Microsoft retired the AI-900 exam on 30 June 2026, which means AI-901 is now the only way to earn the Microsoft Certified: Azure AI Fundamentals credential. If you were part way through AI-900 study, or you are starting fresh, you need to know how to pass AI-901 without wasting weeks on the wrong material. This guide gives you a realistic 4-week study plan, a domain-by-domain breakdown, and clear guidance on the biggest change in the exam: the hands-on Microsoft Foundry content.
AI-901 is not a rebadge of AI-900. The old exam asked you to describe what Azure AI services do. The new one asks you to implement AI solutions using Microsoft Foundry, so more than half the scored questions now sit in a practical, build-something domain. Get that shift right and this is a very passable fundamentals exam.
AI-901 Exam Facts You Need First
Before you plan your study, anchor yourself to the current facts. AI-901 is a fundamentals-level exam, so it is broad rather than deep, but the Foundry weighting makes it more hands-on than most people expect from a "Fundamentals" title.
- Exam name: Microsoft Azure AI Fundamentals (Exam AI-901)
- Credential earned: Microsoft Certified: Azure AI Fundamentals
- Passing score: 700 out of 1000 (scaled, not a raw percentage)
- Questions: roughly 40 to 60, varies by session
- Duration: 45 minutes of exam time
- Cost: 99 USD (price varies by country)
- Prerequisites: none enforced, but Python syntax and basic Azure familiarity are recommended
- Renewal: annually, via a free online assessment on Microsoft Learn
Exam Tip: The 700 passing mark is a scaled score, not 70 percent of questions correct. Do not assume you can drop 30 percent of the paper and still pass. Aim to be comfortable across both domains rather than banking on one.
The skills measured were last updated on 15 April 2026, when AI-901 entered beta, so any study material written for AI-900 that does not mention Microsoft Foundry is now out of date. That is the single most common trap in 2026: revising from a syllabus that no longer matches the exam.
What Changed From AI-900 to AI-901
The credential is identical. If you already hold AI-900, it stays valid and you do not need to retake anything. The change is in the exam content, and it is significant.
| Area | AI-900 (retired) | AI-901 (current) |
|---|---|---|
| Core question style | Describe what a service does | Implement a solution using Foundry |
| Generative AI | Light coverage | Core topic, including prompts and agents |
| Platform focus | Catalogue of separate Azure AI services | Unified Microsoft Foundry platform |
| Hands-on weighting | Minimal | 55 to 60 percent of scored questions |
| Python expectation | Not really needed | Recommended, SDK snippets appear |
The practical half of AI-901 is built around Microsoft Foundry (formerly Azure AI Foundry). You are expected to understand deploying a model in the Foundry portal, writing effective system and user prompts, building a lightweight chat client with the Foundry SDK, and creating and testing a single-agent solution. This is the part AI-900 veterans underestimate.
If you want the full side-by-side, we cover it in detail in AI-900 vs AI-901: What's Changed in Microsoft's Azure AI Fundamentals Exam. This post assumes you have decided to sit AI-901 and now want a plan to pass it.
The Two AI-901 Domains and Their Weightings
AI-901 has just two skill areas, but their weightings tell you exactly where to spend your time.
Domain 1: Identify AI concepts and capabilities (40 to 45 percent)
This is the knowledge half. It covers the principles of responsible AI (fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability), how generative AI models work, choosing an appropriate model, and identifying common AI workloads. Workloads include generative and agentic AI, text analysis, speech, computer vision, and information extraction.
Most of this domain is recognition and scenario matching. You will see questions like "which principle of responsible AI applies here" or "which workload solves this problem." It rewards broad familiarity and clear definitions.
Domain 2: Implement AI solutions by using Microsoft Foundry (55 to 60 percent)
This is the larger and harder half. It splits into four practical themes:
- Generative AI apps and agents: create system and user prompts, deploy and interact with a model in the Foundry portal, build a lightweight chat client with the Foundry SDK, and create and test a single-agent solution.
- Text and speech: build a lightweight text-analysis app, respond to spoken prompts with a deployed multimodal model, and use Azure Speech in Foundry Tools.
- Computer vision and image generation: interpret visual input with a multimodal model and create new visual outputs with generative models.
- Information extraction: use Azure Content Understanding in Foundry Tools to extract information from documents, forms, images, audio, and video.
Exam Tip: Because Domain 2 is worth up to 60 percent of your marks, one afternoon actually clicking through the Foundry portal is worth more than a full day of passive reading. You cannot fake familiarity with a tool you have never opened.
