Free AI-103 Practice Test – 20 Questions with Explanations

A DevOps agent can query repositories and open pull requests through an MCP server. The team wants read operations to stay frictionless, but any pull-request creation must require human approval.

Which value should replace (MISSING CONFIG)?

A team wants a repeatable promotion path for a Foundry solution across environments.

What is the correct order?

You have an Azure subscription named Sub1.
Sub1 contains a Microsoft Foundry project named Project1.
Project1 contains an agent named Agent1.
Agent1 must support tool use, runtime troubleshooting, evaluation, and controlled approval for high-impact actions.
You need to recommend the correct Microsoft Foundry capability or configuration for each requirement.

Which mapping should you use? Answer options:

Match the following:

Answer options:

A developer is defining a function tool for an agent that checks order status. The tool must require an order_id argument and should not accept unexpected extra fields.

Which missing configuration best completes the tool schema?

You need to choose the correct Content Understanding task configuration for each workflow.

Answer options:

A team runs a nightly document summarization pipeline for EU customers. The job is asynchronous, can complete any time within 24 hours, and the company wants the lower-cost batch option while keeping processing within the EU data zone.

Which value should replace (MISSING CONFIG)?

A team wants to build a workflow that turns uploaded videos into a retrieval-ready knowledge asset by using Azure Content Understanding.

What is the correct order?

You are designing speech integration for a Foundry agent that supports spoken customer interactions. Match each requirement to the most appropriate capability.

Answer options:

Match the following:

Answer options:

A team wants to build a RAG application over PDF files in Azure Blob Storage by using integrated vectorization and minimal custom code.

What is the correct order?

You are designing a Foundry-based agentic pipeline for employee onboarding. The solution must use tools, conditional workflow steps, and post-run review before production rollout.

What is the correct order?

A global support assistant uses Azure AI Search as its grounding store. The engineering lead wants a monitoring
design that meets all of the following requirements:
• Detect failed or skipped documents during scheduled ingestion
• Alert when search latency or throttling increases
• Let engineers inspect user query strings and identify long-running searches
• Use managed Azure capabilities with minimal custom code
Which design best meets the requirements?

A finance assistant agent can call internal tools to create wire requests. The security team wants harmful or policy-violating tool invocations to be detected before the action and payload are sent to the tool. Foundry guardrails support multiple intervention points for models and agents.
Which intervention point should you configure?

A Content Understanding analyzer processes annual report PDFs for a downstream reasoning agent. The agent often misses values that appear inside embedded charts and loses meaning from extracted tables. The current output is mostly plain text with limited figure context.
What should you change first?

A platform team is designing a visual moderation workflow.

Answer options:

A Foundry agent retrieves email content from a tool and then decides which action to take. During testing, the security team hides malicious instructions inside a retrieved document, and the agent follows them even though the original user question is benign. Microsoft documents this as an indirect prompt attack pattern and notes that document content should be identified clearly when passed in context.
What is the best next step?

A generative text analysis app must return a fixed JSON structure that includes entities, topics, summary, and risk category. The downstream workflow rejects responses when required properties are missing.

Which missing configuration should you use?

Match each requirement to the most appropriate translation capability or evaluation component.

Answer options:

Match each requirement to the best Foundry monitoring or evaluation capability.

Answer options: