What AI-901 Actually Tests
Level Up Team
This post summarizes What AI-901 Actually Tests, module 1.1 of Level Up’s Azure AI Fundamentals series. The lesson is on YouTube as Azure AI Fundamentals AI-901 Exam: 1.1: What AI-901 Actually Tests.
The exam changed in April 2026
AI-901 now tests understanding plus lightweight implementation.
Before April 2026 it was a broad survey of Azure AI services: primarily conceptual, and split across several workload domains. After April 2026 there are two consolidated exam domains. Microsoft Foundry sits at the center. Python and client-application concepts matter, and generative and agentic AI receive greater emphasis.
Two domains cover the exam
| Domain | Weight | Main question |
|---|---|---|
| Identify AI concepts and capabilities | 40–45% | What kind of AI capability fits this need? |
| Implement AI solutions using Microsoft Foundry | 55–60% | How would you create or use the solution? |
Those are published ranges, not an exact question count. The midpoint of each range is about 42.5% concepts and 57.5% implementation. More than half of the exam concerns implementation using Microsoft Foundry.
What “Fundamentals” means now
You should be able to:
- Explain what an AI capability does
- Choose an appropriate model or service
- Recognize the main steps in a Foundry workflow
- Interpret a short implementation example
- Identify the correct tool, endpoint, or configuration
- Apply responsible AI principles to a scenario
The exam does not ask you to train foundation models from scratch, derive machine learning algorithms, design production infrastructure, or write a complete application during the exam.
Two levels of knowledge
Concept questions ask what something means. Implementation questions ask how it fits into a working solution.
| Conceptual knowledge | Implementation knowledge |
|---|---|
| Explain what grounding does | Select grounding for an answer that must use company data |
| Recognize speech synthesis | Choose the capability that returns spoken audio |
| Describe an AI agent | Identify the components needed for a single-agent solution |
| Explain transparency | Choose a design that communicates AI use and limitations |
| Recognize information extraction | Select Content Understanding for documents, images, audio, or video |
Familiarity, not mastery
The exam can ask which code, command, or API pattern correctly performs a task. The expected depth is recognition, not production engineering.
| Technology | Expected familiarity |
|---|---|
| Python | Read basic syntax, variables, function calls, objects, and returned values |
| Azure resources | Understand resources, deployments, endpoints, keys, and authentication context |
| REST APIs | Recognize requests, endpoints, headers, payloads, and responses |
| SDKs | Understand how an application creates a client and calls a service |
| CLIs | Recognize command-based resource or deployment operations |
Definitions tell you what. Scenarios ask you to choose.
A definition question tests direct recall: “Which AI workload converts spoken language into text?”
A scenario question tests whether you can apply that recall: “A call center needs searchable transcripts of recorded conversations. Which capability should the team use?” To answer it, identify the required output, ignore irrelevant details, match the requirement to a capability, and distinguish it from plausible alternatives.
For every topic in the course, learn three things: what the capability does, when to use it, and how Microsoft Foundry exposes it.
Bring this to your team
Level Up can run a personalized Azure AI Fundamentals workshop for your team. Tell us a little about your group and we’ll reply with pricing.
Loading the form…