Azure AI-900 30-, 60-, and 90-day study plan with topic order, review loops, and final-week priorities.
Use this study plan when you want a disciplined path through AI-900 instead of studying Azure AI services as disconnected buzzwords. The exam gets easier when you build the stack in order: classify the workload, separate the ML vocabulary, learn the service boundaries, then finish with generative-AI judgment and current Microsoft Learn facts.
Miss log: A short record of what you misunderstood, why the distractor looked tempting, and what rule would have prevented the miss.
| If your background is… | Best plan |
|---|---|
| non-technical or business-first | use the full four-week track |
| cloud-fundamentals learner with some Azure exposure | use the four-week track and compress only after your misses get narrow |
| already comfortable with AI workload categories | use the ten-day compression track |
| Week | Focus | Main pages |
|---|---|---|
| 1 | classify workloads and learn the six responsible-AI principles | 1. AI Workloads and Responsible AI, 1.1, 1.2 |
| 2 | lock in ML vocabulary and Azure Machine Learning basics | 2. Machine Learning Fundamentals on Azure, 2.1, 2.2, 2.3 |
| 3 | separate vision workloads from NLP and speech workloads cleanly | 3. Computer Vision on Azure, 4. Natural Language Processing on Azure |
| 4 | finish with generative AI, Microsoft Foundry naming context, and final review | 5. Generative AI on Azure, cheat sheet, faq, resources |
flowchart LR
R["Read one lesson"] --> N["Write 3 to 5 notes"]
N --> Q["Do the lesson quiz"]
Q --> M["Log misses as rules"]
M --> C["Revisit the matching cheat-sheet lane"]
C --> X["Run mixed scenario review"]
| If your misses cluster around… | Go back here first |
|---|---|
| what kind of problem the scenario is really describing | 1.1 Workload Classification and Scenario Fit |
| fairness, transparency, accountability, or human review | 1.2 Responsible AI Principles and Risk Signals |
| regression vs classification vs clustering | 2.1 Regression, Classification, Clustering and Deep Learning |
| features, labels, training, validation, or overfitting | 2.2 Features, Labels, Training and Validation |
| prebuilt AI service vs Azure Machine Learning | 2.3 Azure Machine Learning Capabilities |
| image analysis vs document extraction vs face analysis | 3.1 Image Classification, Object Detection, OCR and Face Analysis |
| speech vs text-analysis vs translation | 4.1 Text Analysis, Speech, Translation and Language Modeling |
| generative AI vs extraction, classification, or search | 5.1 Generative AI Scenarios, Grounding and Responsible Use |
| current official facts, retirement, or naming changes | FAQ and Resources |
AI-900 instead of AI-901.AI-900 study-guide wording for objective coverage and treat the naming difference as a current transition note.