Do not start here without Azure basics.
When a candidate with very basic cloud knowledge asked where to begin, an experienced commenter pointed them to AI-901 and AZ-900 first, because AI-200 assumes working Azure knowledge.

Validates knowledge of building AI-enabled Azure back-end solutions using containers, Cosmos DB, PostgreSQL, Redis, messaging, security and monitoring.
AI-200 tests whether you can build, connect, secure and monitor the back end of an AI solution on Azure. Most of the work is code, so build each service in your own subscription with the Python SDKs, then practise reading scenarios and choosing the right service and setting.
Example scenario: A queue-processing container should scale to zero overnight and scale out when messages arrive. Where would you host it and how would you configure scaling?
Example scenario: A RAG app must return the five most similar documents but only from the customer's own region. How would you write the PostgreSQL query?
Example scenario: Some orders fail processing three times in a row. How would you keep them for investigation without blocking the queue?
Example scenario: A request is slow somewhere between a Function, Service Bus and Cosmos DB. How would you find which hop adds the delay?
AI-200 study guide · Skills measured as of October 7, 2026
Choose the best answer from the options, or the stated number of answers when a question asks for more than one.
A function must process each Service Bus message at least once and keep messages that fail repeatedly for later review. What should you rely on?
A. After the maximum delivery count, Service Bus moves the message to the dead-letter queue where you can inspect it without blocking the queue.
Drag items from a list into the matching places; an item can be used once, more than once or not at all.
Drag each requirement to the service that meets it. The requirements are store JSON documents with vector search, cache frequent lookups with expiry, and run code when a blob is created.
Store JSON documents with vector search → Azure Cosmos DB for NoSQL, cache lookups with expiry → Azure Managed Redis, run code when a blob is created → Azure Functions.
Pick one value from each drop-down list in a block of text, code or a diagram to complete it.
Select the correct value in each list. A Container Apps service scales on queue length by using a [KEDA scale rule / manual replica count], and can scale down to [zero replicas / one replica] when the queue is empty.
KEDA scale rule and zero replicas. A scale rule reacts to queue length and Container Apps can scale to zero.
Read a description of a fictional company, then answer several questions that share the same background.
Northwind runs a support bot that must find similar past tickets quickly. Which data service stores the ticket embeddings so a query can filter by customer region?
A. A vector index returns similar tickets, and the filter limits results to the right region.
Complete tasks in a live environment by using the portal or code, and the exam checks the result.
Create a container app from an existing image and set it to scale on queue length.
Deploy the image to a Container Apps environment and add a queue-based scale rule with a minimum of zero replicas.
Sign in to Microsoft Learn with a personal Microsoft account, complete your certification profile, then schedule AI-200 with Pearson VUE from the certification page.
Take the free practice assessment to see where you start. Then work through one course and build every study guide bullet in your own subscription.
Most candidates spend 60–100 hours. Plan for the low end if you already build containerized apps and use Azure data and messaging services; plan for the high end if several of these services are new to you.
Use the official course for coverage and treat the study guide as a checklist. A video course with code walkthroughs helps you build along.
Tick off every bullet. Third-party material is still catching up with this exam. Check each study guide bullet against Microsoft Learn documentation and build a small example for it.
Start with Microsoft's free practice assessment, then a larger question bank. For every miss, find the matching study guide bullet and rebuild that piece.
Choose the checklist for your test-centre or online appointment.
When a candidate with very basic cloud knowledge asked where to begin, an experienced commenter pointed them to AI-901 and AZ-900 first, because AI-200 assumes working Azure knowledge.
One candidate estimated about two thirds overlap with AZ-204 and used AZ-204 practice tests to warm up. The container orchestration and vector database topics are new.
With prep material thin in the exam's first months, a candidate advised working through every task listed in the study guide rather than relying on one course.