AI-103 is the exam for Microsoft Certified: Azure AI Apps and Agents Developer Associate, the certification that replaced Azure AI Engineer Associate after AI-102 retired on June 30, 2026. It is an intermediate exam for developers who build generative AI applications and agents with Python and Microsoft Foundry. You have 120 minutes, and you need a score of 700 or higher to pass.

This guide explains what changed from AI-102, what the exam covers, who it suits, and how to prepare.

Looking for AI-102? AI-102 retired on June 30, 2026, and you can no longer take it. AI-103 is now the exam for Microsoft’s Azure AI developer certification. Our guide to Microsoft certifications retiring in 2026 covers the other exams that changed this year.

What Is the AI-103 Exam?

The full title is Exam AI-103: Developing AI Apps and Agents on Azure. Passing it earns the Microsoft Certified: Azure AI Apps and Agents Developer Associate certification. Microsoft classifies it as an intermediate certification for AI engineers and developers.

The role it describes is an Azure AI engineer who builds, manages, and deploys AI solutions and agents on Microsoft Foundry. That means choosing the right models and services, building retrieval augmented generation (RAG) apps and agents that call tools, monitoring and securing them in production, and applying responsible AI controls throughout. The role also covers vision, speech, text analysis, and document extraction, increasingly as capabilities that agents use rather than as standalone projects.

Exam detail AI-103
Level Intermediate (Associate)
Exam time 120 minutes
Passing score 700 or higher
Programming language Python
Prerequisite certifications None
Renewal Yearly, free online

AI-103 vs. AI-102: What Changed?

AI-103 keeps the shape of AI-102: planning and securing Azure AI solutions, generative AI, computer vision, language, and information extraction. The biggest change is that agents moved from a small part of the exam to its center. In AI-102, building agents was 5% to 10% of the exam. In AI-103, generative AI and agentic solutions together make up 30% to 35%, the largest skill area, and include multiagent orchestration, conversation memory, and approval controls.

Area Change AI-103 compared with AI-102
Certification Renamed Azure AI Engineer Associate is now Azure AI Apps and Agents Developer Associate.
Exam Replaced AI-102 retired on June 30, 2026. AI-103 is the current exam.
Generative AI and agents Expanded Generative AI (15% to 20%) and agents (5% to 10%) are now one area weighted 30% to 35%, covering RAG, tool calling, memory, multiagent solutions, and safeguards.
Planning and operations Expanded Up from 20% to 25% to 25% to 30%, with more on CI/CD for Foundry projects, cost and quota management, monitoring grounding quality, and governing agent behavior.
Computer vision Refocused Same weight, but now centered on image and video generation and multimodal understanding. Custom vision models are no longer listed.
Language and extraction Reduced Language (15% to 20%) and knowledge mining (15% to 20%) become text analysis and information extraction at 10% to 15% each, built around language models and Content Understanding.
Programming language Narrowed AI-102 named Python and C#. AI-103 expects Python.

If you already hold Azure AI Engineer Associate, it stays on your Microsoft Learn transcript, but Microsoft does not renew certifications after they retire. There is no automatic transition to the new credential. To hold Azure AI Apps and Agents Developer Associate, you need to pass AI-103.

Who Should Take AI-103?

AI-103 suits developers who build AI features into applications, engineers moving from traditional Azure development into generative AI, and former AI-102 candidates. Microsoft expects experience developing apps with Python and familiarity with general AI, generative AI, and Azure services.

A quick readiness check: could you build a Python app that answers questions from your own documents and cites where each answer came from? Could you explain how you would stop an agent from calling a tool it should not use, or from acting without a person’s approval? If those questions feel unfamiliar, spend time on Python and Azure development first. If AI concepts themselves are new to you, Azure AI Fundamentals (AI-901) is an optional starting point, though it is not required.

AI-103 is also the prerequisite for Multi-Agent AI Solutions Expert (AI-500), Microsoft’s expert certification for designing multiagent systems. If your work is the infrastructure that AI apps run on, such as containers, data services, and messaging, the Azure AI Cloud Developer (AI-200) certification may fit better. And if your job is securing AI workloads rather than building them, see our SC-500 exam guide.

To compare AI-103 with the other AI credentials, see our list of the best Microsoft AI certifications, or see where it sits among the Azure roles in our Azure certification path.

