The short version
Most developers with some Azure experience need roughly 40 to 80 hours in total for AI-200, spread over four to eight weeks. Give the most time to data management with vector search (30-35% of the exam), and plan real hands-on time: this exam rewards people who have deployed things. Your background changes these numbers a lot, so treat them as a starting point, not a promise.
Before you start
Open the Microsoft Learn AI-200 study guide and keep it beside you. This plan follows the version titled “Skills measured as of October 7, 2026”. Create an Azure subscription you can afford to break, and set a budget alert before you deploy anything. Several services here (Kubernetes, PostgreSQL, Redis, Cosmos DB) cost money while running, so delete resources when a session ends.
Keep a notes table with one row per decision: for each pair of look-alike options (Container Apps or App Service, HNSW or IVFFlat, Service Bus or Event Grid), write the one constraint that picks each. Our how to study guide has more on notes and spacing.
The 4-week plan (about 10 to 15 hours per week)
This plan suits people who already deploy containers and use at least one Azure database.
Week 1: Containers. Build an image with Azure Container Registry Tasks, deploy it to App Service with environment variables and secrets, then to Container Apps with revisions and a KEDA scale rule. Read an AKS manifest and use kubectl-style logs and events to find a broken deployment. About 10 hours.
Week 2: Data, part one. Cosmos DB for NoSQL: SDK queries, indexing policy, consistency levels, request units, vector search, and a change feed processor. Then Azure Managed Redis: caching, expiration, invalidation, and a vector index. About 12 hours.
Week 3: Data, part two and messaging. PostgreSQL with pgvector: schema design, HNSW and IVFFlat indexes, filters for retrieval-augmented generation, and connection pooling. Then Service Bus (dead-letter queues, topics, subscriptions) and Event Grid (filters, retries). About 14 hours.
Week 4: Functions, security, monitoring, review. Triggers and bindings, deployment settings, Key Vault with rotation, App Configuration, OpenTelemetry, and KQL queries. Spend the last two sessions on your weakest domain. About 12 hours.
The 8-week plan (about 5 to 8 hours per week)
| Week | Focus |
|---|---|
| 1 | Container Registry, Registry Tasks, App Service containers |
| 2 | Container Apps: environments, revisions, KEDA scaling |
| 3 | AKS manifests, logs, events, connectivity checks |
| 4 | Cosmos DB for NoSQL: queries, indexing, consistency, change feed |
| 5 | Cosmos DB vector search, then Azure Managed Redis |
| 6 | PostgreSQL and pgvector, then Service Bus and Event Grid |
| 7 | Azure Functions, Key Vault, App Configuration |
| 8 | OpenTelemetry, KQL, then full review and practice exams |
If Python is your weak spot, add two or three hours in week 1 and week 2 for SDK practice. The exam expects you to read Python code.
How to use practice exams
- At the start of week 1: take one practice set cold to find your baseline. A low score is normal and useful.
- After each domain: answer questions on that domain only and read every explanation, including for options you did not pick.
- Two weeks before the exam: take full-length sets under time pressure, and mark every question where you guessed.
- Keep an error log: write down why you missed each question: a fact you did not know, a service you mixed up, or a constraint you read too fast.
Practice exams measure readiness. They do not replace the Microsoft Learn material or hands-on work. Our practice tests are original questions, so they will not match the wording of the real exam, and that is on purpose. Read how we write our questions if you are curious.
The last 3 days
- Three days out: one full-length practice set, then review only the explanations for missed questions.
- Two days out: reread your notes table of look-alike options and your error log. Skim the study guide bullets and check each one against what you can explain without looking.
- The day before: light review for an hour at most. Try the exam sandbox so the interface is familiar. Confirm your booking time, ID, and room or test center. Sleep.
Watch the date
Microsoft updated the skills measured on October 7, 2026, and said the data management domain grew. Older blog posts and videos may show a different list. Check the study guide date before you rely on any resource.
More guides for AI-200
- AI-200 Exam Format, Scoring and Retakes
- AI-200 Sample Questions With Answers
- Is the AI-200 Exam Hard?
- AI-200 Prerequisites: What to Know First
- AI-200 Renewal and Retakes Explained
- Is AI-200 Worth It? An Honest Look
- AI-200 Cheat Sheet: Key Facts to Review
- AI-200 vs AI-300: Which Exam First?
- AI-103 vs AI-200: Which Exam Should You Take?
Next step
Use the domains explained page to see what each domain means in practice, then take a baseline with our AI-200 practice tests.
Frequently asked questions
How many hours do I need?
It varies. Developers who already run containers and use Cosmos DB or PostgreSQL may need 40 hours. Someone new to Azure may need well over 80.
Should I do hands-on labs?
Yes. Deploy a container app, create a vector index, send a dead-lettered message, and read a KQL result yourself. The exam describes real systems, and hands-on time makes the options easier to judge.
Is a 4-week plan enough?
It can be if you already work with most of these services. If three or more domains feel new, choose the 8-week plan.