Skip to content

Microsoft Azure AI Fundamentals (AI-901)

Microsoft's entry-level AI certification, rebuilt for the generative AI era. AI-901 replaced AI-900 (retired 2026-06-30) and earns the same credential, Microsoft Certified: Azure AI Fundamentals. It is still a fundamentals exam, but it is no longer only a vocabulary test: more than half of it is about building small things with Microsoft Foundry: deploying a model, writing prompts, a lightweight chat client and agent, and apps that use text analysis, speech, vision and Content Understanding.

Exam Details

  • Exam code: AI-901 (skills measured as of 2026-04-15)
  • Passing score: 700
  • Cost: varies by country or region; shown when you schedule
  • Duration and question count: not stated on the exam page as of 2026-10-21; check when booking
  • Prerequisites: none required. Microsoft expects conceptual AI knowledge, basic Python syntax, and familiarity with creating Azure resources
  • Renewal: Fundamentals certifications do not expire

Full detail in the fact sheet.

Skills measured

Domain Weight Notes
Identify AI concepts and capabilities 40-45% 01 AI concepts and responsible AI
Implement AI solutions by using Microsoft Foundry 55-60% 02 Generative AI apps and agents in Foundry, 03 Text, speech and vision, 04 Information extraction

What changed from AI-900

AI-900 (retired) AI-901
Five "describe" domains: workloads, ML on Azure, vision, NLP, generative AI Two domains: concepts (40-45%) and implementing with Foundry (55-60%)
Classic ML on Azure Machine Learning (regression, classification, clustering, AutoML, designer) was a full domain Classic ML is not a named domain; generative and agentic AI lead
Separate services: Azure AI Vision, Language, Speech, Document Intelligence One platform: Microsoft Foundry and Foundry Tools (formerly Azure AI Services); Content Understanding for extraction from documents, images, audio and video
No code "Create a lightweight chat client application by using the Foundry SDK": expect to read short Python

If you studied for AI-900, keep the responsible AI principles and the workload vocabulary; relearn the product names and add Foundry, agents and Content Understanding.

The habits that pass this exam

  1. Know which workload a scenario describes. Sentiment from reviews is text analysis; turning a call recording into text is speech recognition; pulling the invoice total out of a PDF is information extraction; answering questions in conversation is generative AI; a system that takes actions with tools is an agent.
  2. Match the six responsible AI principles to the concern in the question: fairness, reliability and safety, privacy and security, inclusiveness, transparency, accountability.
  3. Know the Foundry flow end to end: create a project, deploy a model, test it in the playground, write a system prompt, call it from code with the project endpoint, create a single agent with instructions and tools.
  4. Prefer a multimodal model or Content Understanding over stitching specialist services together, unless the scenario asks for a specific capability (for example, a custom voice or real-time transcription).

Study sequence

  1. AI concepts, generative AI basics and responsible AI (the concept domain)
  2. Foundry: projects, model deployment, prompts, SDK, agents
  3. Text analysis and speech
  4. Vision and image generation
  5. Content Understanding for information extraction

Schedule in the practice plan.

Study resources