AI-900 — Microsoft Azure AI Fundamentals Study Plan
A practical AI-900 study plan for Microsoft Azure AI Fundamentals candidates, with 7-day, 14-day, 30-day, and longer preparation paths.
Who this study plan is for
This plan is for candidates preparing for the real Microsoft Azure AI Fundamentals (AI-900) exam. It is designed for people who need a practical schedule, not a broad introduction to artificial intelligence.
AI-900 is a fundamentals-level exam, so your preparation should emphasize:
- AI workload recognition and responsible AI concepts
- Basic machine learning concepts and Azure Machine Learning use cases
- Azure AI services and when to choose each service
- Computer vision, natural language processing, speech, translation, document processing, search, and generative AI scenarios
- Scenario-based service selection
- Clear understanding of terms, capabilities, and limitations at a conceptual level
Use the current Microsoft AI-900 skills outline as your source of truth while following the schedule below.
Which plan should you use?
| Time available | Best for | Main goal | Mock exam timing |
|---|---|---|---|
| 7 days | You have studied before or need final review | Close weak areas and build exam readiness | 1 diagnostic early, 1 timed mock near the end |
| 14 days | You know basic AI terms but need structure | Cover each objective area once, then review misses | Diagnostic on Day 1, timed mock around Day 11 or 12 |
| 30 days | Most candidates | Balanced learning, practice, and retention | Diagnostic in Week 1, timed mocks in Weeks 3 and 4 |
| 60 days | New to Azure or AI | Build concepts slowly with repeated practice | Diagnostic early, mocks in final 3 weeks |
| 90 days | Very new to cloud, data, or technical exams | Learn fundamentals without rushing | Light diagnostics monthly, mocks in final month |
If you are already comfortable with basic AI concepts but new to Microsoft Azure services, choose the 14-day or 30-day path. If you are new to both AI and Azure, choose 30, 60, or 90 days.
Daily study rhythm
Use the same rhythm regardless of plan length. Short, repeated practice is more effective than reading large sections once.
| Block | Time | What to do |
|---|---|---|
| Objective review | 20-30 min | Read or watch one focused topic from the current AI-900 skills outline |
| Service mapping | 10-15 min | Match scenarios to Azure services and features |
| Practice questions | 20-30 min | Answer targeted questions for the topic you just studied |
| Missed-question review | 15-20 min | Record why each miss happened and what rule would prevent it |
| Quick recall | 5-10 min | Close notes and explain the topic aloud in plain language |
For a weekday schedule, aim for 60-90 minutes per day. On weekends, use one longer block for mixed review or timed practice.
What to study for AI-900
Use these study lanes to organize your preparation. Do not treat them as official weighting; use the current Microsoft exam page for the official skills measured.
| Study lane | What to know | Practice focus |
|---|---|---|
| AI workloads and responsible AI | Common AI scenarios, prediction, classification, regression, anomaly detection, content generation, fairness, reliability, privacy, security, inclusiveness, transparency, accountability | Identify workload type and responsible AI concern from a short scenario |
| Machine learning on Azure | Basic ML terms, features, labels, training, validation, evaluation, supervised and unsupervised learning, classification, regression, clustering | Choose the correct ML approach for a business problem |
| Computer vision | Image analysis, object detection, OCR, facial-related capabilities, document extraction concepts | Match visual tasks to the right Azure AI capability |
| Natural language processing | Language understanding, sentiment, key phrase extraction, entity recognition, translation, speech-to-text, text-to-speech | Select the best NLP, speech, or translation service for a scenario |
| Document and knowledge workloads | Document extraction, search and retrieval concepts, structured vs unstructured data | Recognize when to use document intelligence, search, or language features |
| Generative AI | Prompts, completions, copilots, grounding, content safety, responsible use, Azure generative AI services and tooling at a conceptual level | Interpret use cases, risks, and service-selection questions |
7-day final review plan
Use this if your exam is within one week. The goal is not to relearn everything. The goal is to find weak areas, fix repeated mistakes, and become comfortable answering scenario questions under time pressure.
| Day | Focus | Study actions | Practice target |
|---|---|---|---|
| 1 | Diagnostic and plan | Take a mixed diagnostic without notes. Mark every uncertain question, even if correct. Build a weak-area list. | 40-60 mixed questions |
| 2 | AI workloads and responsible AI | Review workload types and responsible AI principles. Practice identifying the concern in each scenario. | 30-40 targeted questions |
| 3 | Machine learning basics | Review classification, regression, clustering, training, evaluation, features, labels, and model lifecycle concepts. | 30-40 targeted questions |
| 4 | Vision, document, and OCR workloads | Review image analysis, OCR, object detection, document extraction, and when to choose each service category. | 30-40 targeted questions |
| 5 | NLP, speech, translation, search | Review language, speech, translator, and search-style scenarios. Create a service-selection table from memory. | 40-50 targeted questions |
| 6 | Generative AI and mixed mock | Review generative AI concepts, prompts, grounding, safety, and responsible use. Take a timed mock. | 1 timed mock |
| 7 | Final cleanup | Review only missed questions, confusing terms, and service-selection rules. Stop adding new material late in the day. | 20-30 light questions |
7-day rules
- Do not spend the week passively reading all materials from the beginning.
- Keep a single-page “service selection” sheet.
- Prioritize repeated misses over topics you already answer correctly.
- Stop adding new topics the day before the exam unless they are directly tied to repeated mistakes.
- The final evening should be light review, not a full mock.