What Does the AI-103 Exam Cover?

AI-103 measures five skill areas. The percentages show how much of the exam each area represents.

Skill area Exam weighting
Plan and manage an Azure AI solution 25% to 30%
Implement generative AI and agentic solutions 30% to 35%
Implement computer vision solutions 10% to 15%
Implement text analysis solutions 10% to 15%
Implement information extraction solutions 10% to 15%

Plan and Manage an Azure AI Solution

This area covers choosing models (large and small language models, multimodal models, and Foundry Tools) and the right Foundry services for grounding, vector search, memory, and agent workflows. It also covers designing infrastructure and deployments, connecting Foundry projects to CI/CD pipelines, managing quotas, rate limits and cost, and monitoring performance, drift, safety events, and search index health. Security means managed identities, private networking, keyless credentials, and role policies. Responsible AI covers safety filters and guardrails, evaluations, trace logging and approval workflows, and controls on what tools an agent can use.

Implement Generative AI and Agentic Solutions

The most heavily weighted area. On the generative side, expect deploying and consuming models, implementing RAG, building multistep and tool augmented workflows, evaluating apps for fabrications, relevance, quality, and safety, and connecting applications to Foundry projects with the SDKs. On the agent side, it covers defining roles, goals, and tool schemas, building agents with retrieval, function calling, and memory, orchestrating multiagent solutions, adding safeguards and approval flows to autonomous workflows, and monitoring agent behavior. It also covers prompt engineering, model parameters, self critique loops, and observability through tracing, token analytics, and latency breakdowns.

Implement Computer Vision Solutions

This area covers generating and editing images and videos from prompts and reference media, including inpainting and mask based edits. It also covers multimodal understanding: captions, alt text, question answering grounded in images, video analysis, and object detection with Azure Content Understanding. Responsible AI here includes filtering unsafe visual content and detecting prompt injection hidden in text inside images.

Implement Text Analysis Solutions

This area covers extracting entities, topics, summaries, and structured JSON with language models and Foundry Tools, detecting sentiment and sensitive content, and translating text with Azure Translator or language model flows. It also covers speech: speech to text and text to speech for agents, custom speech models, reasoning over audio, and speech translation.

Implement Information Extraction Solutions

This area covers building retrieval pipelines that ingest and index documents, images, audio, and video, configuring semantic, hybrid, and vector search for grounding, and adding enrichment and OCR to RAG ingestion. It also covers extracting content from documents with Content Understanding to produce clean, structured output for agents and RAG.

Treat AI-102 study material with care. Much of the Foundry, vision, language, and search content still applies, but it gives agents only a small share of attention, and it will not cover multiagent orchestration, agent governance, or image and video generation.

Is AI-103 Worth It?

AI-103 is worth it if you build AI applications on Azure, or you want to. It validates the work most teams are doing now: grounding models in company data, building agents that call tools and APIs, and running them safely in production. It is also the current Microsoft certification for the role AI-102 used to cover, and the required step toward the Multi-Agent AI Solutions Expert certification.

It is a weaker fit if you do not write code. AI-103 is a developer exam, and Microsoft expects working Python. If you administer Copilot and AI services in Microsoft 365, or use AI tools without building them, a different certification will match your work better. And as with any certification, it carries the most weight alongside projects you can describe in an interview.

How to Prepare for the AI-103 Exam

1

Map the Objectives Against Your Experience

Go through the five skill areas and mark each topic as something you build regularly, something you have done once, or something you have only read about. Developers who built chat apps with an AI-102 background often know RAG and model deployment well but have little experience with multiagent orchestration, evaluations, or agent governance. Developers new to Azure may need more time on identity, networking, and deployment.

2

Build Small Projects in Foundry

Use a test subscription and a Foundry project you can delete. Write the code in Python, and give each project a way to check that it works:

  • RAG: Index a small set of your own documents, build an app that answers questions from them, and check that answers cite the right source.
  • Agent: Build an agent with two tools, such as a search tool and a custom function, then add memory so it can follow up on an earlier question.
  • Safeguards: Add an approval step before the agent takes an action, and restrict which tools it may call.
  • Multiagent: Split a task between two agents, such as one that researches and one that writes, and trace how they hand off work.
  • Evaluation: Run evaluations for relevance, fabrication, and safety, change a prompt or model parameter, and compare the results.
  • Extraction: Use Content Understanding to pull fields from a scanned document and feed the result into your RAG index.

