold by Contoso.
You need to configure Agent1 to meet the security and compliance requirements.
What should you use?
Show answer and explanation
Correct answer: B. prompt shields
Prompt shields protect agents from malicious instructions potentially hidden within images and user inputs by detecting and mitigating prompt injection attacks. This directly addresses the requirement that Agent1 must be protected from malicious instructions hidden within images in the product sheets. While PII detection (option C) is also important for protecting customer information, the specific threat mentioned, malicious instructions hidden in images, is the primary concern addressed by prompt shields as a content safety mechanism.
Why the other options are wrong
- A. Self-harm content filtering prevents responses about self-harm, not malicious instructions in images.
- C. PII detection protects existing customer data from being revealed, but doesn't prevent malicious instructions hidden in images.
- D. Violence content filtering prevents violent content responses, not malicious instructions in images.
Question 2
You are planning a Microsoft Foundry project named Project1 that will contain multiple agents. Each agent will access the same Azure AI Search resource.
You need to recommend a solution to centrally manage the Azure AI Search credentials within Project1. The solution must be implemented across all the agents.
What should you recommend?
Show answer and explanation
Correct answer: C. Add a connection to the Azure AI Search resource.
Adding a connection to the Azure AI Search resource in Microsoft Foundry provides centralized credential management that can be shared across all agents within Project1. Connections in Foundry securely store and manage credentials, eliminating the need for API keys and enabling all agents to access the same Azure AI Search resource through a single managed credential. This approach follows the principle of least privilege and avoids hardcoding credentials.
Why the other options are wrong
- A. RBAC controls access permissions but doesn't centrally manage credentials for use across agents.
- B. Disabling key-based access alone doesn't provide a centralized credential management solution.
- D. Managed private endpoints secure network connectivity but don't manage credentials centrally.
Question 3
You have a Microsoft Foundry project that contains three agents as shown in the following table.
You need to orchestrate the agents to ensure that the customer requests meet the following requirements:
Support a deterministic, step-based process that uses conditional branching and shared state across the agents.
Optionally trigger a ticket action based on the triage result.
The solution must minimize development effort.
What should you include in the solution?
Show answer and explanation
Correct answer: A. a workflow
A workflow is the appropriate choice because it provides deterministic, step-based process execution with conditional branching and maintains shared state across the three agents (TriageAgent, PolicyAgent, and ActionAgent). Workflows enable sequential control flow where the triage result can conditionally trigger the ActionAgent to create tickets, while minimizing development effort through declarative composition rather than imperative code. This satisfies all stated requirements: deterministic steps, conditional logic, cros-gent state sharing, and optional ticket action triggering based on triage outcomes.
Why the other options are wrong
- B. Threads and runs without a workflow lack deterministic step-based orchestration and would require manual state management between agent invocations.
- C. Multi-agent group chat sessions operate autonomously without the deterministic ste-ased control and conditional branching required for this use case.
- D. Coordinating separate agent runs in application code maximizes development effort by requiring manual orchestration logic rather than using a declarative workflow framework.
Question 4
You have a Microsoft Foundry project that contains an agent. The agent uses Azure Speech in Foundry Tools.
You fine-tune a baseline speech to text model for the en-us locale and publish the model.
The agent calls the Speech to text REST API and returns an error message indicating that the project ID is invalid.
You need to set the project property to the correct ID.
To what should you set the project property?
Show answer and explanation
Correct answer: B. the custom speech project ID
The project property must be set to the custom speech project ID, which is the identifier created when you fine-tuned and published the baseline speech-to-text model. This ID is used by the Speech-to-text REST API to locate and reference the specific custom model deployment for the Speech service to use during inference.
Why the other options are wrong
- A. The project URL is used for browser access, not for API calls requiring the project property.
- C. The project ID refers to the Foundry project identifier, not the custom speech project identifier needed by the Speech API.
- D. The custom speech endpoint URL is used for direct service calls, not as the project property value.
Question 5
You have an Azure Speech in Foundry Tools resource that hosts a custom speech to text model deployed to a custom endpoint. An agent uses the endpoint to perform rea-ime speech recognition.
You are approaching the expiration date of the custom speech to text model.
What is the expected behavior when the model expires?
Show answer and explanation
Correct answer: A. Speech recognition requests will return a 4xx error until a new custom model is deployed.
Custom speech to text models have a published expiration date, and Azure Speech does not silently substitute another model when that date passes. Once the model behind the custom endpoint expires, recognition calls to that endpoint fail with a 4xx error, so you must retrain or redeploy the endpoint with a supported base or custom model to restore service. The correct planning action is to update the deployment before the expiration date arrives.
Why the other options are wrong
- B. An expired model is no longer usable for transcription, so the endpoint cannot keep serving requests with it until someone deletes it.
- C. There is no automatic fallback to the latest base model for the locale; the request simply fails until the endpoint is pointed at a supported model.
- D. Expiration only ends the model's supported lifetime for transcription, it does not trigger automatic deletion of the model or the endpoint.
Question 6
You have a Microsoft Foundry project that contains a model deployment.
You have an application that calls the deployment by using the Azure OpenAI v1 API and DefaultAzureCredential.
