AI-901 — Microsoft Azure AI Fundamentals Study Plan
A practical 7-, 14-, 30-, and 60/90-day study plan for the Microsoft Azure AI Fundamentals (AI-901) exam.
How to use this AI-901 study plan
This study plan is for candidates preparing for the real Microsoft Azure AI Fundamentals (AI-901) exam from Microsoft. The AI-901 exam is a fundamentals-level exam, so your preparation should focus on recognizing AI workloads, choosing appropriate Azure AI services, understanding responsible AI concepts, and answering short scenario-based questions accurately.
Use this page to turn your available time into a practical schedule. If you are already close to exam day, start with the 7-day or 14-day path. If you are starting earlier, use the 30-day or 60/90-day path and build in more review, practice, and light hands-on familiarity.
Which plan should you use?
| Time available | Best plan | Weekly time target | Use this if | Main goal |
|---|---|---|---|---|
| 7 days | Final review sprint | 8-12 hours total | You already studied or have Azure/AI exposure | Patch weak areas and build exam timing |
| 14 days | Focused plan | 10-16 hours total | You know basic cloud terms but need structured AI-901 review | Cover all major objective areas once, then review misses |
| 30 days | Balanced plan | 3-5 hours per week | You are new to Azure AI or want low-stress preparation | Learn, drill, review, and complete timed practice |
| 60/90 days | Full preparation path | 2-4 hours per week | You are new to both AI concepts and Azure services | Build durable understanding with spaced review |
If you are unsure, take a short diagnostic practice set first. Your plan should be based on missed-question patterns, not on how many pages or videos you have completed.
What AI-901 preparation should cover
Use these study blocks throughout your plan. The exam is conceptual, but the questions often test whether you can identify the right AI workload, feature, or Azure service from a scenario.
| Study block | You should be able to answer | Practice actions |
|---|---|---|
| AI workloads and responsible AI | What type of AI workload is described? What responsible AI principle applies? | Classify scenarios as prediction, vision, language, speech, search, automation, or generative AI. Review fairness, reliability, privacy, security, inclusiveness, transparency, and accountability. |
| Machine learning fundamentals | Is the problem classification, regression, clustering, or forecasting? What does training/evaluation mean? | Drill basic ML vocabulary: features, labels, datasets, training, validation, inference, model evaluation, supervised vs. unsupervised learning. |
| Azure Machine Learning concepts | When would Azure Machine Learning be used instead of a prebuilt AI service? | Review model training, experimentation, automated ML concepts, notebooks, datasets, endpoints, and responsible model management at a high level. |
| Computer vision and document intelligence | Which service fits image analysis, OCR, object detection, or form/document extraction? | Match scenarios to Azure AI Vision and Azure AI Document Intelligence-style capabilities. Focus on inputs, outputs, and use cases. |
| Natural language, speech, and translation | Which service fits sentiment, key phrase extraction, entity recognition, language detection, speech-to-text, text-to-speech, or translation? | Build a comparison chart for language, speech, translator, and conversational scenarios. |
| Generative AI concepts | What are prompts, completions, grounding, copilots, content filtering, and responsible generative AI practices? | Practice scenario questions involving summarization, content generation, chat, retrieval-augmented responses, and safe deployment. |
| Azure service selection | Which Azure AI service is the best fit for a business requirement? | Use service-selection drills. Avoid memorizing product limits; focus on purpose, workload type, and decision clues. |
| Security, governance, and operations basics | How are AI services accessed and managed at a fundamental level? | Review keys/endpoints, identity concepts, access control, monitoring, cost awareness, and data privacy considerations conceptually. |
Daily practice rhythm
Use the same rhythm on most study days. This keeps preparation active and prevents passive reading from taking over.
