AI-103 — Developing AI Apps and Agents on Azure
Microsoft

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

★★★★★★ 4.8 110 verified reviews
88 questions
June 15, 2026 updated
✓ Online quiz simulator
▶ Start Free Practice Exam

Domain coverage

  • Plan and manage an Azure AI solution
  • Implement generative AI and agentic solutions
  • Implement computer vision solutions
  • Implement text analysis solutions
  • Implement information extraction solutions

Sample Questions (9 of 88 shown)

Q1 Plan and Manage an Azure AI Solution (25–30%)
A Python support agent deploys to pre-production and production environments through the same pipeline. Each environment has its own Foundry project and model deployments, but both deployments use the same model family. Release gates must verify that production traffic uses the production project and expected deployment. What should you implement?
  1. Emit traces with the Foundry project endpoint and deployment name on each call
  2. Run safety evaluations on production responses
  3. Monitor aggregate token usage by model family
  4. Compare generated answers against a fixed golden dataset
✓ Correct Answer: A
The goal is to verify connection context, not model quality or safety. Trace logs that record the Foundry project endpoint and deployment name used by each request directly prove the application is using the expected production context.
Q2 Plan and Manage an Azure AI Solution (25–30%)
A Foundry project using CI/CD deploys an enterprise procurement agent. The approved baseline requires business system tools to use user-assigned managed identity and least-privilege role policies, and must prohibit storing API keys. A drift scan shows the erpLookup tool in production now authenticates via an API key stored in app configuration. What should you do?
  1. Add a prompt guardrail telling the model not to leak secrets
  2. Disable the erpLookup tool until the next monthly release
  3. Redeploy the approved managed identity tool configuration and remove the API key
  4. Leave the setting unchanged since endpoints and tool schema match
✓ Correct Answer: C
Drift occurred on the tool's authentication setting, not the tool schema. Restoring the approved managed identity and role policy eliminates the stored secret risk while not blocking authorized procurement order queries.
Q3 Plan and Manage an Azure AI Solution (25–30%)
A Foundry project hosts a support agent that uses Azure AI Search for knowledge retrieval and calls a refund-eligibility function. Before production, stakeholders ask if the agent gives correct policy answers, cites retrieved sources, and blocks unsafe refund guidance. Which evaluation plan best meets this request?
  1. Track token spend per conversation and select the lowest-cost model
  2. Compare p95 latency across different model deployments
  3. Run concurrent user load tests against the agent endpoint
  4. Evaluate traces for groundedness, answer correctness, safety, and tool call correctness
✓ Correct Answer: D
The scenario asks about output quality and agent behavior, not operational efficiency. A suitable Foundry evaluation plan should use conversation traces to test correctness, groundedness, safety, and whether the refund function was called appropriately.
Q4 Plan and Manage an Azure AI Solution (25–30%)
You maintain a Foundry RAG application that answers HR policy questions from documents indexed in Azure AI Search. The team changes retrieval from top-3 semantic hybrid results to top-5 and modifies the base prompt. You must verify this change does not reduce answer grounding or citation accuracy on known issues. What should you do?
  1. Run a baseline-candidate regression evaluation with groundedness checks
  2. Replace hybrid search with vector-only search for all queries
  3. Increase chunk size and redeploy after smoke testing
  4. Monitor production user feedback after deployment
✓ Correct Answer: A
Prompt and retrieval changes alter which passages are used and how answers cite them. Regression evaluation uses the same test cases against the current and candidate workflows to detect regressions in groundedness or citation before deployment.
Q5 Plan and Manage an Azure AI Solution (25–30%)
A Foundry project uses CI/CD to deploy an enterprise procurement agent. A drift scan reveals the erpLookup tool authenticates via API key instead of managed identity. The endpoint and tool schema remain unchanged. What should you do?
