Update (September 2026): AI-900 and AI-102 retired on June 30, 2026. Their replacements, AI-901 (Azure AI Fundamentals) and AI-103 (Azure AI App and Agent Developer Associate), are the current exams. For the current path, see the AI-103 Study Guide.
AI-102 and AI-900 both retired on June 30, 2026. If you were preparing for either, or you are choosing a Microsoft AI certification now, this explains what changed and which exam to take today.
If you've been Googling "Microsoft AI certification" lately, you've probably ended up more confused than when you started: two retired exam codes, their replacements, and Microsoft's documentation spread across a dozen pages.
This post cuts through all of it. By the end you'll know exactly which cert to take, in what order, and - critically - whether you should rush to finish an old exam or wait for the new ones.
What Just Happened (The Context You Need)
On March 24, 2026, Microsoft announced one of the biggest certification overhauls in years. The short version:
- AI-900 (Azure AI Fundamentals) → retired June 30, 2026, replaced by AI-901
- AI-102 (Azure AI Engineer Associate) → retired June 30, 2026, replaced by AI-103
- AI-103 (Azure AI App and Agent Developer Associate) → the current associate exam, after a beta that opened in April 2026
This is not a minor update. Microsoft is repositioning its entire AI certification stack around generative AI, agentic systems, and Microsoft Foundry - the platform where Azure OpenAI, Azure AI Services, and agent orchestration all come together. New to MCP? Read our explainer: What is MCP? The 2026 Developer Guide.
The Three Certs, Plainly Explained
AI-900: Azure AI Fundamentals
- Level: Beginner
- Coding required: No
- Status: Retired June 30, 2026, replaced by AI-901
AI-900 is the entry point. It tests your understanding of AI concepts, not your ability to build anything. You'll learn what machine learning is, how cognitive services work, what responsible AI means, and how to identify which AI tool solves which business problem.
Think of it as the AI equivalent of getting your driving theory certificate before you touch a steering wheel. It proves you understand the landscape. Employers in non-technical roles - project managers, business analysts, consultants - use it as a credibility signal when working on AI initiatives.
Who it's for: Non-developers, people new to AI entirely, anyone who needs to talk intelligently about AI without building it.
Who it's NOT for: Developers. If you're writing code, skip straight to AI-103.
Time to prepare: 2-4 weeks of part-time study. No hands-on lab environment needed.
AI-102: Azure AI Engineer Associate
- Level: Intermediate
- Coding required: Yes (Python or C#)
- Status: Retired June 30, 2026, replaced by AI-103
AI-102 is the current workhorse cert for developers building AI solutions. It covers Azure OpenAI Service, Azure AI Vision, Azure AI Language, Document Intelligence, Bot Service, and more. You need hands-on experience provisioning services, calling APIs, implementing content filters, and building end-to-end solutions.
It's a solid cert - but it was designed for a world where you were wiring together individual AI services. The new exam (AI-103) is designed for a world where you're building autonomous agents and multi-step reasoning workflows.
Who it's for now: Nobody new, because the exam can no longer be taken. If you already hold AI-102, it still validates real skills and stays on your transcript.
What to do instead: Start with AI-103, which covers the same ground plus agents, retrieval and evaluation on Azure AI Foundry.
Time to prepare: 6-10 weeks with hands-on lab practice.
AI-103: Azure AI App and Agent Developer Associate
- Level: Intermediate
- Coding required: Yes
- Status: Live, the current associate-level exam (beta opened April 2026)
AI-103 is the replacement for AI-102, built ground-up around where AI development is actually heading. The focus shifts from "configure this service" to "build and deploy production-ready AI applications and agents."
The exam covers Microsoft Foundry (Microsoft's unified AI development platform), generative AI application development, multi-step reasoning workflows, multi-agent orchestration, responsible AI in production, and deploying at scale.
This is the cert that hiring managers will be looking for from mid-2026 onwards. If you're a developer building anything with AI, this is where you should be heading.
Who it's for: Developers building AI apps and agents. Anyone who already has AI-102 and wants to stay current. Anyone starting from scratch who has 3+ months to prepare.
Time to prepare: 8-12 weeks. Hands-on practice with Microsoft Foundry and Azure OpenAI is essential.
Try a real Azure AI lab, free
A pre-provisioned Azure environment in your browser. No Azure account, no credit card, no setup.
Most large enterprises already run on Microsoft infrastructure. Azure AI Foundry is the platform those teams use to build and deploy AI apps, so the skills you practice here map directly to real job demand.
The Decision Matrix: Which One Should YOU Take?
"I'm completely new to AI and I'm not a developer"
Take AI-901.