The developers at your company receive HTTP 403 errors when they send inference requests, even after running az login.
You need to ensure that the developers can perform model inference. The solution must follow the principle of least privilege.
Which role-based access control (RBAC) role should you assign to the developers?
Show answer and explanation
Correct answer: B. Cognitive Services OpenAI User
The Cognitive Services OpenAI User role grants the minimum permissions necessary to perform model inference operations on Azure OpenAI deployments. This role allows developers to make inference requests while adhering to the principle of least privilege, avoiding the overly broad permissions of Contributor and the read-only limitations of Data Reader.
Why the other options are wrong
- A. Cognitive Services User is a broader role that doesn't specifically grant OpenAI inference permissions.
- C. Contributor provides excessive permissions beyond what is needed for inference only.
- D. Cognitive Services Data Reader provides read access to data but not permissions to invoke model inference.
Question 7
You have a Microsoft Foundry project that contains an agent. The agent has a Model Context Protocol (MCP) tool that queries a knowledge base stored in Azure AI Search.
Some agent runs return answers from the base model without invoking the knowledge base, which results in responses without grounded citations.
You are provided with the following code snippet that runs the agent.
You need to add the correct tool _choice parameter to the code to deterministically force the agent to invoke the MCP tool on each run.
What should you add?
Show answer and explanation
Correct answer: D. tool_choice ={“type”:“mcp”}
To make tool invocation deterministic you pass a tool choice object that names the specific hosted tool, and for a Model Context Protocol tool that object is {"type":"mcp"}. With this set on the run, the agent must call the MCP tool that queries Azure AI Search on every execution, so answers are grounded and carry citations instead of coming from the base model.
Why the other options are wrong
- A. The value shown is a bare set literal rather than a valid tool choice object, and forcing any tool generically does not pin the run to the MCP tool.
- B. "auto" leaves the decision to the model, which is exactly the behavior producing ungrounded answers on some runs.
- C. "knowledge_base" is not a recognized tool type value, so the run would not be bound to the MCP tool.
Question 8
You have a Microsoft Foundry project named Project1 that contains an agent. The agent uses an OpenAPI 3.0 specification to call an external weather service.
The weather service requires a key to be passed in an HTTP header. The key value is stored as a connection in Project1.
You need to ensure that the key value from the connection is included automatically whenever the OpenAPI tool is invoked.
What should you configure in the OpenAPI specification?
Show answer and explanation
Correct answer: C. an API key security scheme
An API key security scheme in the OpenAPI 3.0 specification defines how security credentials should be passed to the service. When configured with a header parameter, the API key security scheme automatically includes the key value from the Project1 connection in the HTTP header for every operation invocation, without requiring manual inclusion in each call.
Why the other options are wrong
- A. Defining header parameters in each operation is manual configuration and doesn't leverage the connection automatically.
- B. Azure Key Vault connections are for credential storage, not for automatic inclusion in OpenAPI calls.
- D. Bearer token security scheme is used for bearer token authentication, not API key authentication in headers.
Question 9
You have a Microsoft Foundry project that serves a high-volume chat app.
Most requests are simple FAQs, but some require advanced reasoning.
You need to reduce costs and latency for common queries, without degrading the quality of the responses to complex questions.
What should you do?
Show answer and explanation
Correct answer: B. Use a model cascade that routes the requests to different models.
A model cascade routes different requests to different models based on complexity: simpler FAQ requests go to a smaller, faster, cheaper model, while complex questions requiring advanced reasoning go to a more capable model. This optimizes both cost and latency by matching request complexity to model capability, reducing costs for simple queries while maintaining quality for complex ones.
Why the other options are wrong
- A. Routing all requests to a smaller model would degrade quality for complex questions requiring advanced reasoning.
- C. Increasing max_tokens would increase latency and costs without addressing the routing of requests by complexity.
- D. Routing all requests to the most capable model eliminates cost optimization for simple FAQ queries.
Question 10
You have a Microsoft Foundry project that contains a high-traffic agent.
After a recent update, operational costs increase significantly.
Monitoring confirms that the volume of user traffic to the agent remains unchanged.
You suspect that changes to the request or response characteristics are causing the increase. You need to identify whether the additional costs are driven by the model input size, the model output size, or expanded tool usage.
Which observability capability should you use?
Show answer and explanation
Correct answer: D. token usage
Token usage observability provides detailed metrics on the number of tokens consumed by model input and output, as well as tool invocation counts. By examining token usage patterns, you can identify whether increased costs are driven by larger input sizes, longer output sizes, or expanded tool usage, directly pinpointing the cause of the cost increase after the recent update.
Why the other options are wrong
- A. Latency metrics measure response time, not the factors driving cost increases.
- B. Evaluation metrics assess response quality, not input/output size or tool usage characteristics.
- C. Run success rate indicates operational reliability, not the factors driving increased costs.
That was 10 of 135.
The full Microsoft AI-103 pack has all 135 questions, each with the answer, the explanation and why the other options are wrong, plus a questions-only copy for timed runs. US$39, paid once, with free monthly updates and a pass-or-your-money-back guarantee.