You can pass the AI-901 in two weeks if you split the time the way the exam is weighted: spend week one on the concepts (40 to 45% of the marks) and week two getting hands-on with Microsoft Foundry (55 to 60%), with a timed mock at the start, middle and end. This plan assumes about one focused hour a day plus a longer session at each weekend, and it works to the official skills outline that Microsoft last updated on 15 April 2026.
How much study does the AI-901 really need?
Most candidates pass the AI-901 with one to two weeks of focused study. It is a Fundamentals exam with no prerequisites, so the depth stays shallow, but the 2026 refresh added a large hands-on Microsoft Foundry section, so pure reading no longer carries you the way it did on the old AI-900. If you already write a little Python and have clicked around the Azure portal, two weeks is comfortable. If you are starting from zero coding, give yourself the full two weeks and protect the hands-on days.
For how the difficulty actually feels, see how hard the AI-901 really is. For the complete skills list this plan is built on, see what’s on the AI-901 exam.
What does each week cover?
| Week | Focus | Exam weight | Goal by Sunday |
|---|---|---|---|
| Week 1 | Identify AI concepts and capabilities | 40-45% | Responsible AI and AI workloads locked in |
| Week 2 | Implement AI solutions with Microsoft Foundry | 55-60% | Deployed a model, sent prompts, built a tiny chat client |
The headline is the same as the exam’s: more than half the marks live in the Microsoft Foundry implementation half, so week two is the one you cannot skip.
Week 1: lock in the concepts
The conceptual half is three groups: responsible AI, AI model basics, and the common AI workloads. Knock them out in order.
- Day 1 - Baseline mock. Sit a full timed mock cold, before you revise anything. You will score low, and that is the point: the per-domain breakdown tells you which half is weaker so you can spend your two weeks where it counts. Use the free AI-901 practice test for this.
- Day 2 - Responsible AI. Learn the six principles: fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability. Know one plain example of each. This group looks soft, so people under-revise it and then meet it repeatedly on the exam.
- Day 3 - AI models. How generative AI models work at a high level, choosing an appropriate model for a job, and the main deployment options and configuration parameters.
- Day 4 - AI workloads, part one. Generative and agentic AI, plus text analysis (keyword extraction, entity detection, sentiment analysis, summarisation).
- Day 5 - AI workloads, part two. Speech (recognition and synthesis), computer vision and image generation, and information extraction from text, images, audio and video.
- Weekend - Concepts mock + review. Re-sit a timed mock. You should see the concepts domain climb. Review every wrong answer and read the linked Microsoft Learn reference for it.
By the end of week one the concepts half should feel routine. Resist the urge to keep polishing it: diminishing returns set in fast, and week two is where most candidates actually lose marks.
Week 2: get hands-on with Microsoft Foundry
This is the half that did not exist on the AI-900, and the exam expects you to have actually built small things, not just described them. You do not need to ship production code; you need the tasks to feel familiar.
- Day 8 - Deploy and prompt. Deploy a model in the Microsoft Foundry portal and interact with it. Practise writing clear system and user prompts.
- Day 9 - Build a chat client. Create a lightweight chat client using the Foundry SDK. This is the single highest-value hands-on task; having done it once removes a whole cluster of exam questions.
- Day 10 - Agents. Create and test a single-agent solution in the Foundry portal, then a small client app for an agent.
- Day 11 - Text, speech and vision. Try a text-analysis app, respond to a spoken prompt with a deployed multimodal model using Azure Speech in Foundry Tools, and interpret a visual input in a prompt.
- Day 12 - Information extraction. Extract information from a document or form using Azure Content Understanding in Foundry Tools. Know that the Foundry portal, the Foundry SDK and Content Understanding do different jobs, because questions check that you reach for the right one.
- Weekend - Full mock + weak-spot drill. Sit a final timed mock under real conditions. Drill whichever domain is still under 700, then re-sit until you clear 700 out of 1000 consistently.
Microsoft’s audience profile for the AI-901 expects “knowledge of Python coding syntax and programming techniques” and familiarity with “REST APIs, SDKs, and CLIs,” so reading along with the SDK calls in week two matters more than memorising them. Source: the official AI-901 study guide.
How should you use mock exams across the plan?
Three mocks, spaced out, beat ten mocks crammed at the end:
- Mock 1 (Day 1): diagnostic. Find your weak half before you spend any time.
- Mock 2 (weekend one): checkpoint. Confirm concepts are landing and spot any stragglers.
- Mock 3 (weekend two): dress rehearsal. Full timed conditions, then drill the gap.
The pass mark is 700 out of 1000, and there is no penalty for a wrong answer, so on every mock answer every question rather than leaving blanks. Check the official AI-901 exam page for current pricing and to book, since the fee is set by the country or region where the exam is proctored.
Is two weeks too aggressive?
If you can only manage thirty minutes a day, stretch the same sequence over three to four weeks rather than cutting the hands-on days. The one thing that breaks a two-week plan is treating the AI-901 like the AI-900 and skipping Foundry because “it is only Fundamentals.” The implementation half is the bigger half. For more on that shift, see how hard the AI-901 really is.
Practise the way this plan is structured
PassTheVoid mirrors the official 40-45 / 55-60 split, so your daily questions are weighted like the real exam, not skewed toward whichever domain was easiest to write. You get a free timed mock for your Day 1 diagnostic, a per-domain breakdown that tells you whether to revise concepts or Foundry next, and offline study so the daily habit survives a commute. It is free and ad-free, with a one-time Pro unlock for the full question bank and unlimited mock exams.