Prep
AI-900 study guide: first, the exam has changed
If you are searching for an AI-900 study guide, the most important fact is that you can no longer take AI-900. Microsoft's official AI-900 study guide page now says the exam was retired on June 30, 2026. To earn the Microsoft Certified: Azure AI Fundamentals certification today, you pass Exam AI-901: Microsoft Azure AI Fundamentals. The certification name is the same; the exam behind it is new.
This matters for your preparation. Most AI-900 courses, question banks and videos online were written for the old exam, which was almost entirely about describing AI concepts and Azure AI services. AI-901 is built differently, and older material covers less than half of it.
Everything below comes from Microsoft Learn's AI-901 study guide (skills measured as of April 15, 2026), the certification and exam pages, and Microsoft's announcement on its Tech Community blog, all checked on 7 November 2026.
| Skill area | Official weight | Days |
|---|---|---|
| Identify AI concepts and capabilities | 40-45% | 1-3 |
| Implement generative AI apps and agents with Foundry | Part of 55-60% | 4-6 |
| Implement text, speech and vision solutions with Foundry | Part of 55-60% | 8-9 |
| Implement information extraction with Foundry | Part of 55-60% | 10-11 |
| Review and timed practice | - | 7, 12-14 |
AI-900 vs AI-901: what changed
| AI-900 (retired June 30, 2026) | AI-901 (current) | |
|---|---|---|
| Audience in Microsoft's comparison | Nontechnical or technical beginners | Technical beginners planning to build AI solutions |
| Coding knowledge | None | Basic Python syntax and programming concepts |
| Skill areas | 5 "describe" areas: AI workloads, machine learning, computer vision, NLP, generative AI | 2 areas: Identify AI concepts and capabilities (40-45%); Implement AI solutions by using Microsoft Foundry (55-60%) |
| Style | Conceptual | Concepts plus building lightweight apps and agents, with REST APIs, SDKs and CLIs |
| Certification | Azure AI Fundamentals | Azure AI Fundamentals (same name) |
If you already passed AI-900, Microsoft says no action is required: retirement does not revoke a certification already earned, and Fundamentals certifications do not expire. Taking AI-901 would not change your credential's name.
AI-901 exam facts
| Item | AI-901 |
|---|---|
| Certification | Microsoft Certified: Azure AI Fundamentals |
| Exam time | Microsoft's standard for Fundamentals exams is 45 minutes (65 minutes seat time); the AI-901 page does not list a separate time |
| Questions | Not published. Microsoft says most of its exams have 40 to 60 questions |
| Passing score | 700 on a scale of 1 to 1,000 (scaled, not a percentage) |
| Microsoft Learn during the exam | Not allowed on Fundamentals exams |
| Languages | 13: English, Arabic, Chinese (Simplified and Traditional), French, German, Indonesian, Italian, Japanese, Korean, Portuguese (Brazil), Russian, Spanish |
| Price | Set by the country where the exam is proctored; Microsoft shows your price when you book |
| Prerequisites | None |
| Validity | Does not expire |
Note one inconsistency on Microsoft's exam page: its short description still lists the old AI-900 topics. The skills measured list and the study guide are the authoritative version, and that is what this plan follows.
The AI-901 skills measured
Identify AI concepts and capabilities (40-45%)
- Responsible AI: fairness, reliability and safety, privacy and security, inclusiveness, transparency and accountability.
- AI model components and configurations: how generative AI models work, choosing a model by capability, and deployment options and configuration parameters.
- AI workloads: scenarios for generative and agentic AI, text analysis, speech, computer vision and information extraction; text analysis techniques such as keyword extraction, entity detection, sentiment analysis and summarisation; speech recognition and synthesis; computer vision and image generation.
Implement AI solutions by using Microsoft Foundry (55-60%)
- Generative AI apps and agents: writing system and user prompts, deploying a model and using it in the Foundry portal, a lightweight chat client with the Foundry SDK, and creating and testing a single-agent solution and its client app.
