AI-901 prep
For people starting out in AI solution development who want to prove they understand AI concepts and can build simple generative AI, agent, speech, vision and extraction apps with Microsoft Foundry and basic Python.
The real exam
| Questions | Typically 40 to 60 (our mock uses 45) |
|---|---|
| Time | 45 minutes of exam time |
| Pass mark | 700 on a 1 to 1000 scale |
| Certification | Microsoft Certified: Azure AI Fundamentals |
| Replaces | AI-900, retired 30 June 2026 |
| Skills outline | Skills measured as of 15 April 2026 |
| Question types | Single and multiple choice, Yes/No statements, matching, ordering, code reading |
| Open book | No Microsoft Learn access on Fundamentals exams |
| Validity | Fundamentals certifications do not expire |
What you get
Mocks follow the real rules: 45 questions in 45 minutes, flag and review.
Syllabus
Fairness, reliability and safety, privacy and security, inclusiveness, transparency and accountability. Part of Identify AI concepts and capabilities (40 to 45%).
How generative models work, choosing a model by capability, deployment types and parameters. Part of Identify AI concepts and capabilities (40 to 45%).
Generative and agentic AI, text analysis, speech, computer vision, image generation and information extraction scenarios. Part of Identify AI concepts and capabilities (40 to 45%).
System and user prompts, deploying and testing models in the portal, chat clients with the Foundry SDK, single agents and agent client apps. Part of Implement AI solutions by using Microsoft Foundry (55 to 60%).
Text analysis apps, spoken prompts to multimodal models and apps using Azure Speech in Foundry Tools. Part of Implement AI solutions by using Microsoft Foundry (55 to 60%).
Visual input to multimodal models, generating images and lightweight vision apps. Part of Implement AI solutions by using Microsoft Foundry (55 to 60%).
Azure Content Understanding in Foundry Tools for documents, forms, images, audio and video. Part of Implement AI solutions by using Microsoft Foundry (55 to 60%).
Course outline
- Responsible AI · 4 lessons
- How generative AI models work · 5 lessons
- Text, speech, vision and extraction workloads · 4 lessons
- Microsoft Foundry and prompting · 4 lessons
- Chat clients and agents · 4 lessons
- Text, speech, vision and extraction apps · 5 lessons
- Weak spots · 6 lessons
- Worked scenarios · 3 lessons
- Exam day · 1 lessons
Try a question
A team removes the gender column from its hiring data set, but the model still favours one gender. What is the most likely cause?A. The model was deployed to the wrong regionB. The temperature parameter is set too highC. Other features act as proxies for genderD. The training data set was encrypted at restShow answer
Harbourline Clinics wants an AI assistant that suggests triage levels. The team must make sure the assistant behaves predictably when it receives unusual or incomplete symptom descriptions. Which principle does this requirement address?A. TransparencyB. InclusivenessC. AccountabilityD. Reliability and safetyShow answer
Passwell is an independent provider and is not affiliated with, sponsored or endorsed by Microsoft. Microsoft, Azure, Microsoft Entra and Microsoft Foundry are trademarks of the Microsoft group of companies. Practice questions are original and are not actual exam questions. Facts checked 2026-11-14.