  1. Add a prompt guardrail
  2. Disable erpLookup
  3. Redeploy the approved managed identity tool configuration and remove the API key
  4. Leave as-is since endpoints match
✓ Correct Answer: C
The drift is in authentication settings, not tool schema. Redeploying the approved managed identity configuration restores compliance without changing functionality.
Q6 Plan and Manage an Azure AI Solution (25–30%)
You are designing a Microsoft Foundry HR benefits agent. The agent uses Azure AI Search for handbook knowledge retrieval and a payroll tool that can submit compensation change requests. The solution must block unsafe or prompt-injected input, avoid retrieval based on restricted content, limit tool calls to approved actions, prevent unsafe responses, and require HR approval before any payroll change is submitted. Which guardrail placement strategy should you use?
  1. Use system prompts to describe all restrictions and rely on conversation memory to avoid unsafe requests
  2. Run one safety check on the final model response only and send payroll changes to HR review after execution
  3. Guard input before retrieval, filter content after retrieval, enforce tool policy before execution, scan output before returning, and gate approval before submission
  4. Require HR approval for every user conversation turn and skip retrieval and tool guardrails
✓ Correct Answer: C
Guardrails should be placed as close to the risk they control as possible. Each checkpoint in this workflow requires a separate guardrail so unsafe content or unauthorized actions are blocked before they affect the next stage.
Q7 Plan and Manage an Azure AI Solution (25–30%)
A team in a Microsoft Foundry project builds a RAG-based enterprise assistant. New policy PDFs sync from SharePoint, undergo OCR/layout extraction, chunking, embedding, and indexing into Azure AI Search. Users report that some newly uploaded policies are not being referenced. The team has evaluated generated answer quality and needs a data ingestion quality observability signal. What should they monitor?
  1. Search click-through rate for referenced documents
  2. Source-to-index completeness and freshness per ingestion run
  3. LLM token usage and response latency per conversation
  4. Safety evaluator scores on generated responses
✓ Correct Answer: B
Ingestion quality monitoring should verify the path from source content to searchable grounded data. A useful signal is whether each source PDF is processed, chunked, embedded, indexed, and refreshed on time.
Q8 Plan and Manage an Azure AI Solution (25–30%)
A Foundry project support agent sees intermittent latency spikes during morning login. Model deployment traces show a 429 RateLimitExceeded pattern occurring in 2-minute bursts. The application currently retries up to three times immediately on failed calls. What is the best next fix?
  1. Increase the model temperature for the agent deployment
  2. Increase retry concurrency to clear the backlog
  3. Throttle and queue model calls respecting Retry-After headers
  4. Rebuild the search index with larger vectors
✓ Correct Answer: C
The symptom is a model deployment rate limit. The 429 status with token-per-minute limit message and retry-after-ms header indicates the application should slow down and schedule requests rather than retrying immediately.
Q9 Plan and Manage an Azure AI Solution (25–30%)
You are evaluating a Microsoft Foundry RAG assistant that uses Azure AI Search for knowledge retrieval to answer HR policy questions. Groundedness scores are low, but trace logs show answers are consistent with retrieved chunks. For failing questions, the relevant policy pages exist in the source files but are not in the top retrieval results. What should you prioritize?
  1. Run content safety evaluations on failing answers
  2. Increase system prompt verbosity and retest for fluency
  3. Compare answer length against model token limits
  4. Measure retrieval recall@k against the marked relevant policy pages
✓ Correct Answer: D
Trace evidence points to a retrieval failure, not a generation failure. If the model is answering from retrieved context but the correct source passages are missing, measuring retrieval recall against marked relevant content is the key verification.

