AI-900 to AI-901: What Changed and How to Prepare
Keep the fundamentals that still matter, but rebuild your preparation around Microsoft Foundry and small implementation tasks.
The short answer
Microsoft retired exam AI-900 on June 30, 2026. AI-901 is now the exam for the Microsoft Certified: Azure AI Fundamentals credential. The credential remains; its assessment moved from mostly recognising AI concepts and services toward understanding concepts and carrying out lightweight work in Microsoft Foundry.
Verify the date in Microsoft's official exam retirement list, then use the current AI-901 study guide as your source of truth.
AI-901 has two objective domains
| Current objective | Weight | Preparation consequence |
|---|---|---|
| Identify AI concepts and capabilities | 40–45% | Know responsible AI, model behaviour and configuration, and the fit of generative AI, agents, language, speech, vision, and information extraction. |
| Implement AI solutions by using Microsoft Foundry | 55–60% | Spend more than half your preparation on small Foundry tasks rather than definitions alone. |
The weighting is the clearest reason not to reuse an AI-900-only plan unchanged. AI-901 still tests foundational understanding, but implementation with Foundry now carries the majority of the blueprint.
What changes from an AI-900 study plan
| Preparation area | How to update it for AI-901 |
|---|---|
| Candidate baseline | Add basic Python syntax and programming techniques, plus familiarity with Azure resources. Microsoft includes both in the current audience profile. |
| Generative AI | Move from recognising generative AI to writing effective prompts, selecting a model and deployment, and interacting with it in Foundry. |
| Agents | Create and test a single-agent solution, then understand how a lightweight client communicates with it. |
| Applied AI | Connect text, speech, vision, image generation, and Content Understanding to small solution scenarios. |
| Service memorisation | Replace long product lists with capability selection: choose the simplest tool that satisfies the input, output, safety, and integration requirements. |
What still transfers
- The six responsible AI principles and the ability to apply them to a scenario.
- Recognising language, speech, computer vision, and information-extraction workloads.
- Understanding training versus inference and the role of model inputs, outputs, and configuration.
- Choosing an Azure AI capability from a business requirement.
Keep these concepts, but verify terminology against the current guide. An older course may use retired portals, service names, or workflows even when the underlying idea remains useful.
A focused preparation sequence
Map your existing knowledge
Mark every AI-901 bullet as strong, weak, or untouched. Do not estimate readiness from an AI-900 score because the objective balance has changed.
Build one Foundry mini-project
Provision the required Azure resources, open a Foundry project, deploy a suitable model, test prompts, and call it from a short Python client. Keep the project small enough that you can explain every component.
Add an agent and one applied capability
Create and test a single agent, then complete one focused task with text, speech, vision, image generation, or Content Understanding. Explain why that capability fits the scenario.
Test, review, repeat
Use scenario practice and Microsoft's free Practice Assessment to find weak objectives. Review the documentation behind every miss and repeat the corresponding task in Foundry.
AI-901 study resources worth using
- The official AI-901 study guide for the current objective wording and weights.
- Microsoft's free AI-901 Practice Assessment in AI Skills Navigator; sign-in is required.
- Microsoft Foundry training for the hands-on modules behind the 55–60% implementation domain.
Avoid these transition mistakes
- Booking AI-900 material under a new exam code without checking the AI-901 blueprint.
- Treating “fundamentals” as a no-code exam when the audience profile explicitly expects basic Python.
- Learning a portal walkthrough by position instead of understanding the resource and outcome.
- Using recalled exam content instead of current documentation and original scenarios.
- Ignoring the 55–60% Foundry implementation domain until the final week.
Check your AI-901 baseline
Start with ten original scenarios, then review every miss and its explanation to decide what to build or study next.
Start free AI-901 practiceAfter the fundamentals
Review the complete AI-901 practice exam overview. If your goal is to build production-oriented apps and agents, compare the next step in the AI-103 study plan.