AI-103 courseLesson 1 of 18 · 9 min
AI-103 Exam Guide: Azure AI Apps and Agents Developer (2026)
XUtilHub Team ·
Exam skills covered
- Exam overview and study plan
AI-103: Developing AI Apps and Agents on Azure is the exam for the Microsoft Certified: Azure AI Apps and Agents Developer Associate certification. It's for engineers who build, deploy, secure and operate generative AI apps and agents on Microsoft Foundry, and it replaces the older AI-102 exam. This lesson explains what's on it, how it's weighted, and how to study.
Who the exam is for
The audience profile describes an Azure AI engineer who builds, manages and deploys agents and AI solutions with Microsoft Foundry. You're expected to:
- Have experience developing apps in Python
- Know the capabilities of general AI, generative AI and Azure services
- Work with architects, data scientists, DevOps and security engineers
If you're new to Azure AI, start with our free AI-901 Azure AI Fundamentals course. AI-103 assumes you already know those concepts and asks you to build and operate real solutions.
The five skill areas
| Skill area | Weight |
|---|---|
| Plan and manage an Azure AI solution | 25–30% |
| Implement generative AI and agentic solutions | 30–35% |
| Implement computer vision solutions | 10–15% |
| Implement text analysis solutions | 10–15% |
| Implement information extraction solutions | 10–15% |
More than half the exam is about planning, operating and building generative AI and agents. The other three areas are smaller, but they still decide whether you pass.
What changed from AI-102
AI-102 was built around individual services: Vision, Language, Speech, Document Intelligence, Search, each with its own SDK. AI-103 is built around Foundry:
- Agents are central: tools, memory, multi-agent orchestration, approvals and safeguards
- Operating AI gets a whole domain: quotas, cost, monitoring, tracing, private networking and keyless auth
- Evaluation is tested explicitly: groundedness, relevance, quality, safety and agent evaluators
- Vision now includes image and video generation, and multimodal models do much of the analysis
- Information extraction centres on Azure AI Search pipelines and Content Understanding
Classic services still appear as Foundry Tools (Language, Translator, Speech, Content Understanding), but often as one choice alongside "prompt a language model".
Exam facts
- Passing score: 700 out of 1000 (scaled)
- Question types: multiple choice, multiple select, drag-and-drop ordering, case studies and code completion. Expect Python snippets and configuration you have to read.
- Preview features: most questions cover GA features, but commonly used preview features can appear
- Renewal: associate certifications expire every year; renew with a free online assessment on Microsoft Learn
- Language: English is updated first; if your language isn't offered you can request 30 extra minutes
How this course is organised
The 18 lessons follow the official skills outline:
- Plan and manage: choosing models and services, setting up deployments and CI/CD, quotas, monitoring and security, and responsible AI guardrails
- Generative AI and agents: the Foundry SDK, RAG, evaluation, agents with tools and memory, multi-agent workflows, tuning and observability
- Computer vision: image and video generation, multimodal understanding and visual safety
- Text analysis: LLM and Azure Language text analysis, translation, and speech for agents
- Information extraction: Azure AI Search pipelines and Content Understanding
- Cheat sheet: everything on one page
Every lesson ends with a short quiz, and the AI-103 practice test mixes all the questions with extra exam-style ones.
A 6-week study plan
| Week | Focus |
|---|---|
| 1 | Lessons 1–5: planning, deployments, security, responsible AI. Create a Foundry project and deploy two models. |
| 2 | Lessons 6–7: Foundry SDK and RAG. Build a small RAG app over your own PDFs. |
| 3 | Lessons 8–11: evaluation, agents, multi-agent workflows, tracing. Build an agent with a function tool. |
| 4 | Lessons 12–15: vision, text and speech. Try image generation, image Q&A and speech to text. |
| 5 | Lessons 16–17: Azure AI Search and Content Understanding. Index scanned documents with OCR. |
| 6 | Cheat sheet, practice tests, Microsoft's free practice assessment, and fixing weak areas. |
Hands-on time matters. Questions often describe a symptom ("the app gets 429 errors", "answers ignore the documents") and ask for the fix, which is much easier once you've seen it happen.
Key takeaways
- AI-103 = build, deploy and operate AI apps and agents on Microsoft Foundry, in Python.
- Weighting: generative AI and agents 30–35%, plan and manage 25–30%, and 10–15% each for vision, text and extraction.
- Pass mark 700; renew yearly with a free assessment.
- Study by building: deploy models, a RAG app, an agent with tools, and a search index.
Next lesson: choose the right models and Foundry services.
Check your understanding
Exam-style questions on this lesson. Get them all right to mark the lesson complete.
Finished reading?