The 4-Week AI-901 Study Plan
This plan assumes around 6 to 8 hours of study per week, which is realistic for someone in full-time work. If you already work with Azure or Python daily, you can compress it into 3 weeks. If AI, Python, and Azure are all new to you, stretch it to 5 or 6 weeks and repeat weeks 2 and 3.
Week 1: Foundations and responsible AI
Start with Domain 1 because it builds the vocabulary you need later. Work through the Microsoft Learn modules on core AI concepts and responsible AI. Your goal this week is to be able to define each of the six responsible AI principles and give one example of each.
- Learn the six responsible AI principles cold: fairness, reliability and safety, privacy and security, inclusiveness, transparency, accountability
- Understand how generative AI models work at a conceptual level, including tokens, prompts, and grounding
- Learn the common AI workloads and be able to match a scenario to the right one
- Set up a free Azure account so you are ready for hands-on work next week
End the week with a short set of practice questions on Domain 1 concepts to confirm the definitions have stuck.
Week 2: Get hands-on with Microsoft Foundry
This is the most important week. Open the Foundry portal and actually use it. Reading about Foundry will not carry you through 55 to 60 percent of the exam.
- Deploy a model in the Foundry portal and chat with it directly
- Write and test system prompts and user prompts, and observe how changing the system prompt changes behaviour
- Create and test a single agent in the Foundry portal
- If you know a little Python, build a lightweight chat client using the Foundry SDK
You do not need to become a developer. You need to recognise what each step looks like and why you would do it, because the exam asks scenario questions grounded in this workflow.
Week 3: Text, speech, vision and information extraction
Now cover the remaining Domain 2 themes. Spend a session on each so none of them surprise you.
- Text analysis: keyword extraction, entity detection, sentiment analysis, and summarisation
- Speech: speech recognition and speech synthesis, and responding to spoken prompts with a multimodal model
- Computer vision: interpreting images with a multimodal model and generating new images
- Information extraction: using Azure Content Understanding to pull data from documents, forms, images, audio, and video
Build a simple service-selection matrix as you go: a one-page table that maps a scenario to the correct workload and Foundry capability. This single artefact answers a large share of exam questions.
Week 4: Practice exams and weak-spot repair
The final week is about exam readiness, not new content. Take full-length, timed practice exams under realistic conditions. Your aim is to score consistently above the 700 equivalent before you book.
- Sit at least two or three full timed practice papers
- Review every wrong answer and, more importantly, every lucky guess
- Revisit your service-selection matrix and the six responsible AI principles the night before
- Book the exam for a morning when you are fresh, and use the sandbox to get used to the interface
Exam Tip: Track not just your score but your timing. AI-901 gives you 45 minutes, which is comfortable, but multi-step Foundry scenario questions can eat time. Practise flagging and moving on rather than stalling.
Common Reasons People Fail AI-901
Fundamentals exams have a reputation for being easy, which is exactly why people underprepare and fail. The most common causes in 2026 are specific to this refreshed exam.
- Revising from AI-900 material. If your notes never mention Foundry, they are out of date. This is the number one trap.
- Treating Foundry as theory. Domain 2 is a build domain. Candidates who never open the portal lose the majority of their marks.
- Ignoring Python entirely. You do not need to be a coder, but you should be able to read a short SDK snippet and understand what it does.
- Cramming definitions but skipping scenarios. The exam is scenario-led. Rote definitions without application do not transfer.
If this pattern sounds familiar from other exams, we break down the psychology of it in Why Most People Fail Certification Exams.
Should You Sit AI-901 First, or a Different AI Cert?
AI-901 is the natural entry point into Microsoft's AI certification track. From here, candidates typically progress to associate-level exams that build on Foundry, such as the Azure AI developer and agent paths. If you are weighing Microsoft against Amazon at the fundamentals level, our AWS AI Practitioner vs Azure AI Fundamentals comparison will help you choose which to take first.
If you have a free Microsoft voucher burning a hole in your pocket, AI-901 is one of the smartest fundamentals exams to spend it on right now, precisely because the AI-900 retirement has created a fresh, in-demand credential. We ranked the options in Got a Free Microsoft Certification Voucher? Here's the Smartest Exam to Spend It On in 2026.
Ready to Start Practising?
The fastest way to pass AI-901 is to combine hands-on Foundry time with realistic, exam-style practice questions that mirror the two-domain structure and the scenario format. Reading alone will not get you to a consistent 700.
CertCrush gives you focused AI-901 practice with detailed explanations, so you can drill responsible AI principles, workload matching, and Foundry scenarios until they are automatic. Create your free account to start practising, browse our full course catalogue to see everything available, and turn this 4-week plan into a first-time pass.
Study the right syllabus, open the Foundry portal, practise under timing, and AI-901 becomes a straightforward win.