You've viewed 3 of 88 questions. Start the free practice exam to answer all questions with instant feedback.

Exam overview

Microsoft's AI-103 exam replaces the older AI-102 track as the new baseline for Azure AI Engineers. The exam code AI-103 corresponds to the full title "Developing AI Apps and Agents on Azure," an associate-level assessment that covers generative AI agent development, RAG pipeline architecture, and Responsible AI implementation using Azure AI services. Unlike AI-102 which focused on pre-built cognitive services, AI-103 shifts emphasis to building custom AI agents using Microsoft Foundry Python SDK, constructing RAG pipelines with vectorization and semantic re-ranking, and applying content filters via Azure AI Content Safety — the skills required for modern AI app development on Azure.

You need solid Python coding skills and at least 1–2 years of practical experience implementing software systems, consuming REST APIs/SDKs, and configuring Azure services before the material here will make sense. The official preparation course is AI-103T00-A (4-day intermediate training), and hands-on work with Microsoft Foundry projects, Prompt Flow monitoring, and Azure AI Search vector index configuration is strongly recommended before sitting the exam. Because this is an associate-level exam, Microsoft Learn is available during the test via a built-in split screen, but with only about two minutes per question, you cannot rely on documentation for architectural decisions.

Five domains shape the AI-103 blueprint, with the heaviest weight on implementing generative AI and agentic solutions (30–35%). The remaining domains — planning and managing an Azure AI solution (25–30%), implementing computer vision solutions (10–15%), implementing text analysis solutions (10–15%), and implementing information extraction solutions (10–15%) — complete the assessment. A distinctive feature of this exam is the code-heavy question format: you will see dropdown Python SDK code completion items, JSON schema matching questions for RAG pipeline configuration, and multi-part scenario questions that require you to choose between Azure AI Document Intelligence and Azure AI Content Understanding for a given extraction workload.

The practice questions here mirror the five-domain distribution exactly, with code-heavy scenario items that reproduce the dropdown Python SDK code completion and JSON schema matching formats you will face at the test center. Each question includes a detailed explanation that walks through the code logic — why a specific Foundry Python SDK method signature is required, how to configure vectorization and semantic re-ranking in an Azure AI Search RAG pipeline, and when to use Content Understanding versus Document Intelligence for multimodal document parsing. The downloadable PDF packages the same question bank for offline review, so you can practice Foundry Python SDK code patterns, Prompt Flow monitoring configuration steps, and Responsible AI content filter setup procedures during commutes or in environments without reliable internet.

Official Exam Domains & Weighting

To successfully pass the AI-103 exam, candidates must master the following core domains:
  • Domain 1: Plan and manage an Azure AI solution — 25–30%
Covers configuring RBAC and secure endpoints, managing API keys and SAS tokens, implementing Responsible AI practices using Azure AI Content Safety (prompt shields, jailbreak/injection blocking, content filters), creating and sharing Azure AI resources between Microsoft Foundry projects, and monitoring and evaluating AI pipeline performance using Prompt Flow.
  • Domain 2: Implement generative AI and agentic solutions — 30–35%
Focuses on developing AI Agents using Microsoft Foundry Python SDK, building Retrieval-Augmented Generation (RAG) pipelines over Azure AI Search, managing conversation states, threads, messages, and persistent agent runs, and applying integrated vectorization, hybrid search, and semantic re-ranking. The exam tests your ability to configure agent memory, manage conversation thread lifecycle, and optimize RAG pipeline retrieval accuracy with semantic re-ranking.
  • Domain 3: Implement computer vision solutions — 10–15%
Tests applying multimodal models to analyze image, video, and visual context, building solutions involving object detection, image captioning, and visual question answering, and addressing inpainting and image-fidelity requirements within Azure AI Vision. You must know how to call the multimodal model API with both image URL and base64-encoded image inputs.
  • Domain 4: Implement text analysis solutions — 10–15%
Covers processing natural language processing (NLP) workloads utilizing Large Language Models (LLMs), building solutions leveraging Azure AI Language capabilities for translation, sentiment analysis, and keyword classification, and configuring tokenization, customized parsing, and analyzer settings. The exam includes questions on setting up custom text analysis models and configuring Language studio projects.
  • Domain 5: Implement information extraction solutions — 10–15%
Focuses on discerning when to use Azure AI Document Intelligence versus Azure AI Content Understanding, extracting key-value pairs, tables, and structured text from forms, invoices, and contracts, and building schemas in natural language for multimodal unstructured document parsing. Document Intelligence is evaluated for deterministic, tabular layouts; Content Understanding is utilized for complex multimodal sources containing text, images, and audio/video files.

