AI-901 Exam Guide 2026: Azure AI Fundamentals, Rebuilt
AI-900 was a conceptual exam you could pass in a weekend without writing a line of code. AI-901 carries the same “fundamentals” label and expects Python syntax and hands-on Foundry work. That is a much bigger change than the number suggests.
On this page
AI-901 is the current exam for Microsoft Certified: Azure AI Fundamentals, replacing AI-900. If you are about to book it on the strength of AI-900's reputation as the easiest AI certification going, stop and read this first.
Built against the official skills-measured document published 15 April 2026.
Why AI-901 is not AI-900 renamed
The audience profile is the giveaway. Microsoft says you should have conceptual knowledge of AI solutions in Azure and the foundational technical skills to work with them — then adds that you need knowledge of Python coding syntax and programming techniques and familiarity with Azure resources. The exam notes add that you should be familiar with REST APIs, SDKs and CLIs.
None of that was true of AI-900, which was a describe-and-recognise exam with no coding whatsoever.
| AI-900 (previous) | AI-901 (current) | |
|---|---|---|
| Nature | Conceptual — describe workloads and services | Conceptual + implementation |
| Coding | None required | Python syntax expected |
| Domain count | Five workload-shaped areas | Two |
| Centre of gravity | Azure AI service awareness | Microsoft Foundry, at 55–60% |
| Typical prep | A weekend of reading | Hands-on time in the Foundry portal and SDK |
Non-technical professionals who took AI-900 for AI literacy — managers, sales, analysts — are no longer the natural audience. If that is you and you want vocabulary rather than build skills, look at AB-900 for the Copilot administration angle, or SC-900 for security and compliance concepts, before defaulting to AI-901.
Exam format at a glance
As a fundamentals certification it does not expire — Microsoft's annual free renewal requirement applies to associate, expert and specialty credentials only. The free practice assessment has moved to AI Skills Navigator; you need to be signed in there to launch it. There is also the usual free exam sandbox for previewing the interface.
Two domains, and the 60/40 split
Only two domains, and the larger one is hands-on. Plan your study time the same way: more than half of it in the Foundry portal and SDK, not in reading.
Domain 1 — AI concepts and capabilities (40–45%)
Three sub-objectives.
Responsible AI principles — considerations for fairness, reliability and safety, privacy and security, inclusiveness, transparency and accountability. Six named principles; learn them as a list and be able to match a scenario to the right one.
AI model components and configurations — how generative AI models work, identifying an appropriate model based on capabilities, and identifying appropriate deployment options and configuration parameters. Model selection is now testable, not just model awareness.
AI workloads — scenarios for common workloads including generative and agentic AI, text analysis, speech, computer vision and information extraction; text-analysis techniques (keyword extraction, entity detection, sentiment analysis, summarisation); speech recognition and synthesis; computer vision and image-generation models; and techniques to extract information from text, images, audio and video.
Domain 2 — Building with Microsoft Foundry (55–60%)
Four sub-objectives, all phrased as implement or build. This is where the exam changed.
Reading will not get you there. Open the Foundry portal, deploy a model, and actually build the four things the objectives name: a chat client, a single agent, something that reads an image, and something that pulls fields out of a document with Content Understanding. The phrase “lightweight application” recurs deliberately — small working samples, not production systems.
What AI-901 does not test
Worth knowing, because AI-900-era material covers ground AI-901 has dropped or demoted.
- Service-by-service tours. AI-900 prep walked through each Azure AI service in turn. AI-901 asks you to pick an appropriate model and build with Foundry instead — the framing is task-first, not catalogue-first.
- Classical machine learning depth. Regression, clustering and the mechanics of training your own models are not the focus. That belongs to DP-100 and, operationally, to AI-300.
- Production engineering. Scaling, monitoring, cost control and deployment pipelines are out of scope. The objectives say lightweight application repeatedly, and they mean it.
The exam narrowed and deepened at once: fewer services described, more actually built.
Five things to build before you sit it
Every objective in domain 2 maps to something you can build in an afternoon. Build these five and you have covered most of the exam's larger half.
- A chat client using the Foundry SDK, against a model you deployed yourself in the Foundry portal.
- A single agent, created and tested in the portal, plus a lightweight client that calls it.
- Something that listens and answers — a spoken prompt handled by a deployed multimodal model, or Azure Speech in Foundry Tools.
- Something that sees — pass an image into a prompt and interpret it, then generate an image with a generative model.
- Something that extracts — pull fields from a document, an image, and an audio or video file using Content Understanding.
Keep each one small, and keep the code. Reading it back the night before beats any summary sheet, because the exam asks about the choices you made while building — which model, which deployment option, which parameter.
A four-week study plan
Coming from AI-900 or weighing it against other fundamentals? The comparison tool puts them side by side. Our AI-900 guide still explains the underlying Azure AI concepts well — use it for domain 1, and hands-on Foundry time for domain 2.
Frequently Asked Questions
Does AI-901 replace AI-900?
Yes. AI-901 is the current exam for the Microsoft Certified: Azure AI Fundamentals certification. The skills measured were published as of 15 April 2026.
Do I need to know Python for AI-901?
Yes. Unlike AI-900, the audience profile explicitly expects knowledge of Python coding syntax and programming techniques, plus familiarity with Azure resources, REST APIs, SDKs and CLIs.
How many domains does AI-901 have?
Two. Identify AI concepts and capabilities is 40 to 45 percent, and Implement AI solutions by using Microsoft Foundry is 55 to 60 percent.
Is AI-901 still a beginner exam?
It is still positioned at the start of an AI development career, but it is meaningfully more hands-on than AI-900. The majority of the exam is implementation using Microsoft Foundry rather than describing concepts.
Practise AI-901 Before You Book
500–1,000+ AI-901 practice questions with worked explanations across both domains.
AI-901 Practice Test
Practice Before You Book
500–1,000+ practice questions per exam with detailed explanations, across Azure, AWS, GCP, security, and AI certifications.