Microsoft Certified – Azure AI Fundamentals (AI-901)

Learning plan

About 22 hours of focused study, split into 4 phases. Pick your weekly pace and the timeline adjusts.

Estimated finish: 3 weeks

What the exam covers

Source: official exam guide. Weights can change, so check before you start.

Before you start

• An Azure free account (or a Microsoft Learn sandbox) so you can create a Foundry project and deploy a model
• Enough Python to read a short script that imports an SDK, sets an endpoint and prints a response. AI-901 expects this, unlike the old AI-900

1. Week 1 · 6 h

   AI concepts and responsible AI

   • Name Microsoft's six responsible AI principles (fairness, reliability and safety, privacy and security, inclusiveness, transparency, accountability) and give a design consideration for each
   • Explain how a generative model produces text (tokens, embeddings, transformer attention, context window) and why it can fabricate answers
   • Match workloads to scenarios - generative vs agentic AI, keyword extraction, entity detection, sentiment, summarization, speech recognition and synthesis, OCR and image generation
   • Pick a model from the Foundry model catalog based on modality, size and cost, and explain deployment options and parameters such as temperature and max tokens

   Use: AI-901 study guide (skills measured) · AI-901T00 Introduction to AI in Azure (free self-paced paths)

2. Weeks 1–2 · 6 h

   Generative AI apps and agents in Foundry

   • Deploy a chat model in the Foundry portal and test it in the playground with a system prompt and few-shot user examples
   • Read and run a minimal Python chat client that uses the Foundry SDK and a project endpoint
   • Create a single agent in the Foundry portal with instructions and a tool, then call it from a small client app

   Hands-on: Create a Foundry project, deploy a small chat model, write a system prompt that answers only questions about a product FAQ, then run the SDK quickstart from Python and change the prompt until off-topic questions are refused.

   Use: Microsoft Foundry documentation · Foundry SDK quickstart · MicrosoftLearning AI-901 lab repo

3. Week 2 · 6 h

   Speech, vision and information extraction

   • Build a lightweight text-analysis app (sentiment, entities, summary) and send a spoken prompt to a deployed multimodal model
   • Use Azure Speech in Foundry Tools for speech-to-text and text-to-speech
   • Send an image to a multimodal model for description, and generate an image from a text prompt
   • Use Azure Content Understanding to extract fields from a form, text from an image and a transcript from audio or video

   Hands-on: In Content Understanding, build an analyzer for a scanned invoice that returns vendor, date and total as fields. Then run the same invoice image through a multimodal chat model and compare the two outputs.

   Use: Azure Speech documentation · Azure Content Understanding overview

4. Weeks 2–3 · 4 h

   Practice assessment and review

   • Take the free AI-901 practice assessment on AI Skills Navigator and note every miss by skill area
   • Re-read the Foundry half of the study guide, since it carries 55–60% of the exam
   • Walk through the exam sandbox so the question types (drag and drop, hot area, build list) hold no surprises

   Use: AI Skills Navigator practice assessment · Microsoft exam sandbox

You're ready to book when…

• You can deploy a model, test it in the playground and call it from Python without following the docs step by step
• You can say which Foundry tool (Speech, Content Understanding, a multimodal model) fits a given text, audio, image or document scenario
• You can state all six responsible AI principles and a concrete risk for each
• You score 80% or higher on the official practice assessment twice in a row

Exam tips

• Fundamentals exams give you 45 minutes and, unlike associate exams, no access to Microsoft Learn during the test.
• Questions can show short Python snippets, so expect to read SDK calls rather than just name services.
• If you fail you can retake after 24 hours; later retakes have longer waits under the Microsoft retake policy.