45-minute weekday session
| Time | Activity | Output |
|---|---|---|
| 5 min | Quick recall | Write 3-5 facts from yesterday without notes. |
| 15 min | Focused concept review | Review one AI-901 topic only. |
| 15 min | Practice questions | Complete a small set on that topic. |
| 10 min | Missed-question review | Log misses, fix the reason, and schedule a retry. |
75- to 90-minute session
| Time | Activity | Output |
|---|---|---|
| 10 min | Warm-up recall | Rebuild a service-selection table from memory. |
| 25 min | Learn/review | Study one objective area or Azure AI service family. |
| 25 min | Mixed practice | Answer mixed questions without notes. |
| 20-30 min | Review | Analyze every missed and guessed question. |
Weekend or long session
| Time | Activity | Output |
|---|---|---|
| 30 min | Weak-area review | Revisit your highest-error topic. |
| 45-60 min | Timed practice block | Build pacing and accuracy. |
| 30-45 min | Missed-question repair | Update notes and create retest items. |
| 15 min | Plan next sessions | Choose the next 2-3 study targets. |
Diagnostic-first setup
Before starting any plan longer than 7 days, take a diagnostic practice set.
| Step | What to do | Why it matters |
|---|---|---|
| 1 | Answer a mixed set without notes | Reveals actual readiness, not perceived familiarity |
| 2 | Mark each answer as known, guessed, or unsure | Separates lucky answers from stable knowledge |
| 3 | Categorize every miss | Shows whether the issue is concepts, service selection, vocabulary, or question reading |
| 4 | Build your weak-area list | Determines what goes into the next study sessions |
| 5 | Retest later with new questions | Confirms improvement instead of memorization |
Do not spend your first week only reading. For AI-901, you need repeated exposure to short scenario questions: “Which service?”, “Which workload?”, “Which principle?”, and “Which feature?”
Missed-question review method
Every missed or guessed question should produce a repair action. Use a simple log like this:
| Field | What to record | Example |
|---|---|---|
| Topic | The exam area involved | Natural language processing |
| Scenario clue | The phrase that mattered | “Extract key phrases from customer reviews” |
| Your answer | What you selected | Translator |
| Correct idea | What you should have recognized | Language service capability |
| Miss reason | Why you missed it | Confused translation with text analytics |
| Fix | What you will do | Add to service-selection chart and retry 5 similar items |
| Retest date | When to revisit | In 2-3 days |
Common miss categories
| Miss type | What it usually means | Repair action |
|---|---|---|
| Vocabulary miss | You did not know a term | Create a flashcard or one-line definition |
| Service-selection miss | You knew the concept but chose the wrong Azure service | Add the scenario to a comparison chart |
| Responsible AI miss | You confused two principles | Write a short example for each principle |
| ML concept miss | You mixed up classification, regression, or clustering | Drill 10 scenario classifications |
| Question-reading miss | You missed a qualifier like “prebuilt,” “custom,” “speech,” or “document” | Slow down and underline decision words during practice |
| Guess-correct | You got it right but were not sure | Treat it as a miss and review it |
When to use timed mock exams
Timed mock exams are most useful after you have completed at least one pass through the major topics. Do not use all full-length practice too early; you need some timed sets later for readiness checks.
| Plan | First diagnostic | First timed mock | Final timed mock |
|---|---|---|---|
| 7-day plan | Day 1 | Day 3 or 4 | Day 5 or 6 |
| 14-day plan | Day 1 | Day 8-10 | Day 12 or 13 |
| 30-day plan | Day 1-2 | Week 3 | Final week |
| 60/90-day plan | Week 1 | Around midpoint | Final 7-10 days |
Mock exam rules
- Use a quiet setting and complete the set without notes.
- Review every missed and guessed question before taking another mock.
- Track weak areas by category, not just by score.
- Do not rely on repeated exposure to the same questions as evidence of readiness.
- If a mock exposes a weak domain, pause new content and repair that domain first.
