AI-901 vs AI-900: What Changed in the New Microsoft Azure AI Fundamentals Exam?
Microsoft has made a major update to its entry-level Azure AI certification path. The AI-900: Microsoft Azure AI Fundamentals exam retired on June 30, 2026, and candidates pursuing the Microsoft Certified: Azure AI Fundamentals certification should now focus on AI-901.
Although both exams are associated with Azure AI Fundamentals, AI-901 is more than a simple exam-code update. The new exam reflects how Microsoft’s AI ecosystem has changed, placing much greater emphasis on Microsoft Foundry, generative AI, AI agents, multimodal AI, and actually implementing AI solutions.
So, what changed from AI-900 to AI-901, and what should candidates study differently?
Is AI-901 the Replacement for AI-900?
Yes. AI-900 is now retired, while AI-901 is the current exam associated with Microsoft Certified: Azure AI Fundamentals. Microsoft’s retired-exam list confirms that AI-900 reached retirement on June 30, 2026, while the current AI-901 exam page lists no retirement date.
The important point for new candidates is simple:
If you are starting your Azure AI Fundamentals preparation now, study for AI-901 rather than AI-900.
Candidates who previously prepared for AI-900 should not assume that all of their existing preparation transfers directly to the new exam. Many foundational AI concepts remain relevant, but the emphasis has changed substantially.
AI-901 vs AI-900: What Is the Biggest Difference?
The biggest change is the move from primarily understanding AI concepts and Azure AI workloads toward a combination of conceptual understanding and hands-on AI implementation.
The current AI-901 blueprint is divided into only two major areas:
| AI-901 Skills Measured | Weight |
|---|---|
| Identify AI concepts and capabilities | 40–45% |
| Implement AI solutions by using Microsoft Foundry | 55–60% |
In other words, more than half of AI-901 focuses on implementing AI solutions with Microsoft Foundry.
That is an important distinction for anyone approaching AI-901 as if it were simply a renamed AI-900.
Why Does Microsoft Foundry Matter So Much in AI-901?
Microsoft Foundry sits at the center of the new exam.
Candidates are expected to understand how Foundry can be used to work with AI models and create practical AI solutions. The official objectives include deploying and interacting with models, creating lightweight applications with the Foundry SDK, creating and testing single-agent solutions, and building client applications for agents.
This means your preparation should go beyond questions such as:
“What is generative AI?”
You should also be comfortable with questions such as:
“How would I use Microsoft Foundry to implement this AI solution?”
That difference captures much of the shift from AI-900 to AI-901.
Does AI-901 Include AI Agents?
Yes, and this is one of the most notable additions for candidates familiar with the older Azure AI Fundamentals approach.
Microsoft explicitly includes generative and agentic AI among the workloads candidates should recognize. The implementation objectives also include creating and testing a single-agent solution in the Foundry portal and creating a lightweight client application for an agent.
This reflects the broader evolution of AI applications.
Modern AI solutions increasingly involve systems that can interact with models, instructions, data, tools, and other resources instead of simply sending a prompt to a language model and displaying the response.
For AI-901 candidates, understanding the basic role of AI agents is therefore important.
What AI Concepts Are Still Important?
The move toward implementation does not mean fundamentals have disappeared.
AI concepts and capabilities still account for 40–45% of AI-901. Candidates should understand responsible AI principles such as fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability.
You should also be able to recognize different AI workloads and capabilities, including:
- Generative and agentic AI
- Text analysis
- Speech
- Computer vision
- Image generation
- Information extraction
AI-901 therefore still expects a broad understanding of artificial intelligence. The difference is that conceptual knowledge now sits alongside a much larger implementation component.
Is There More Generative AI in AI-901?
Yes. Generative AI has a much more central role.
Candidates should understand how generative AI models work, how to select an appropriate model based on its capabilities, and how deployment options and configuration parameters affect a solution.
Prompting is also explicitly included.
The AI-901 objectives expect candidates to understand how to create effective system and user prompts for generative AI models.
For preparation purposes, this means generative AI should not be treated as one small topic near the end of your study plan. It is connected to multiple parts of the new exam.