AI-900 is no longer available. AI-901 replaced it at the same fundamentals level, and it adds generative AI, Azure OpenAI and Microsoft Foundry, which is what most beginners want to learn anyway.
"I'm a developer who started studying for AI-102 and I'm halfway through"
Move to AI-103.
The skills you're learning for AI-102 are not wasted. AI Vision, AI Language, Azure OpenAI, Document Intelligence - all of that knowledge carries forward into AI-103. Taking AI-102 gives you a certified foundation and the hands-on experience that makes AI-103 prep significantly faster.
AI-102 can no longer be booked, but little of your preparation is wasted: roughly half of it carries over. Our AI-102 to AI-103 guide maps it objective by objective.
"I'm a developer starting from scratch today"
Skip AI-102. Start AI-103 prep now.
AI-102 is no longer available, and AI-103 is where the industry is going. Invest your time in the cert that will still be valid and valued in 2027 and beyond.
"I already have AI-102 - what do I do?"
Your cert remains valid until it expires on your transcript. You don't lose it.
But it can no longer be renewed. When it expires, AI-103 is how you stay current, so start AI-103 prep before that date rather than after it.
"Should I do AI-900 before AI-103?"
Only if you're a complete beginner with no cloud or coding experience.
For developers, AI-900 is optional at best. The fundamentals knowledge it covers will come naturally as you prepare for AI-103. The 2-4 weeks you'd spend on AI-900 is almost always better spent starting AI-103 prep earlier.
Still here? You clearly care, grab the exam study guide
One-page quick reference plus the hands-on labs that go with it.
The Honest Answer to "Which One First"
For the majority of people reading this, the answer is:
- If you're a developer: AI-103.
- If you're non-technical: AI-901.
- If you were mid-prep on AI-102: move to AI-103 and keep what carries over.
The certification landscape is changing fast. Microsoft's 2026 overhaul is the clearest signal yet that the industry has moved on from "configure individual AI services" to "build and deploy AI-powered applications and agents end-to-end." The exams are following that shift.
The One Thing That Actually Determines If You Pass
Every certification forum has the same post: "I watched all the videos and read the docs but still failed." The reason is almost always the same - no hands-on practice.
AI-102 and AI-103 are not multiple choice recall tests. They include scenario-based questions that test whether you've actually built these solutions, not just read about them. Can you identify why a deployed model is returning inappropriate responses? Can you debug an agent that's failing mid-workflow? Can you select the right AI service for a given business constraint?
You can't answer those questions from documentation alone. You need to have actually done it.
That's exactly why we built GetHandsOn.ai - browser-based AI labs in cloud that mirror the real exam scenarios. No cloud subscription needed, no setup, no teardown. You launch a lab, complete the task, and the environment handles the rest.
We have a free lab for AI-103 available right now - no payment required. If you're starting your AI-103 journey, it's the fastest way to get a feel for what the exam actually demands.
Quick Reference: AI-900 vs AI-102 vs AI-103
| AI-900 | AI-102 | AI-103 | |
|---|---|---|---|
| Level | Beginner | Intermediate | Intermediate |
| Coding needed | No | Yes | Yes |
| Status | Retired June 30, 2026 | Retired June 30, 2026 | Live |
| Replacement | AI-901 | AI-103 | - |
| Focus | AI concepts & literacy | AI services | AI apps & agents |
| Hands-on lab practice | Optional | Essential | Essential |
| Time to prepare | 2-4 weeks | 6-10 weeks | 8-12 weeks |
Bottom Line
The June 30 deadline has passed. If you were halfway through AI-102, that preparation still counts toward AI-103. And if you're starting fresh, AI-103 is the cert that will still matter when you're renewing it two years from now.
The fundamentals don't change: hands-on practice beats passive study every time. Whatever cert you're targeting, build things, break things, and fix them.
GetHandsOn.ai provides browser-based hands-on AI labs in cloud for Microsoft AI certifications. No cloud account required. Start with a free lab today.
Related: What Is an AI Engineer in 2026? Role, Skills & Path
The 2026 position, in one paragraph
Two of the three exams compared above are gone. AI-900 and AI-102 both retired on June 30, 2026, which means the real choice today is between AI-901 at fundamentals level and AI-103 at associate level. AI-901 assumes no Azure background and is the sensible entry point if generative AI is new to you. AI-103 assumes you can already write and deploy code against cloud APIs, and it tests agents, tool calling, evaluation and retrieval design rather than service trivia. Almost nobody needs both: pick the level that matches how much application development you already do. If you hold a retired AI-900 or AI-102 certification, it stays valid until its expiry date and neither replacement resets that.