7-day final review sprint
Use this plan if your exam is one week away. It assumes you have some familiarity with AI concepts or Azure services. If you are starting from zero, extend to the 14-day plan if possible.
| Day | Focus | Study actions | Output |
|---|---|---|---|
| 7 | Diagnostic and triage | Take a mixed diagnostic set. Categorize misses by topic. | Top 3 weak areas identified |
| 6 | AI workloads, responsible AI, and ML basics | Review workload types, responsible AI principles, classification/regression/clustering, training vs. inference. | One-page concept sheet |
| 5 | Azure AI service selection | Drill Vision, Language, Speech, Translator, Document Intelligence, Azure Machine Learning, Azure AI Search, and generative AI scenarios. | Service-selection chart |
| 4 | Computer vision, documents, language, and speech | Practice scenario questions by workload. Focus on inputs and outputs. | Missed-question repairs |
| 3 | Generative AI and applied scenarios | Review prompts, grounding, copilots, content safety, summarization, chat, and responsible use. Take a timed practice block. | Timing and weak-area data |
| 2 | Final weak-area sprint | Rework your highest-error topics. Review all guessed questions. Do one short mixed set. | Clean weak-area list |
| 1 | Light review only | Review notes, service chart, responsible AI examples, and common miss patterns. Avoid heavy new material. | Calm, concise final review |
7-day rules
- Stop adding brand-new material after Day 3 unless it fixes a repeated miss.
- Spend more time reviewing missed questions than watching new lessons.
- Keep a daily service-selection drill: read a scenario and name the best-fit workload or service.
- On the final day, do not overload. Review patterns, not every detail.
14-day focused plan
Use this if you have two weeks and need a complete but compressed path.
| Day | Focus | Practice target |
|---|---|---|
| 1 | Diagnostic set and plan setup | Mixed questions; build miss log |
| 2 | AI workloads and responsible AI | Workload classification and principle matching |
| 3 | ML fundamentals | Classification, regression, clustering, training, evaluation, inference |
| 4 | Azure Machine Learning overview | When to use custom ML vs. prebuilt AI services |
| 5 | Computer vision | Image analysis, object detection, OCR-style scenarios |
| 6 | Document intelligence and extraction | Forms, documents, structured extraction scenarios |
| 7 | Review checkpoint | Mixed practice; repair Days 2-6 misses |
| 8 | Natural language processing | Sentiment, key phrases, entities, language detection |
| 9 | Speech and translation | Speech-to-text, text-to-speech, translation, conversation scenarios |
| 10 | Generative AI | Prompts, chat, summarization, grounding, copilots, responsible use |
| 11 | Azure service-selection drills | Mixed “which service?” practice |
| 12 | Timed mock exam | Full review of misses and guessed answers |
| 13 | Weak-area sprint | Retest weakest 2-3 areas |
| 14 | Final review | Light review, readiness checklist, exam pacing |
14-day emphasis
Prioritize service selection and scenario recognition. Many candidates lose points not because they lack general AI knowledge, but because they confuse similar Azure AI capabilities.
30-day balanced plan
Use this if you want a realistic schedule with time for learning, practice, and review.