What Our Customers Say 110 verified reviews

4.8 ★★★★★★ Based on 110 reviews
★★★★★★
Took the AI-103 exam today and passed with 87%. Used this as my main prep material for about a month.
— Grace L.
★★★★★★
Good investment for the AI-103 exam. My only regret is not buying it sooner — would have saved me a lot of study time.
— Henry C.
★★★★★★
I was really impressed by the depth of the AI-103 answer explanations. Feels like having a tutor walk you through each question.
— Jack W.
★★★★★★
Good range of difficulty levels in the AI-103 bank. Easy questions build confidence, hard ones prepare you for the real thing.
— Avery K.
★★★★★★
I recommend this to everyone preparing for AI-103. The lifetime access is great — I keep coming back for reference.
— Robert C.
★★★★★★
Good bang for your buck on the AI-103 practice exam. Unlimited attempts means you can really drill the material.
— Zoey L.

Log in to rate this exam and leave a review.

Submitted for moderation before publishing. Keep it helpful and respectful.

Frequently Asked Questions

Yes. Because AI-103 is an Associate-level engineering exam, a built-in split screen button allows you to open and search the live Microsoft Learn documentation during the test. No external web searching or personal bookmarks are allowed. Our practice questions reproduce the same split-screen experience by providing detailed answer explanations with references to official Azure AI Foundry Python SDK documentation, Azure AI Search REST API specifications, and Azure AI Content Safety configuration guides — the same documentation you can access during the real exam.

You must understand when to use Azure AI Document Intelligence versus Azure AI Content Understanding. Document Intelligence is evaluated for deterministic, tabular layouts (invoices, receipts, forms). Content Understanding is utilized for complex multimodal sources containing text, images, and audio/video files. Our practice questions reproduce these architectural decision scenarios with detailed answer explanations that reference the exact service selection criteria — for example, when a form has fixed key-value pairs (use Document Intelligence) versus when a contract contains free-form text with images and tables (use Content Understanding).

Upon passing Exam AI-103, you achieve the Azure AI Apps and Agents Developer Associate credential. It remains active for exactly one year. You can renew it entirely for free during the 180 days leading up to its expiration by completing an unproctored assessment on Microsoft Learn. The renewal assessment covers only the features added in the past year, so it is significantly shorter than the full exam. Our practice materials include a dedicated FAQ section explaining the renewal process, and the downloadable PDF includes a timeline checklist to help you track your certification expiration date.

If you do not pass, a mandatory 24-hour cooling-off period is enforced before you can sit for a second attempt. Any subsequent attempts (up to a maximum of 5 tries within a rolling 12-month window) require a 14-day waiting period between tests. Each retake requires a new exam fee ($165 USD) unless you purchased an exam voucher bundle with a retake included. Our mock exam reproduces the exact question formats you will face — including case studies with locked navigation and multi-part scenario questions — so you can identify knowledge gaps before using an attempt.

The mock exam includes dropdown Python SDK code completion items where you must select the correct method signature for creating an AI agent, adding a data source to a RAG pipeline, and configuring semantic re-ranking parameters. Each practice question includes a detailed explanation that walks through the code logic — why a specific Foundry Python SDK method is required, how to configure vectorization and semantic re-ranking in an Azure AI Search RAG pipeline, and when to use hybrid search versus pure vector search for a given retrieval accuracy requirement.

Yes, the complete question bank is available as a downloadable PDF that packages all five domains in the same weighted distribution as the real exam. The PDF includes detailed answer explanations with references to official Azure AI Foundry Python SDK documentation, Azure AI Search REST API specifications, and Azure AI Content Safety configuration guides. The PDF also contains code samples for Foundry Python SDK patterns, Prompt Flow monitoring setup, and Responsible AI content filter configuration — all formatted for offline review during commutes or in environments without reliable internet access.

Set up a Microsoft Foundry project and deploy at least one AI agent using the Foundry Python SDK. Build a RAG pipeline over Azure AI Search with vectorization, hybrid search, and semantic re-ranking enabled, and test it with sample documents. Configure Azure AI Content Safety content filters with prompt shields and jailbreak/injection blocking, then test the filters with adversarial prompts. When you can write Foundry Python SDK code and configure RAG pipeline parameters without referencing documentation, the AI-103 scenario questions become straightforward applications of your hands-on experience.