- Text and speech: an app with text analysis, responding to spoken prompts with a multimodal model, and an app using Azure Speech in Foundry Tools.
- Computer vision and image generation: visual input in prompts, creating images with generative models, and an app with vision capabilities.
- Information extraction: using Azure Content Understanding in Foundry Tools to extract information from documents and forms, images, audio and video, and building an app with it.
The 2-week AI-901 study plan
This plan assumes about 1 to 1.5 hours a day, some familiarity with Azure resources and enough Python to read and edit a short script. If Python is new to you, add a week of basic Python first. If you already build with Azure AI services, you can compress it into one week by doubling up days.
| Days | Skill area (weight) | Focus and hands-on task |
|---|---|---|
| 1 | Concepts (40-45%) | Baseline practice questions. Responsible AI: the six principles, each with an example of it going wrong |
| 2 | Concepts | How generative models work, tokens, choosing a model by capability, deployment options and parameters such as temperature |
| 3 | Concepts | AI workloads: generative vs agentic AI, text analysis techniques, speech, vision and information extraction. Match 20 business scenarios to a workload |
| 4 | Foundry (55-60%) | Set up a Foundry project, deploy a model and test it in the portal. Write system prompts that change its behaviour |
| 5 | Foundry | Build a lightweight chat client with the Foundry SDK in Python |
| 6 | Foundry | Create and test a single agent in the portal, then call it from a small client app |
| 7 | Review | Mixed practice set; rework every miss against the skills list |
| 8 | Foundry: text and speech | A small text analysis app; spoken prompts with a multimodal model; Azure Speech in Foundry Tools |
| 9 | Foundry: vision | Images in prompts, generating an image, a small app with vision capabilities |
| 10 | Foundry: information extraction | Content Understanding on a document or form, an image, and an audio or video file |
| 11 | Foundry: information extraction | Build a small app that uses Content Understanding; review REST, SDK and CLI basics |
| 12 | Mixed review | Timed practice in 45 minutes; list weak objectives |
| 13 | Weak areas | Repeat the labs and reading for your weakest objectives; try the exam sandbox |
| 14 | Final check | A last timed set, then book or sit the exam |
Delete your Azure resources after each session and set a budget alert on day 4. Model deployments and some AI services are billed while they exist.
How to use practice questions for AI-901
- Check the exam code on any question bank. Material labelled AI-900 covers the old exam. It still helps for responsible AI and workload concepts, but not for the Foundry implementation half.
- Use Microsoft's free Practice Assessment. For AI-901 it is on AI Skills Navigator, and you need to be signed in to launch it.
- Turn misses into tasks. If you miss a question about agents or Content Understanding, go and build the thing in Foundry. The exam is written for people who have done it.
- Practise in 45 minutes. That is Microsoft's standard Fundamentals exam time, and there is no Microsoft Learn lookup to fall back on.
Our free Azure AI Fundamentals practice test is a quick way to find your weak areas on day 1 and again on day 12.
Exam day tips for AI-901
- Run the exam sandbox first. Microsoft does not reveal question types in advance, but its sandbox shows the formats and the interface.
- Watch the clock and breaks. Microsoft builds 5 minutes into the exam time for unscheduled breaks, but the clock keeps running, and you cannot return to questions you saw before a break.
- Read code questions slowly. Expect short snippets and configuration choices. Know what each step in a basic Foundry SDK chat client does.
- Book with a personal Microsoft account. Microsoft warns that exam records tied to a work or school account can be lost if you leave the organisation.
AI-901 or something else?
AI-901 suits people at the start of a technical AI career. If you are in a non-technical role and want a conceptual overview, note that Microsoft now describes this certification for technical beginners who will build AI solutions, with Python expected. If you are new to cloud altogether, AZ-900 may be a better first step; we compare the two in AI-900 vs AZ-900.
For general study methods behind this plan, such as retrieval practice and spacing, see our guide on how to study for exams.