What About Computer Vision, Speech, and NLP?
These technologies are still relevant, but AI-901 approaches them in a more implementation-oriented way.
For example, candidates may need to understand how to:
- Build a lightweight application that includes text analysis.
- Respond to spoken prompts with a deployed multimodal model.
- Work with Azure Speech in Foundry Tools.
- Interpret visual input with a multimodal model.
- Generate visual outputs.
- Build applications that include vision capabilities.
This is another useful way to understand the AI-900 → AI-901 transition:
Knowing what an AI service does is no longer enough; candidates should also understand how that capability can be incorporated into an AI solution.
Is Information Extraction Important for AI-901?
Yes. Information extraction has a dedicated place in the current objectives.
Microsoft specifically references Azure Content Understanding in Foundry Tools for extracting information from documents, forms, images, audio, and video. Candidates should also understand how these capabilities can be incorporated into lightweight applications.
This gives AI-901 a noticeably more multimodal character.
AI is no longer presented simply as text, vision, and speech concepts operating independently. Candidates increasingly need to understand how different forms of information can become part of an application.
Do You Need Python for AI-901?
This is another area candidates should pay attention to.
Microsoft’s current audience profile states that AI-901 candidates should have knowledge of Python coding syntax and programming techniques, along with familiarity with Azure resources. The official page also notes that candidates should be familiar with REST APIs, SDKs, and CLIs.
That does not mean AI-901 suddenly becomes an advanced programming exam.
However, candidates coming from AI-900 preparation should not assume that purely conceptual knowledge will be sufficient. Basic technical familiarity is much more useful for the new exam.
How Should Former AI-900 Candidates Prepare for AI-901?
If you already studied AI-900, you do not need to discard everything you learned.
Concepts such as responsible AI, NLP, computer vision, speech, generative AI, and general Azure AI workloads still provide a useful foundation.
The better strategy is to build on that foundation.
Spend additional time on:
Microsoft Foundry → Generative AI → AI Agents → Prompting → Multimodal AI → Content Understanding → Lightweight AI Application Implementation
Most importantly, practice connecting concepts to scenarios.
Instead of only asking:
What is this technology?
also ask:
When would I use it, and how would I implement it in Microsoft Foundry?
That mindset is much closer to the current AI-901 objectives.
How Should You Use AI-901 Practice Questions?
AI-901 practice questions can be especially useful for identifying whether you are still approaching the exam with an AI-900 mindset.
When reviewing each question, determine whether it tests:
Concept recognition or solution implementation.
If you consistently perform well on definitions but struggle with Foundry scenarios, agent implementations, multimodal applications, or selecting the appropriate capability for a solution, that is a strong indication of where additional study is needed.
A useful preparation cycle is:
Review the Objectives → Learn the Concept → Try the Technology → Practice Questions → Analyze Mistakes
Certspots AI-901 practice questions can then serve as an additional way to test your understanding of the current exam topics and become more comfortable with scenario-based questions.
Is AI-901 Harder Than AI-900?
It is more useful to describe AI-901 as more implementation-oriented rather than simply labeling it harder.
Candidates who prefer conceptual learning may find the increased emphasis on Foundry, Python familiarity, SDKs, AI agents, and lightweight application development more demanding.
Candidates who already experiment with generative AI applications may find many of the newer objectives more intuitive.
Either way, the preparation strategy needs to change.
The old question was largely:
“Do I understand the fundamentals of AI on Azure?”
AI-901 increasingly adds another question:
“Can I understand how those AI capabilities are used to build a solution?”
What Should You Remember About the AI-900 to AI-901 Transition?
The transition from AI-900 to AI-901 reflects a broader change in what “AI fundamentals” means in 2026.
AI fundamentals are no longer limited to recognizing machine learning, NLP, computer vision, and responsible AI concepts. Microsoft now expects entry-level candidates to have foundational knowledge of generative AI, agents, multimodal AI, Microsoft Foundry, and practical AI implementation.
For anyone beginning preparation now, the direction is clear: focus on AI-901 and use the current AI-901 objectives rather than older AI-900 study plans.
LEAVE A COMMENT