Week 1: Baseline and AI fundamentals
| Day | Focus | Actions |
|---|---|---|
| 1 | Diagnostic | Take a mixed set and create your miss log. |
| 2 | AI workloads | Review prediction, vision, language, speech, search, automation, and generative AI scenarios. |
| 3 | Responsible AI | Learn the principles and write one practical example for each. |
| 4 | ML concepts | Review supervised vs. unsupervised learning, classification, regression, clustering. |
| 5 | ML lifecycle | Review training, validation, testing, inference, model evaluation, and deployment concepts. |
| 6 | Practice | Topic practice for AI workloads and ML basics. |
| 7 | Review | Repair misses and rebuild notes from memory. |
Week 2: Azure AI service families
| Day | Focus | Actions |
|---|---|---|
| 8 | Azure Machine Learning | Understand when custom ML is appropriate. |
| 9 | Computer vision | Match image scenarios to vision capabilities. |
| 10 | Document intelligence | Review document extraction and OCR-style use cases. |
| 11 | Language | Practice sentiment, key phrase extraction, entity recognition, and language detection scenarios. |
| 12 | Speech and translation | Compare speech, translation, and conversational scenarios. |
| 13 | Mixed practice | Service-selection questions across the week’s topics. |
| 14 | Review | Update comparison charts and retest missed items. |
Week 3: Generative AI, search, and scenario drills
| Day | Focus | Actions |
|---|---|---|
| 15 | Generative AI basics | Review prompts, completions, chat, summarization, and content generation. |
| 16 | Grounding and retrieval concepts | Understand how search and enterprise data can support generated responses. |
| 17 | Responsible generative AI | Review content safety, transparency, privacy, and human oversight. |
| 18 | Azure AI Search and knowledge retrieval | Practice search and retrieval scenario recognition. |
| 19 | Integrated scenarios | Choose between ML, prebuilt AI, search, and generative AI options. |
| 20 | Timed practice block | Complete a timed set and review deeply. |
| 21 | Repair day | Revisit weak topics and guessed questions. |
Week 4: Exam readiness
| Day | Focus | Actions |
|---|---|---|
| 22 | Full objective review | Walk through every major topic and mark confidence levels. |
| 23 | Weak area 1 | Drill your lowest-confidence topic. |
| 24 | Weak area 2 | Drill your second-lowest topic. |
| 25 | Weak area 3 | Drill your third-lowest topic. |
| 26 | Timed mock | Simulate exam conditions and review misses. |
| 27 | Service-selection sprint | Rapid-fire scenario practice across all services. |
| 28 | Responsible AI and generative AI review | Recheck principles, safety, grounding, and prompt concepts. |
| 29 | Final mixed review | Short mixed set; no deep new topics. |
| 30 | Light final review | Review notes, charts, and exam-day pacing. |
60/90-day full preparation path
Use this path if you are new to AI, new to Azure, or studying alongside work and other commitments.
| Phase | 60-day timing | 90-day timing | Goal |
|---|---|---|---|
| Foundation | Days 1-14 | Days 1-21 | Learn AI vocabulary, workloads, responsible AI, and ML basics |
| Azure AI services | Days 15-30 | Days 22-45 | Understand Azure AI service families and use cases |
| Scenario practice | Days 31-42 | Days 46-65 | Build service-selection accuracy |
| Generative AI and integration | Days 43-50 | Days 66-75 | Review prompts, grounding, copilots, search, and responsible AI |
| Mock and repair | Days 51-57 | Days 76-85 | Timed practice, missed-question repair, weak-area sprints |
| Final review | Days 58-60 | Days 86-90 | Light review, readiness checks, exam pacing |
60/90-day weekly rhythm
| Weekly session | Focus | What to produce |
|---|---|---|
| Session 1 | Learn one topic | Notes in your own words |
| Session 2 | Practice that topic | Miss log updates |
| Session 3 | Mixed review | Retest older misses |
| Optional session | Light hands-on or portal review | Better service recognition |
How to use the extra time in a 90-day plan
If you have 90 days, do not simply stretch passive reading. Use the extra time for spaced repetition:
- Revisit each topic at least three times.
- Build and update a service-selection matrix.
- Take smaller timed sets every 2-3 weeks.
- Add light hands-on review so Azure service names and workflows feel familiar.
- Keep the final 10 days focused on practice and repair, not new content.
