Microsoft · AI-200 · Associate
Developing AI Cloud Solutions on Azure (AI-200) — Practice Questions and Mock Exam
Practice real AI-200 questions, never dumps. Alex explains every answer, and your readiness score tells you when you're ready to pass.
Checked against Microsoft · July 2026Current exam version
Overview
About the exam
Exam AI-200, Developing AI Cloud Solutions on Azure, leads to the Microsoft Certified: Azure AI Cloud Developer Associate certification. Microsoft names it as the replacement for Exam AZ-204 (Azure Developer Associate), which retires on 31 July 2026 — Microsoft's own guidance tells AZ-204 candidates who are not close to testing to prepare for AI-200 instead. The shift in scope is the point: where AZ-204 tested general-purpose Azure development, AI-200 assumes the application being built is an AI system. Containers, vector databases, event-driven pipelines and observability are treated as the ordinary furniture of production AI, not as specialisms. Python proficiency is assumed throughout.
Exam Domains
What's on the exam
Expect roughly 40 to 60 questions in about 120 minutes of seat time. Item types include single-answer multiple choice, multiple response, drag-and-drop ordering and matching. Scoring is scaled from 1 to 1000 and 700 is required to pass, so the raw percentage of questions answered correctly is not the reported score. Most questions cover generally available features; Preview features can appear where they are in common use.
Container application hosting and orchestration: building, storing, versioning and managing images in Azure Container Registry and running ACR Tasks; deploying containers to Azure App Service with environment variables and secrets; deploying to Azure Container Apps with environment configuration and revision management; event-driven scaling with KEDA; deploying and managing AKS workloads from manifest files; and troubleshooting both AKS and Container Apps by inspecting logs, events and end-to-end connectivity. Vendor states this domain at 20-25%.
The vector-data core of the exam. Azure Cosmos DB for NoSQL: SDK connections and queries, tuning query performance and Request Unit consumption via indexing policies and consistency levels, storing and retrieving embeddings for vector similarity search, and change feed processors. Azure Database for PostgreSQL: schema modelling and indexing strategies, sizing compute/memory/storage for vector workloads, reducing pgvector compute overhead, vector similarity search and RAG patterns with metadata filtering, and connection optimisation. Azure Managed Redis: caching with expiration and invalidation, plus vector indexing for similarity search. Vendor states this domain at 25-30% — the single heaviest area.
Event- and message-based AI solutions plus serverless compute: queueing and processing back-end operations with Azure Service Bus including dead-letter queue handling, messages, topics and subscriptions; event-driven workflows with Azure Event Grid including filters, custom events and retries; and Azure Functions covering serverless APIs, triggers and bindings, and function app configuration and deployment. Vendor states this domain at 20-25%.
Securing secrets with Azure Key Vault including rotation and retrieval; storing and retrieving application configuration with Azure App Configuration; distributed tracing with OpenTelemetry SDKs; and writing KQL queries to analyse logs and metrics. Vendor states this domain at 20-25%.
SourceMicrosoft study guide
Format
What to expect
Watch out
Where candidates struggle
The most common trap is preparing as though this were AZ-204 with an AI chapter bolted on. It is not: the heaviest single domain is Azure data management services at 25-30%, and it is dominated by vector work — embeddings, similarity search, pgvector cost control and RAG with metadata filtering. Candidates who treat Cosmos DB and PostgreSQL as ordinary CRUD stores lose the most marks here. The second recurring trap is Request Unit consumption in Cosmos DB: indexing policies and consistency levels are examined as cost and performance levers, not as configuration trivia. Third, observability is tested concretely — OpenTelemetry tracing and hand-written KQL, not a general awareness of monitoring.
- 01AZ-204 habits — Preparing as if this were AZ-204 plus AI. The data-management domain is the heaviest at 25-30% and is vector-centric; general-purpose Azure development knowledge does not cover it.
- 02Request Units as trivia — RU consumption in Cosmos DB is examined as a tunable cost and performance lever via indexing policies and consistency levels, not as a configuration detail to memorise.
- 03pgvector cost blindness — PostgreSQL questions ask about reducing pgvector compute overhead and sizing compute, memory and storage for vector workloads — not merely whether a similarity query returns rows.
- 04Hand-waving observability — Monitoring is tested concretely: distributed tracing with OpenTelemetry SDKs and writing KQL queries against logs and metrics.
- 05Ignoring dead-letter handling — Service Bus questions include dead-letter queue handling, topics and subscriptions, not just basic send and receive.
Details
Exam logistics
The exam is delivered through Pearson VUE at a test center or as an online-proctored session. Localised versions are updated roughly eight weeks after the English version, and candidates taking the exam outside their preferred language can request an additional 120 minutes. The exam environment can be explored beforehand through Microsoft's exam sandbox.
SourceMicrosoft study guide
Before you book the exam
Would you pass AI-200 today?
Take the free readiness check. Answer real AI-200 questions and get your readiness score across every domain.
Take the free readiness check20 questions · freeReach 80% readiness by exam day. Pass, or your money back.