AI-901 study guide: domains, format and a 4-week plan
· 7 min read
AI-901, Microsoft Azure AI Fundamentals, is Microsoft's entry-level AI exam and the only exam required for the Microsoft Certified: Azure AI Fundamentals certification. It is aimed at people at the start of a career in AI solution development, and it expects more hands-on familiarity than its predecessor: basic Python and working knowledge of Azure resources. You have 45 minutes, a passing score of 700 on a 1–1,000 scale, and more than half the marks come from building small solutions with Microsoft Foundry.
AI-901 replaced AI-900, which Microsoft retired on 30 June 2026. The certification name is unchanged, but candidates must now pass AI-901 to earn it.
AI-901 at a glance
| Item | Detail |
|---|---|
| Provider | Microsoft |
| Level | Fundamentals |
| Questions | Not published for this exam (Microsoft says most of its exams contain 40–60) |
| Duration | 45 minutes (65 minutes of seat time) |
| Passing score | 700 |
| Exam fee (USD) | US$99 in the United States; the price varies by country or region |
| Languages | English, Arabic, Chinese (Simplified), Chinese (Traditional), French, German, Indonesian, Italian, Japanese, Korean, Portuguese (Brazil), Russian, Spanish |
| Delivery | Proctored, scheduled through Pearson VUE |
| Question formats | Microsoft does not announce formats per exam; its exam sandbox shows multiple choice, drag and drop, hot area, build list and similar types |
Details as of September 2026 — confirm on the official exam page before booking.
Who this exam is for
Microsoft describes the candidate as someone at the beginning of their career in AI solution development, with conceptual knowledge of AI on Azure and the foundational technical skills to work with it. The audience profile also asks for knowledge of Python syntax and programming techniques and familiarity with Azure resources.
There are no formal prerequisites. In practice, if you have never written a few lines of Python or created a resource in the Azure portal, spend a week on both before you start the plan below. AI-901 is not a pure vocabulary test: roughly 55–60% of it is about implementing things in Foundry.
What the exam covers
The AI-901 study guide lists skills measured as of 15 April 2026 in two areas.
Identify AI concepts and capabilities (40–45%)
This domain has three parts. The first is Microsoft's six responsible AI principles: fairness, reliability and safety, privacy and security, inclusiveness, transparency and accountability. Expect scenarios where you must name the principle at stake, so learn what a breach of each one looks like rather than memorising definitions.
The second part covers how generative models work, choosing a model by its capabilities, and the deployment options and configuration parameters you meet when you deploy one. Know what temperature and maximum output length change, and why a multimodal model suits a task a text-only model cannot handle.
The third part is workload recognition: generative and agentic AI, text analysis (keyword extraction, entity detection, sentiment, summarisation), speech recognition and synthesis, computer vision and image generation, and extracting information from text, images, audio and video. Most questions here describe a business need and ask which workload fits.
Implement AI solutions by using Microsoft Foundry (55–60%)
This is the larger domain and the one that separates AI-901 from the old AI-900. It asks you to write system and user prompts, deploy a model and test it in the Foundry portal, build a lightweight chat client with the Foundry SDK, and create and test a single agent with a small client application.
It then extends to text and speech (text analysis in an app, answering spoken prompts with a multimodal model, and Azure Speech in Foundry Tools), to vision (interpreting images in prompts and generating new images), and to information extraction with Azure Content Understanding in Foundry Tools across documents, forms, images, audio and video. Focus on what each step needs before it can work: a deployed model, an endpoint, credentials and the right tool for the input type.
A 4-week study plan
The plan follows the weights: about 40% of your time on concepts and 60% on Foundry work.
Week 1: concepts. Work through the responsible AI principles with your own examples of each. Then study how generative models produce output, what the common configuration parameters do, and how to match a model to a task. Finish the week by sorting twenty business scenarios into workloads.
Week 2: Foundry fundamentals. Create a Foundry project, deploy a chat model, and test prompts in the portal. Write a short Python chat client with the Foundry SDK. Build a single agent in the portal and call it from a small client. Keep notes on each resource and setting you needed.
Week 3: modalities. Build one small app per area: text analysis, a spoken prompt to a multimodal model, a speech-to-text or text-to-speech sample, an image passed to a multimodal model, and an image generation request. Then run Content Understanding against a form, an image and a short audio or video file, and compare what comes back.
Week 4: consolidation. Take a full-length mock under timed conditions, review every explanation, and spend two or three days on the domain that scored lowest. Take a second mock at the end of the week, and try Microsoft's exam sandbox so the interface holds no surprises.
Where candidates lose marks
- Studying for the old exam. A lot of free material online was written for AI-900, which covered machine learning principles in their own domain and no Foundry implementation. Check that anything you use follows the April 2026 skills list.
- Mixing up adjacent responsible AI principles. Transparency (people can understand how the system reaches its output) and accountability (people remain answerable for it) are the pair most often confused, followed by fairness and inclusiveness.
- Choosing a service when the question asks for a capability. Many items describe a need, such as pulling fields from invoices or summarising call recordings. Identify the workload first, then the tool.
- Skipping the build steps. Questions on the SDK and agent client test the order of operations and what must exist first. If you have only read about Foundry, these are guesses.
- Treating information extraction as one thing. Content Understanding handles documents, images, audio and video, and the guide lists each separately. Know what kind of output you get from each input.
- Running out of time. Forty-five minutes is short. Do not linger on one scenario; mark it and come back, and remember that Microsoft Learn is not available during fundamentals exams.
Practising for AI-901 on CertifyAIx
CertifyAIx has 590 original AI-901 questions, written against the April 2026 skills list and organised by its two skill areas. Each question explains why the correct answer is right and why each other option is wrong. You can read more about [how our practice questions are written](/blog/how-certifyaix-practice-questions-are-written).
- 3 full-length mock exams, timed to the real exam's 45 minutes
- 24 mixed domain papers and 27 topic papers of up to 25 questions each (60 minutes per 25)
- Progress tracking by domain and topic
See the [AI-901 practice questions](/certifications/azure-ai-fundamentals-ai-901), and for general method read [how to use practice exams effectively](/blog/how-to-use-practice-exams-effectively).
Frequently asked questions
How hard is AI-901?
It is a fundamentals exam, but harder than AI-900 for someone with no technical background, because more than half of it covers implementing solutions in Foundry. Basic Python and some time in the Azure portal make a large difference.
How long should I prepare for AI-901?
Most people should plan for three to six weeks. The shorter end suits someone who already writes Python and uses Azure; the longer end suits career changers, who need time to build the small apps the guide describes rather than only read about them.
Is AI-900 still available?
No. Microsoft retired AI-900 on 30 June 2026 and replaced it with AI-901. The Azure AI Fundamentals certification now requires AI-901.
Are CertifyAIx questions real AI-901 exam questions?
No. Every CertifyAIx question is original, written to rehearse the skills in Microsoft's published study guide. Using real exam content breaks the candidate agreement, as explained in [why exam dumps put your certification at risk](/blog/why-exam-dumps-put-your-certification-at-risk).
What should I take after AI-901?
If you want to build AI apps and agents on Azure, the associate-level exam is AI-103; see the [AI-103 study guide](/blog/ai-103-study-guide). AI-901 is not a prerequisite for it.
CertifyAIx is independent and not affiliated with Microsoft. Microsoft Azure AI Fundamentals is a trademark of its owner. All CertifyAIx practice questions are original.