AI-200 Prerequisites: What to Know First

The short version

The Microsoft pages we checked list no required certification and no required number of years before AI-200. They do describe the skills a candidate should have, and Python is named explicitly. If you can write simple Python, deploy a container, and use a database from code, you can start. If those are new, add preparation time first.

What Microsoft says

Microsoft describes the candidate as someone who contributes to all phases of implementing AI solutions on Azure, with an emphasis on back-end services and components, and who supports the development lifecycle: requirements, design, development, deployment, security, and monitoring. The skills you should be proficient in are:

  • Azure SDKs and third-party SDKs used in Azure
  • Azure data management services
  • Azure monitoring and troubleshooting
  • Azure messaging and eventing
  • Vector databases
  • Python programming
  • Implementing containerized applications on Azure

Microsoft does not turn this list into a formal requirement, so you can book the exam without any certification. The list tells you what the questions will assume.

Required versus recommended

Required: nothing that Microsoft lists. You do need a way to schedule the exam; Microsoft recommends a personal Microsoft account. Check the ID and check-in rules for the proctored session when you book.

Recommended, based on the skills list:

  • Basic Python: functions, classes, dictionaries, reading a stack trace, and installing packages.
  • A working idea of REST APIs and what an SDK does for you.
  • Some hands-on time in the Azure portal and with a command-line tool.
  • Familiarity with containers: what an image is, what a registry does, and how a container differs from a virtual machine.
  • A general sense of what an embedding is and why you would search by similarity.

Suggested earlier certifications

Microsoft’s AI-200 pages do not name a preceding certification. If you want a foundation, general Azure knowledge helps, but no earlier certification is required for AI-200. If you hold the older Azure Developer Associate credential (AZ-204, now marked as retired), AI-200 is a separate exam with a separate skills list, so do not assume your earlier knowledge covers it.

Skills self-check

Tick what you can do today, without looking anything up:

  • I can explain the difference between a container image and a running container.
  • I can say what a container registry stores and why tags matter.
  • I can describe what a Kubernetes Deployment and a Service do.
  • I can explain scale to zero and what an event-driven scaler watches.
  • I know what a request unit is in Cosmos DB.
  • I can explain what an index does and what it costs on writes.
  • I can explain what an embedding is and what “nearest neighbor” means.
  • I can explain the difference between exact and approximate search.
  • I know why a cache entry needs an expiry or an invalidation rule.
  • I can say when I would use a queue and when I would use an event.
  • I know what a dead-letter queue is for.
  • I can write a small Python function that calls a service through an SDK.
  • I know why an app should use a managed identity instead of a stored password.
  • I can read a simple KQL query and say what it returns.

If you can honestly tick ten or more, start with a quick pass through the study plan and take a baseline test early. If you tick fewer than six, choose the 8-week schedule and begin with the basics below.

What to learn first if you are new to Azure development

  1. Learn Python well enough to read a short script and call an SDK.
  2. Learn what Azure resources, resource groups, and subscriptions are, and set a spending alert.
  3. Learn containers: build one image, run it, and push it to a registry.
  4. Learn one database well. Use it from Python, then add a vector column or index.
  5. Learn queues and events with one small example each.
  6. Learn how identity replaces passwords: managed identity, role assignment, and Key Vault.
  7. Then follow the four areas in the domains explained page.

You do not need to train a machine learning model.

Accessibility and language

If you need accommodations, such as assistive devices or extra time, Microsoft asks you to request them before you register. If the exam is not offered in your language, Microsoft says you can request an additional 30 minutes. The exam is offered in 13 languages.

More guides for AI-200

Next step

Read the exam overview for the format at a glance, then try the AI-200 practice tests to see where your baseline sits.

Frequently asked questions

Do I need AZ-204 or another Azure certification first?

No. The Microsoft pages we checked name no required certification for AI-200.

Do I need to know machine learning?

Not in the sense of training models. You should understand embeddings and similarity search, because vector search appears across the data management domain.

Do I need an Azure account?

It is not required to book the exam, but you will study much faster with one. Set a budget alert first, because some of these services cost money while they run.

Try the AI-200 practice exams