AI-901 is the current exam for Microsoft Certified: Azure AI Fundamentals. AI-900 retired on 30 June 2026, and Microsoft now awards the same certification only to candidates who pass AI-901. This AI-901 exam guide covers the two domains, the cost, the pass mark and a three-week plan. However, if you already hold the certification through AI-900, you keep it and you do not need to take AI-901.
In contrast, if you are starting now, AI-901 is the only way to earn it, and it is a different exam from AI-900: more than half of it asks you to build in Microsoft Foundry, the Azure portal for AI apps and agents.
After this guide, the natural next step is the Microsoft AI-103 exam guide, and the AB-620 and GH-700 guide covers the builder and business exams.
AI-901 exam guide: what changed after AI-900
AI-901 replaced AI-900 for Microsoft Certified: Azure AI Fundamentals. The exam runs 45 minutes, pass mark 700, costs $99 in the US, and tests concepts at 40 to 45 percent and Foundry building at 55 to 60 percent. This guide covers the domains, the cost and a three-week plan.
It became generally available in June 2026.
In contrast, AI-900 was a concepts exam about AI workloads and Azure services, with no coding required. AI-901 keeps the same certification name but changes the test: AI-900 asked what AI is, and AI-901 asks how to build an AI app in Foundry.
However, the audience changed with the exam. AI-901 is aimed at people early in an AI development career, who already understand Azure AI at a conceptual level, can read and write basic Python, and know their way around Azure resources. However, old AI-900 notes do not cover the Foundry work that now carries most of the weight.
AI-901 exam guide: cost, duration and pass mark
The facts below come from the official AI-901 study guide and the published Microsoft price table, checked 21 November 2026. Confirm the price when you book, because Microsoft prices by country.
| Exam fact | AI-901 detail |
|---|---|
| Exam code | AI-901: Azure AI Fundamentals |
| Certification earned | Microsoft Certified: Azure AI Fundamentals |
| Exam time | 45 minutes, inside a 65-minute appointment |
| Question count | Not published by Microsoft |
| Pass mark | 700 on a scale of 1 to 1,000 |
| Price | $99 in the US; varies by country |
| Delivery | Pearson VUE, online proctored or at a test center |
| Languages | 13, including English, French, German, Japanese and Spanish |
| Practice | Free practice assessment on AI Skills Navigator (sign-in required); free exam sandbox |
| Validity | Does not expire |
Microsoft does not publish an exact question count for AI-901, so any source that quotes one is guessing. The pass mark of 700 is a scaled score: Microsoft reports exam scores on a 1 to 1,000 scale, so a scaled score may not equal 70 percent of the points.
The practice assessment is on AI Skills Navigator, Microsoft’s certification practice hub, and the exam sandbox is a free demo of the exam screen.
AI-901 domains: what the exam tests
Overall, the skills outline dated 15 April 2026 splits the exam into two domains.
Domain 1: Identify AI concepts and capabilities (40 to 45 percent)
Three parts. First, the principles of responsible AI: fairness, reliability and safety, privacy and security, inclusiveness, transparency and accountability. Second, model components and configurations: how generative AI models work and how to pick a model and deployment option. Third, AI workloads: generative and agentic AI (AI that plans steps and uses tools), text analysis, speech, computer vision and information extraction, including keyword extraction, entity detection, sentiment analysis and summarization.
Domain 2: Implement AI solutions by using Microsoft Foundry (55 to 60 percent)
First, generative AI apps and agents: effective system and user prompts, deploying a model in the Foundry portal, a lightweight chat client using the Foundry SDK, the code library for calling the model, and creating and testing a single-agent solution in the portal.
Next, text and speech: a lightweight text-analysis application, spoken prompts answered by a deployed multimodal model, and an application using Azure Speech in Foundry Tools.
Computer vision and image generation: visual input in prompts with a deployed multimodal model, new visual outputs from generative models, and information extraction from documents, forms, images, audio and video with Azure Content Understanding.
AI-901 walkthrough 1: deploy a model and build a chat client
Step 1: pick the model by capability. For example, suppose the task is a support assistant that answers questions from product documents. Pick a text generation model, such as GPT-4o; an image-generation model does not fit. The exam tests exactly this: match the model to the capability, then match the deployment option to the use case.
Step 2: deploy the model and talk to it. Deploy the chosen model in the Foundry portal, then test it in the playground, the test chat, with a system prompt such as “Answer only from the product documents” and a user question such as “What is the warranty?” Change the system prompt, then watch the answer change.