Model deployments, search indexes, and some tools are paid services, so check what each costs and delete lab resources when you finish.

3

Give Agents and Operations Extra Time

Agents, evaluation, and monitoring are the newest part of the exam and the least likely to appear in older material. Learn what each control does: guardrails and content filters, evaluators, tracing, approval flows, and tool access limits. Then practice reading traces and evaluation results, because scenario questions often ask what to change when an agent gives a poor or unsafe answer.

4

Test Yourself Under Exam Conditions

Take timed practice exams once you have covered every area. For each question you miss, note which skill area it belongs to and go back to your projects for that topic. Microsoft’s exam sandbox shows the question formats and interface before exam day.

Think like the scenarioAI-103 questions describe an application and ask for the best service, setting, or fix. Before choosing, identify what the solution must do, what data it is grounded in, and which answer meets the requirement with the least cost, access, or risk.

Prepare with Our AI-103 Bootcamp

Our AI-103 Azure AI Apps and Agents Developer bootcamp covers the full exam in four days of live training. Join us in the classroom or attend remotely with a live instructor. You take the AI-103 exam during the bootcamp.

Our bootcamp includes:

  • Live instruction from a Microsoft Certified Trainer.
  • The official AI-103 courseware and hands on labs, with 90 days of lab access.
  • An AI-103 exam voucher.
  • One retake voucher if needed.
  • Practice exams.

Our AI-103 bootcamp costs $2,995 for classroom live or remote classroom live, or $4,095 all inclusive with airfare and lodging. Every option includes the exam voucher.

If you are new to AI, our AI-901 and AI-103 bootcamp adds Azure AI Fundamentals in a five day program. It costs $3,495 for classroom live or remote classroom live, or $4,845 all inclusive, and includes both exam vouchers and one retake voucher per exam.

If you want to go straight on to the expert level, our AI-103 and AI-500 bootcamp covers both exams in six days. It costs $4,995 for classroom live or remote classroom live, or $6,345 all inclusive, and includes both exam vouchers and one retake voucher per exam.

The bootcamp length refers to the training. Earning the certification requires passing the exam, and your readiness depends on your experience and preparation.

AI-103 Exam FAQs

What Replaced AI-102?

AI-103 replaced AI-102, which retired on June 30, 2026. The certification changed from Azure AI Engineer Associate to Azure AI Apps and Agents Developer Associate.

Is AI-103 Hard?

AI-103 is an intermediate exam with a broad scope: five skill areas spanning models, agents, vision, speech, language, search, and document extraction. Candidates who already build applications in Python have a strong advantage. If you have not written code against an AI service before, build that experience before you start.

Do I Need AI-901 Before AI-103?

No. AI-103 has no prerequisite certifications. Azure AI Fundamentals (AI-901) is an optional introduction to AI concepts, useful if generative AI is new to you.

Can I Use C# for AI-103?

Microsoft’s AI-103 candidate profile asks for experience developing apps with Python and does not mention C#, which AI-102 did. Plan to read and write Python for the exam.

Can I Still Renew My AI-102 Certification?

No. Microsoft does not renew certifications after they retire. Your Azure AI Engineer Associate certification stays on your transcript, and to earn the current credential you need to pass AI-103.

Does AI-103 Count Toward AI-500?

Yes. Azure AI Apps and Agents Developer Associate is the prerequisite for Multi-Agent AI Solutions Expert. You earn the expert certification by holding the associate certification and passing AI-500.

How Much Does AI-103 Training Cost?

Our four day AI-103 bootcamp costs $2,995 for classroom live or remote classroom live, including the exam voucher and a retake voucher. The AI-901 and AI-103 bootcamp costs $3,495 and includes both exam vouchers.

Does the Certification Expire?

Yes. Like other associate certifications, it expires after one year unless you renew it. Renewal is free through an online assessment on Microsoft Learn, available in the six months before expiry.

What’s Next

Compare the five AI-103 skill areas with the AI work you already do, then plan projects around the gaps, starting with agents, evaluation, and monitoring. If you want a structured week with an instructor, labs, and the exam voucher included, explore our four day AI-103 bootcamp.