What is covered on the AI-900 exam?
The AI-900 Microsoft Azure AI Fundamentals exam covers five areas: AI workloads and considerations, machine learning fundamentals, computer vision, natural language processing, and generative AI. It is an entry-level exam requiring no prior AI or programming experience. The exam costs $99 USD and requires a passing score of 700 out of 1000.
The AI-900 Microsoft Azure AI Fundamentals certification validates foundational knowledge of artificial intelligence concepts and Microsoft Azure AI services. It is the entry-level credential for AI on Azure, designed for anyone wanting to understand how AI works and how Microsoft has implemented AI capabilities in its cloud platform.
As artificial intelligence becomes embedded in every industry, AI literacy has become a competitive differentiator. The AI-900 provides structured education about AI concepts accessible to professionals without programming or data science backgrounds. The exam costs $99 USD, contains 40-60 questions, and requires a passing score of 700 out of 1000.
Exam Overview
| Detail | Information |
|---|---|
| Exam Code | AI-900 |
| Full Name | Microsoft Azure AI Fundamentals |
| Number of Questions | 40-60 |
| Time Limit | 45 minutes |
| Passing Score | 700/1000 |
| Cost | $99 USD |
| Prerequisites | None |
| Related Advanced Exams | AI-102, DP-100 |
The exam covers five domains:
- Describe AI workloads and considerations (15-20%)
- Describe fundamental principles of machine learning on Azure (20-25%)
- Describe features of computer vision workloads on Azure (15-20%)
- Describe features of NLP workloads on Azure (15-20%)
- Describe features of generative AI workloads on Azure (15-20%)
"AI-900 was created because organizations were adopting AI tools without their workforce understanding what AI can and cannot do. It provides a common vocabulary and realistic understanding of AI capabilities, limitations, and ethical considerations. Even non-technical employees benefit from this foundation." -- Microsoft AI learning team guidance
Domain 1: AI Workloads and Considerations (15-20%)
Types of AI Workloads
Machine learning uses algorithms that learn from data to make predictions or decisions without explicit programming. Common workloads:
- Predictive analytics: Forecasting future values (sales forecasting, demand prediction)
- Classification: Categorizing data into discrete classes (spam detection, disease diagnosis)
- Clustering: Grouping similar data points without predefined categories (customer segmentation)
- Anomaly detection: Identifying unusual patterns (fraud detection, equipment failure prediction)
Computer vision enables machines to interpret and understand visual information:
- Image classification, object detection, facial recognition, optical character recognition
Natural language processing (NLP) enables machines to understand and generate human language:
- Text analysis, sentiment analysis, translation, speech recognition, text-to-speech
Responsible AI Principles
Microsoft's Responsible AI framework consists of six principles that the exam tests conceptually:
| Principle | Description |
|---|---|
| Fairness | AI systems should treat all people fairly without bias based on gender, race, or other characteristics |
| Reliability and Safety | AI systems should perform reliably and safely across a range of conditions |
| Privacy and Security | AI systems should respect privacy and maintain data security |
| Inclusiveness | AI should empower everyone and engage people broadly |
| Transparency | AI systems should be understandable; people should know how they work |
| Accountability | People should be accountable for AI systems they develop and deploy |
Domain 2: Machine Learning on Azure (20-25%)
Machine Learning Concepts
Supervised learning uses labeled training data (input-output pairs) to learn a mapping function. Examples: linear regression, decision trees, neural networks.
Unsupervised learning finds patterns in unlabeled data. Examples: clustering (k-means), dimensionality reduction (PCA).
Reinforcement learning learns through trial and error, receiving rewards for correct actions. Examples: game playing, robot navigation.