Lightweight hands-on review for AI-901
AI-901 is not a deep implementation exam, but light hands-on review can help you understand what each service is for. Use free training environments, demos, documentation walkthroughs, or your own Azure environment if available. Avoid creating resources you do not understand, and monitor cost if you use a real subscription.
| Area | Hands-on or visual review | What to learn |
|---|---|---|
| Azure AI services overview | Browse service categories in the Azure portal or Microsoft learning environment | Which services are prebuilt AI services |
| Azure Machine Learning | Review workspace concepts and model lifecycle examples | When custom ML is used |
| Vision | Review sample image analysis and OCR flows | Inputs, outputs, and common use cases |
| Language | Review examples for sentiment, key phrases, and entity recognition | How text analytics scenarios are described |
| Speech | Review speech-to-text and text-to-speech examples | How speech differs from language and translation |
| Document Intelligence | Review document extraction examples | When forms, receipts, invoices, or documents are the clue |
| Azure AI Search | Review search and retrieval examples | When knowledge mining or retrieval is the clue |
| Generative AI | Review prompt, chat, grounding, and content safety examples | How generated responses are created and controlled |
Service-selection drill
Create a chart like this and fill it in from memory. Rebuild it several times during your plan.
| Scenario clue | Likely workload | Service family to review |
|---|---|---|
| Predict whether a customer will churn | Machine learning | Azure Machine Learning concepts |
| Group customers by similar behavior | Machine learning | Clustering concepts |
| Extract text from scanned forms | Vision/document processing | Azure AI Vision or Document Intelligence concepts |
| Identify sentiment in reviews | Natural language processing | Azure AI Language concepts |
| Convert spoken audio to written text | Speech | Azure AI Speech concepts |
| Translate text between languages | Translation | Translator concepts |
| Search enterprise content and retrieve relevant passages | Search/retrieval | Azure AI Search concepts |
| Generate a draft email or summarize a document | Generative AI | Azure OpenAI/generative AI concepts |
| Add safeguards for generated content | Responsible AI/safety | Responsible AI and content safety concepts |
Final-week rules
Use these rules regardless of whether you followed the 7-, 14-, 30-, or 60/90-day plan.
| Rule | What it means |
|---|---|
| Stop adding broad new material 2-3 days before the exam | Only fix repeated misses or clarify essential concepts. |
| Review guessed answers as seriously as wrong answers | A lucky correct answer is not stable knowledge. |
| Keep practice mixed | The real exam will not announce the topic before each question. |
| Prioritize service-selection accuracy | Many AI-901 questions are scenario-based. |
| Do short recall drills | Rebuild charts and definitions from memory. |
| Sleep and pacing matter | Do not trade the final night’s rest for low-quality cramming. |
Exam-readiness checks
You are more likely to be ready when you can do the following without notes:
- Explain the difference between classification, regression, and clustering.
- Identify whether a scenario is vision, language, speech, search, generative AI, or custom ML.
- Match common business requirements to the appropriate Azure AI service family.
- Explain responsible AI principles with examples.
- Recognize when a prebuilt AI service is more appropriate than building a custom model.
- Understand basic training, evaluation, deployment, and inference vocabulary.
- Describe prompt, grounding, retrieval, and content safety concepts at a fundamentals level.
- Complete timed mixed practice without running out of time.
- Review missed questions and explain why the correct answer is better than the distractors.
If your practice results are not improving
If your scores stay flat, change the review method before adding more content.
| Symptom | Likely cause | Fix |
|---|---|---|
| You miss many “which service?” questions | Service names and workloads are blended together | Build a service-selection chart and drill scenarios daily |
| You understand AI generally but miss Azure questions | Not enough Azure service familiarity | Review service families and example use cases |
| You miss responsible AI questions | Principles are memorized but not applied | Write practical examples for each principle |
| You miss ML questions | Core ML vocabulary is weak | Drill classification, regression, clustering, training, and inference |
| You run out of time | Too much rereading during questions | Practice timed sets and mark uncertain items quickly |
| You repeat the same mistakes | Miss log is not being used | Retest old misses every 2-3 days |
Practical next step
Start with a mixed AI-901 diagnostic practice set. Then build a missed-question log, choose the plan that matches your exam date, and spend your next session repairing the highest-impact weak area before adding more new material.