Azure AI-900 exam guide covering AI workload concepts, vision, language, generative AI, and responsible AI decisions.
This guide targets Microsoft Azure AI Fundamentals (AI-900) at the exact level Microsoft currently tests: workload recognition, responsible-AI judgment, core machine-learning vocabulary, and high-level Azure service fit. On April 13, 2026, Microsoft Learn says the certification is still active, the exam is scheduled to retire on June 30, 2026, the successor will be AI-901, and the platform naming is shifting from Azure AI Foundry to Microsoft Foundry. For exam prep, use the current AI-900 study-guide wording first and treat the Foundry rename as a branding transition, not a different objective set.
AI workload: The primary kind of problem a system is solving, such as image analysis, speech recognition, document extraction, text analysis, or content generation.
Service fit: Choosing the Azure capability that matches the problem class the question is really asking about.
At a glance
Exam fact
Current official signal
Level
Beginner
Certification page last updated
May 2, 2025
Time limit
45 minutes
Delivery
Proctored exam, with possible interactive components
Retirement date
June 30, 2026
Replacement note
Microsoft says AI-901 will replace AI-900 after retirement
Skills measured
AI workloads and considerations; machine learning fundamentals; computer vision; NLP; generative AI
The right answer on AI-900 is usually not the most technical-sounding one. It is the answer that keeps the correct abstraction level. First classify the workload. Then choose the concept or Azure service family that fits it. Only after that should you care about closer distractors.
How to use this guide
Start with the study plan if you want a structured route through the five official skill areas.
Work through the chapter map in order, because the later chapters assume you can already separate workloads, ML basics, and service families.
Use the cheat sheet for last-mile recall and confusion-pair cleanup.
Use the faq for retirement timing, Foundry rename context, and exam-day judgment.
Use the resources page to re-check live Microsoft Learn facts before you book or sit the exam.
Use the glossary only when terms blur together; do not let it replace the actual lessons.
Blueprint-aligned chapter map
The current Microsoft study guide measures skills across five domains. This guide follows that structure directly.
Skill area
Weight
Chapter
Start here
Describe Artificial Intelligence workloads and considerations
flowchart LR
A["1. Workload classification"] --> B["2. ML fundamentals"]
B --> C["3. Vision and NLP service fit"]
C --> D["4. Generative AI on Azure"]
D --> E["Cheat sheet, FAQ, and resources"]
What this guide assumes
AI-900 does not assume you are already building production ML systems. It assumes you can:
recognize what kind of problem a scenario describes
separate predictive ML from prebuilt AI services and from generative AI
identify the responsible-AI principle most directly at risk
keep Microsoft service categories straight at fundamentals depth
What strong answers usually do
classify the scenario before naming a product
choose the broad service family or concept that solves the problem cleanly
reject generative-AI answers when the need is really extraction, classification, transcription, or search
keep document processing separate from generic image analysis when the clue is about forms, receipts, or structured files
remember that AI-900 uses current Microsoft terminology but still tests fundamentals, not engineering implementation detail
Best fit for this guide
If you are coming from…
Bias your review toward…
business, analyst, or operations roles
workload classification, responsible AI, and service-boundary decisions
help desk, support, or cloud-fundamentals background
ML vocabulary, Azure AI service fit, and common confusion pairs
future Azure AI Engineer path
fundamentals now, then deeper build workflows after you can classify the problem correctly
Before you schedule the exam
confirm the live Microsoft Learn status on the resources page, especially if your target date is near or after June 30, 2026
use the study plan to narrow weak areas instead of rereading everything evenly
keep the cheat sheet nearby for final review, but do the actual learning in the chapter lessons first