Microsoft updated the Azure AI Fundamentals certification: AI-901 replaces and expands on AI-900. AI-900 remains valid for holders until its retirement date, but new candidates should prepare for AI-901 which emphasizes practical deployment with Microsoft Foundry, Python integration, and hands-on scenarios.

Quick facts and timeline

AI-901 English release: April 15, 2026.
AI-900 retirement date: June 30, 2026. AI-900 holders retain their credential but new candidates should take AI-901.
Voucher / beta discount: an 80% beta discount has been offered for early candidates; availability and regional exclusions vary. Note: this voucher is not valid in Pakistan.
Check the official exam pages for the most current dates and availability.

Topic comparison: AI-900 (legacy) vs AI-901 (refreshed)

Topic	AI-900 (legacy)	AI-901 (refreshed)
Core AI concepts	Fundamentals: supervised/unsupervised learning, model evaluation metrics, basic ML lifecycle.	Same fundamentals but updated examples and emphasis on applying concepts in real deployments.
Computer vision	Image classification, object detection, common use cases and service overviews.	Practical pipelines: image preprocessing, Foundry deployment patterns, inference at scale.
Natural language processing	Text classification, entity recognition, sentiment analysis, LLM basics.	LLM usage patterns, prompt design, retrieval-augmented generation, Foundry orchestration for language flows.
Generative AI	Concepts and ethical considerations; high-level service descriptions.	Generative model workflows, safety and guardrails, evaluation of outputs, Foundry integration for multi-step generation.
Platform focus	Overview of Azure AI services and low-code options.	Microsoft Foundry as a primary focus: model deployment, single-agent solutions, orchestration, monitoring.
Developer skills	Conceptual understanding; low-code examples.	Hands-on Python integration, SDK usage, sample code for calling services and automating Foundry flows.
Hands-on emphasis	Minimal practical labs.	Significant practical scenarios: deploy model, create Foundry flow, integrate with a client app.
Audience	Non-technical and technical beginners.	Entry-level developers and technical learners who will build or integrate AI apps.

Detailed topic breakdown for AI-901 (what to study)

Foundry and deployment

Foundry concepts: agents, flows, connectors, orchestration patterns.
Deployment: packaging models, versioning, environment configuration, CI/CD basics for Foundry artifacts.
Monitoring and observability: telemetry, logging, performance metrics, cost considerations.

Python integration and SDKs

SDK usage: calling Azure AI services from Python, authentication patterns, error handling.
Sample tasks: text generation, image inference, document extraction via Python scripts.
End-to-end: build a small client app that calls a Foundry endpoint or an Azure AI service.

Generative AI and LLMs

Prompt engineering basics and prompt templates.
Retrieval-augmented generation (RAG) patterns and vector stores.
Safety, hallucination mitigation, and responsible AI principles.

Vision, Speech, and Document Intelligence

Image preprocessing, common model outputs, and evaluation metrics.
Speech-to-text and text-to-speech basics and integration scenarios.
Document extraction, OCR, structured data extraction and validation.

Responsible AI and governance

Bias identification and mitigation strategies.
Privacy, data handling, and compliance considerations for AI solutions.
Explainability and user-facing transparency patterns.

4-week practical study plan (detailed)

Week 1 — Foundations and core concepts

Complete Microsoft Learn modules on AI fundamentals: ML lifecycle, model evaluation, and responsible AI.
Read concise summaries of vision, language, and speech workloads.
Take short quizzes to verify conceptual understanding.

Week 2 — Azure AI services and hands-on labs

Work through labs for Vision, Language, Speech, and Document Intelligence.
Deploy a prebuilt model or use a managed service endpoint for inference.
Document one end-to-end example for your portfolio (repo or notebook).

Week 3 — Microsoft Foundry and deployment scenarios

Follow Foundry tutorials: create a simple agent or flow, connect a model, and run test inputs.
Practice orchestration: chain a retrieval step with a generation step and add basic validation.
Capture screenshots and code snippets for LinkedIn or portfolio posts.

Week 4 — Python integration, mock exams, and review

Write Python scripts that call Azure AI endpoints and Foundry flows; handle auth and errors.
Run timed practice tests and use the exam sandbox to get comfortable with the interface.
Review responsible AI topics and common scenario-based questions.

Short exam prep checklist

Understand Foundry architecture and common deployment patterns.
Be able to read and reason about short Python snippets that call AI services.
Know core ML concepts and evaluation metrics (accuracy, precision, recall, F1, ROC/AUC).
Practice scenario-based questions: choose the right service or pattern for a given requirement.
Review responsible AI: bias mitigation, privacy, and explainability.

AI-901 Study Resources — Topics and Labs

Official exam pages

Core learning paths and overview

Microsoft Learn: Azure AI Fundamentals learning path — end-to-end modules covering fundamentals, workloads, and responsible AI.
Azure AI Services overview — catalog and conceptual docs for vision, language, speech, and applied AI services.

Microsoft Foundry (deployment, agents, orchestration)

Microsoft Foundry documentation and quickstarts — Foundry concepts, agents, flows, and deployment patterns.
Microsoft GitHub (search Foundry samples) — search for Foundry sample repos and workshops on GitHub.

Generative AI and Azure OpenAI

Azure OpenAI Service documentation — API reference, quickstarts, and examples for chat and completions.
Architecture guidance for generative AI — patterns such as RAG, prompt design, and safety considerations.

Python SDKs and code samples

Azure SDK for Python — overview and samples — authentication, client libraries, and examples.
Azure SDK for Python on GitHub — sample code and quickstarts for Cognitive Services and OpenAI clients.

Vision, Speech, and Document Intelligence labs

Computer Vision documentation and quickstarts — image analysis, object detection, and SDK examples.
Speech Service quickstarts — speech-to-text, text-to-speech, and real-time scenarios.
Document Intelligence / Form Recognizer quickstarts — OCR, form extraction, and structured data extraction labs.

Responsible AI and governance

Responsible AI in Azure Machine Learning — principles, fairness, interpretability, and mitigation strategies.
Microsoft Learn module: Responsible AI principles — short, practical module for exam-relevant topics.

Hands-on labs, workshops and sample repos

Azure-Samples on GitHub — search for sample projects: OpenAI, Cognitive Services, Form Recognizer, and Foundry workshops.
Microsoft AI GitHub org — community labs and reference implementations for generative AI and RAG patterns.
Microsoft Learn workshop: Build AI solutions with Azure OpenAI — hands-on exercises and code samples.