Step 3: build a lightweight chat client with the Foundry SDK. Write a short Python program that sends a user prompt to the deployed model, then prints the reply. Finally, keep it small: connect, send, print. Creating a client like this is itself a measured skill in the study guide.
AI-901 walkthrough 2: build a single agent and extract information
This mirrors the second Foundry skill set: agents and information extraction, the part AI-900 holders have never been tested on.
Step 1: create a single agent in the Foundry portal. First, give the agent clear instructions stating its role, what it may and must not do. Add one tool, such as a search or document lookup, and test the agent with two or three prompts. Then test each prompt twice: inside the role and outside it.
Step 2: build a lightweight client for the agent. It works like the chat client, but it calls the named agent from step 1 instead of a model, so read the reply plus any tool output it returns.
Step 3: extract information from documents and images. Take a sample invoice, then run it through Azure Content Understanding in Foundry Tools. Note the fields it pulls out: invoice number, date, total and line items. Then try a photo of a form.
Step 4: extract from audio and video, then try speech. Extract information from a short audio clip and a video, then answer a spoken prompt with a deployed multimodal model. Finish with the small speech application built with Azure Speech in Foundry Tools.
AI-901 exam traps to avoid
Trap 1: studying AI-900 material only. AI-901 keeps the concepts but adds Foundry implementation at 55 to 60 percent.
Trap 2: skipping Python. In fact, the study guide says candidates need basic Python and Azure familiarity, and it lists building a lightweight chat client as a measured skill.
Trap 3: memorizing service names instead of building. Domain 2 asks you to deploy, create, build and test, so naming features without deploying a model is not enough.
Trap 4: trusting a quoted question count. However, Microsoft does not publish one for AI-901. Instead, plan on the 45-minute clock, not on a question count Microsoft has not published.
Trap 5: treating the certification as expiring. Fundamentals certifications do not expire. AI-900 holders keep the certification and do not need to take AI-901.
Trap 6: confusing AI-901 with AI-103. It is the fundamentals exam: concepts plus small Foundry builds. AI-103 is the associate exam for developers building AI apps and agents.
AI-901 study plan: three weeks
Week 1: concepts. First, work through the responsible AI principles, model components and workloads in Domain 1 with the free Microsoft Learn Azure AI training. End the week by writing one-line definitions for the six responsible AI principles from memory, then matching five workloads to their techniques.
Week 2: Foundry hands-on. Do both walkthroughs above: deploy a model, build the chat client, create and test a single agent, then run Content Understanding on a document, an image, audio and video.
Week 3: practice and review. Take the free practice assessment on AI Skills Navigator, then try the free exam sandbox. Review every wrong answer against the study guide, then run both walkthroughs once more. Finally, book only after you finish the practice assessment inside 45 minutes and can explain every wrong answer.
AI-901 questions candidates ask
How many questions does AI-901 have? Microsoft does not publish a question count, so plan on the 45-minute clock.
Do AI-900 holders need to take AI-901? No. The certification earned through AI-900 stays valid.
Does the AI-901 certification expire? No. Fundamentals certifications do not expire.
AI-901 exam guide: one-page revision
AI-901 exam guide in one page. It replaced AI-900 as the route to Microsoft Certified: Azure AI Fundamentals. Two domains: concepts 40 to 45 percent, Foundry implementation 55 to 60 percent. Exam time 45 minutes inside a 65-minute appointment; pass mark 700; price $99 in the US; 13 languages; Pearson VUE online or at a test center. Basic Python and Azure familiarity expected. The certification does not expire. Free preparation: the practice assessment and the exam sandbox.
AI-901 rewards candidates who build. Deploy a model, build the chat client, test an agent, then extract information before exam day. The source for the syllabus and pass mark in this guide is the official AI-901 study guide on Microsoft Learn; the price and booking facts come from the published Microsoft price table. For the next rung up, read the Microsoft AI-103 exam guide, then the AB-620 and GH-700 guide.
Enjoying this? Add StudyFromNotes as a preferred source on Google to see our latest notes and updates more often:

Daniel Mercer is a finance writer focused on the ACCA and CPA exams, financial reporting, and career moves in accounting and finance. Daniel writes from the same syllabuses covered on this site and favors clear explanations, comparison tables, and practical checklists over theory alone. The beat covers exam technique, accounting standards in plain language, and how finance professionals plan their next role.
Leave a Reply