Azure AI work is where the hiring is right now.
Companies spent the last two years running generative AI as small pilots. Now they're building it into real products, and that takes a different kind of developer: someone who can take an AI feature from a demo and turn it into something that runs safely, at scale, in production. That's a change in what the job asks of you, not just a new tool on the pile.
That change is probably why you're reading this. Either you already work in Azure and want to move into that AI-focused role, or you had AZ-204 on your plan and just found out it's gone. Both routes run through the same exam now.
Microsoft retired AZ-204, its general-purpose Azure Developer certification, on 31 July 2026. AI-200, Microsoft Certified: Azure AI Cloud Developer Associate is what replaced it, and there's no plan for a new general Azure developer exam to take over what AZ-204 used to be. AI-200 is the route now. That decision is worth sitting with for a second, because Microsoft replacing its main developer cert with an AI-focused one is a signal about where the Azure developer role itself is going. The demo-to-production gap is exactly where most teams are stuck: they've shown leadership a proof of concept, but nobody can take it further. AI-200 is built to certify the person who can.
This guide covers what the certification is, who it's for, how it compares to AI-102, what each of the four exam domains actually asks of you, and where developers coming from AZ-204 tend to have the hardest adjustment.
AI-200 at a glance
Before the detail, the facts you're probably scanning for:
- Certification: Microsoft Certified: Azure AI Cloud Developer Associate
- Exam: AI-200, Developing AI Cloud Solutions on Azure
- Length: 120 minutes, proctored, scheduled through Pearson VUE
- Level: intermediate
- Domains: four (weightings below)
- Prerequisites: none formal, but real hands-on Azure development experience is assumed
- Languages: English
- Replaces: AZ-204, retired 31 July 2026
- Fee: varies by country, so check the official AI-200 page before booking
Everything below expands on those points. If you just needed the facts, you have them.
What passing AI-200 actually proves
It proves you can build, secure, and run AI-powered applications on Azure end to end, rather than prototype something in a notebook and hand it to someone else to productionise. That's the distinction that matters to employers, because it's the exact gap most organisations are stuck in.
It's also not AZ-204 with a new number stuck on the front. The exam still covers Azure development fundamentals like containerisation and connecting to Azure services, but the weighting has shifted hard towards building and securing AI solutions. If you already hold AZ-204, it doesn't vanish either: it moves to Microsoft's historical certifications list once your renewal window closes, though you can't renew it going forward.
Who should take AI-200?
Developers who already work across the Azure lifecycle: requirements, design, build, deployment, security, and monitoring. Microsoft lists no formal prerequisite certification, but the skills the exam assumes (Azure SDKs, containerised applications, Python, Azure data services, messaging and eventing) point at someone with genuine hands-on Azure experience, not someone starting cold.
In practice, you're probably one of three people:
- An Azure developer whose AZ-204 plans just got interrupted, and who needs the direct replacement.
- A backend or cloud engineer being asked to build generative AI features into an existing production application.
- Someone responsible for securing, monitoring, and troubleshooting AI workloads in Azure, not just building them once and moving on.
If you were still working towards AZ-204 using our AZ-204 certification guide, this is where that effort goes now.
AI-200 or AI-102: which one fits you?
The difference is what you're building. AI-102 (Azure AI Engineer Associate) tests you on the AI itself: model selection, agents, and how you design retrieval. AI-200 tests you on wiring that AI into a production application, the plumbing rather than the model: the containerisation, the data layer, the messaging, and keeping the whole thing secure and running.
So the quick version:
- Choose AI-200 if your day job is building and running complete cloud applications, and the AI features are one part of a bigger developer role you already own.
- Choose AI-102 if your focus is the AI services themselves, and the surrounding infrastructure and data layer belong to someone else.
If AI-102 sounds closer to what you actually do, our guide to becoming a Microsoft Certified AI Engineer walks through that path in detail.
The four domains, and what each one asks of you
AI-200 groups its content into four domains. The percentages tell you where to put your study time, but the percentages alone don't tell you what to actually practise, so here's both. Weightings come from Microsoft's AI-200 study guide:
- Develop containerised solutions on Azure (20 to 25%). You should be comfortable packaging an application into a container and running it on Azure's container services: Container Registry, Container Apps, and AKS. If you came through AZ-204, this is likely your strongest ground already.
- Develop AI solutions using Azure data management services (25 to 30%). The heaviest domain, and the newest for most Azure developers. This is where vector search lives, across services like Cosmos DB, PostgreSQL with pgvector, and Azure Managed Redis, along with the retrieval patterns built on them. It's where reading a module and actually building something pull apart fastest.
- Connect to and consume Azure services (20 to 25%). Wiring services together: SDKs, messaging and eventing, calling one Azure service from another. Familiar territory for an experienced Azure developer, applied to an AI context.
- Secure, monitor, and troubleshoot Azure solutions (20 to 25%). Keeping the application safe and observable once it's live: identity and access, securing your AI workload, and diagnosing what's gone wrong when it misbehaves in production.
Two things stand out from that list. The containerisation and general connectivity domains cover ground an AZ-204-background developer usually knows well. What's genuinely new is the second domain, the AI data management one, and that's where your preparation time earns the most.
Where AZ-204 developers get caught out
The gap is almost always the data management domain. Vector search and Azure's AI-specific data services are new ground for developers whose Azure experience predates the AI-heavy part of the stack, and that domain is the single biggest slice of the exam. It's also the part that rewards people who've actually built something over people who've only read about it, because the concepts don't fully land until you've handled real data in a real service.
So if your Azure experience comes from the AZ-204 world, don't spread your revision evenly. Containerisation and service connectivity should feel familiar and need a lighter pass. The bigger adjustment is the vector-search and AI data layer, and that's where hands-on practice, not module-reading, closes the gap.
Preparing for the exam
Start with Microsoft's own AI-200 study guide, which maps every objective to a free Microsoft Learn module. It's a sound starting point, but reading modules and passing the exam are two different skills, especially in that AI data management domain.
Microsoft's exam sandbox lets you get used to the interactive question format before the day itself. A practice assessment wasn't available at the time of writing; Microsoft usually releases one around eight weeks after an exam goes out of beta, so check again nearer your exam date rather than assuming there isn't one.
If your Azure experience predates the AI-heavy part of this syllabus, which is true for a lot of developers moving over from AZ-204, structured teaching closes that gap faster than self-study alone. Our instructor-led Microsoft Azure AI Cloud Developer AI-200 Course is built around the same four domains, with the AI data management content given the extra classroom time it needs. And if AI-200 is one of several certifications your team needs this year rather than a one-off, Unlimited Microsoft Training gives your developers ongoing access across the Azure and Microsoft stack instead of booking each exam's preparation separately.
Put the preparation in properly and AI-200 is a real opening, not just another badge. The developers who can build and ship AI features are getting asked first. This is how you prove you're one of them.