Azure Machine Learning Studio
Azure Machine Learning (Azure ML) is the platform for building, training, and deploying machine learning models. Key components:
- Designer: Visual drag-and-drop interface for building ML pipelines without code
- Automated ML (AutoML): Automatically tries multiple algorithms and hyperparameters to find the best model
- Notebooks: Jupyter notebook environment for code-based ML development
- Compute: Managed compute instances and clusters for training and inference
- Model registry: Versioned storage for trained models
- Endpoints: Deployment targets for real-time and batch inference
Domain 3: Computer Vision Workloads (15-20%)
Azure AI Vision Services
Azure AI Vision (formerly Computer Vision) provides:
- Image analysis: Detecting objects, describing scenes, extracting text (OCR), identifying colors
- Face detection and recognition: Detecting faces and their attributes, face verification, face identification
- Custom Vision: Training custom image classification and object detection models without ML expertise
- Document Intelligence (formerly Form Recognizer): Extracting structured data from forms, invoices, receipts, and documents
Key concepts the exam tests:
- Object detection vs. image classification: Detection finds objects with bounding boxes; classification assigns the whole image to a category
- OCR (Optical Character Recognition): Extracting printed and handwritten text from images
Domain 4: NLP Workloads on Azure (15-20%)
Azure AI Language Services
Azure AI Language provides:
- Sentiment analysis: Determining whether text expresses positive, negative, or neutral sentiment
- Key phrase extraction: Identifying the most important phrases in a document
- Named entity recognition (NER): Identifying people, places, organizations, dates, and other entities in text
- Language detection: Identifying what language a text is written in
- Translation: Azure AI Translator for translating text across 100+ languages
Azure AI Speech
Azure AI Speech provides:
- Speech-to-text: Converting spoken audio to text transcriptions
- Text-to-speech: Converting text to natural-sounding spoken audio
- Speaker recognition: Identifying or verifying speakers based on voice characteristics
- Speech translation: Real-time speech translation across languages
Language Understanding (CLU)
Conversational Language Understanding (CLU) enables building natural language understanding models that interpret user intent and extract entities from utterances. Used in chatbots and virtual assistants.
Domain 5: Generative AI Workloads (15-20%)
Understanding Generative AI
Generative AI refers to AI models capable of creating new content (text, images, code, audio) by learning patterns from large training datasets. Large language models (LLMs) are the foundation of most generative AI tools.
Key concepts:
- Large language models (LLMs): Neural networks trained on vast text corpora that can generate, summarize, translate, and reason about text
- Tokens: The basic units LLMs process -- approximately 0.75 words per token in English
- Prompts and responses: The input-output interaction model for LLMs
- Grounding: Providing LLMs with specific context or data to constrain responses to factual information
Azure OpenAI Service
Azure OpenAI Service provides access to OpenAI's models (including GPT-4, DALL-E, and Whisper) with Azure enterprise features:
- Enterprise security and compliance
- Private network connectivity
- Responsible AI content filtering
- Usage monitoring and cost controls
Copilot products (Microsoft 365 Copilot, GitHub Copilot, Security Copilot) are built on Azure OpenAI and provide AI assistance integrated into Microsoft applications.
"The generative AI domain was added to AI-900 in late 2023 and now constitutes approximately 15-20% of the exam. Candidates studying from older materials may be underprepared for questions about LLMs, prompt engineering, Azure OpenAI Service, and Copilot. Use the most current Microsoft Learn content." -- AI-900 certification community note
Frequently Asked Questions
Do I need programming experience for AI-900? No programming experience is required for AI-900. The exam tests conceptual understanding of AI types, Azure AI services, and responsible AI principles at a foundational level. You should understand what each service does and when to use it, but you will not be asked to write or debug code.
Is AI-900 worth taking in 2025 with so many AI changes happening? Yes, AI-900 is worth taking because it covers fundamental AI concepts that remain stable even as specific tools evolve. The certification demonstrates structured AI literacy to employers and provides a conceptual foundation for understanding more advanced AI tools. Microsoft updates the exam objectives to include current topics like generative AI and Copilot products.
What is the best study path after AI-900? The two main advanced paths from AI-900 are AI-102 (Azure AI Engineer Associate), which covers building and implementing Azure AI solutions using code, and DP-100 (Azure Data Scientist Associate), which covers machine learning model development. Professionals interested in the business and compliance aspects of AI may also find SC-900 complementary.
References
- Microsoft. (2025). Exam AI-900: Microsoft Azure AI Fundamentals. https://learn.microsoft.com/en-us/credentials/certifications/exams/ai-900/
- Microsoft. (2025). Azure AI Services Documentation. https://learn.microsoft.com/en-us/azure/ai-services/
- Microsoft. (2025). Responsible AI Principles and Approach. https://www.microsoft.com/en-us/ai/responsible-ai
- Microsoft. (2025). Azure OpenAI Service Documentation. https://learn.microsoft.com/en-us/azure/ai-services/openai/
- Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press. https://www.deeplearningbook.org/
- Russell, S., & Norvig, P. (2021). Artificial Intelligence: A Modern Approach (4th ed.). Pearson.