14-day focused plan
Use this if you need a compact but complete pass through the exam content.
| Day | Focus | Main task |
|---|---|---|
| 1 | Diagnostic | Take a mixed diagnostic. Create your topic tracker: strong, uncertain, weak. |
| 2 | AI workloads | Review prediction, classification, regression, anomaly detection, language, vision, and generative workloads. |
| 3 | Responsible AI | Study fairness, reliability, safety, privacy, security, inclusiveness, transparency, and accountability. |
| 4 | ML concepts | Review features, labels, training, validation, testing, model evaluation, and basic algorithm categories. |
| 5 | Azure Machine Learning concepts | Review what Azure Machine Learning is used for at a fundamentals level. Focus on lifecycle and use cases. |
| 6 | Computer vision | Review image analysis, OCR, object detection, and related scenario selection. |
| 7 | Vision practice and catch-up | Drill mixed vision/document scenarios and review misses from Days 1-6. |
| 8 | NLP | Review sentiment, key phrases, entities, language understanding, and text analytics scenarios. |
| 9 | Speech and translation | Review speech-to-text, text-to-speech, translation, and when to combine services. |
| 10 | Search, documents, and knowledge | Review document extraction and search/retrieval scenarios. |
| 11 | Generative AI | Review prompts, completions, copilots, grounding, responsible use, and content safety concepts. |
| 12 | Timed mock | Take a timed mixed mock. Review every miss and every lucky guess. |
| 13 | Weak-area sprint | Re-study only weak domains. Build final service-selection and vocabulary notes. |
| 14 | Final review | Light mixed practice, exam-readiness check, and rest. No heavy new material. |
30-day balanced plan
This is the best default plan for most AI-900 candidates. It gives enough time to learn concepts, practice service selection, and complete timed review without cramming.
Week 1: Baseline and AI fundamentals
| Day | Focus | Actions |
|---|---|---|
| 1 | Diagnostic | Take a short mixed diagnostic. Record weak areas by topic. |
| 2 | AI workload types | Review prediction, classification, regression, clustering, anomaly detection, vision, language, and generative tasks. |
| 3 | Responsible AI | Learn the responsible AI principles and practice identifying them in scenarios. |
| 4 | AI terminology | Review model, dataset, feature, label, inference, confidence, training, and evaluation. |
| 5 | Workload drills | Answer scenario questions that ask “What type of AI workload is this?” |
| 6 | Weekly mixed practice | Complete a mixed set and review misses. |
| 7 | Rest or catch-up | Revisit only weak Week 1 topics. |
Week 2: Machine learning and Azure AI service selection
| Day | Focus | Actions |
|---|---|---|
| 8 | ML concepts | Review supervised and unsupervised learning, classification, regression, and clustering. |
| 9 | Model lifecycle | Review training, validation, testing, deployment, monitoring, and evaluation concepts. |
| 10 | Azure Machine Learning | Study the role of Azure Machine Learning in building and managing ML solutions. |
| 11 | Computer vision | Review image analysis, OCR, object detection, and visual recognition scenarios. |
| 12 | Document workloads | Review document extraction and forms/document processing concepts. |
| 13 | Targeted practice | Drill ML, vision, and document questions. |
| 14 | Weekly review | Update your missed-question log and service-selection sheet. |
Week 3: Language, speech, search, and generative AI
| Day | Focus | Actions |
|---|---|---|
| 15 | NLP basics | Review sentiment, key phrase extraction, entity recognition, and language understanding. |
| 16 | Speech and translation | Review speech-to-text, text-to-speech, translation, and common combined scenarios. |
| 17 | Search and retrieval | Review search-style workloads, indexing concepts, and knowledge retrieval scenarios. |
| 18 | Generative AI basics | Review prompts, completions, copilots, grounding, and responsible generative AI use. |
| 19 | Generative AI on Azure | Review Azure generative AI tooling and service-selection scenarios at a conceptual level. |
| 20 | Timed mock 1 | Take your first timed mixed mock. |
| 21 | Mock review | Spend more time reviewing the mock than taking it. Rewrite rules for every miss. |
Week 4: Exam readiness and weak-area repair
| Day | Focus | Actions |
|---|---|---|
| 22 | Weakest domain 1 | Re-study your lowest-scoring topic and complete targeted practice. |
| 23 | Weakest domain 2 | Repeat for the second-lowest topic. |
| 24 | Service-selection sprint | Practice scenarios that ask which Azure AI capability best fits a requirement. |
| 25 | Responsible AI and generative AI review | Review safety, privacy, fairness, transparency, grounding, and content safety themes. |
| 26 | Timed mock 2 | Take a second timed mixed mock. |
| 27 | Deep review | Review all misses, guessed questions, and slow questions. |
| 28 | Final notes | Build a one-page summary of terms, service choices, and responsible AI rules. |
| 29 | Light mixed practice | Do a short mixed set. Avoid exhausting yourself. |
| 30 | Final readiness | Review your one-page summary and missed-question log. Stop heavy studying. |
60/90-day full preparation path
Use this path if you are new to AI, new to Azure, or balancing study with a full workload. For 60 days, use the weekly structure below. For 90 days, stretch the first four weeks into six weeks and add more review days.
| Phase | 60-day timing | 90-day timing | Goal |
|---|---|---|---|
| Foundation | Weeks 1-2 | Weeks 1-3 | Learn basic AI vocabulary, workloads, and responsible AI concepts |
| Core services | Weeks 3-4 | Weeks 4-6 | Learn Azure AI service categories and common scenarios |
| Applied selection | Weeks 5-6 | Weeks 7-8 | Practice choosing services from business requirements |
| Mock and repair | Weeks 7-8 | Weeks 9-12 | Timed mocks, weak-area review, final readiness |
Weeks 1-2: AI foundation
| Focus area | Study actions |
|---|---|
| AI workload types | Build examples for prediction, classification, regression, clustering, anomaly detection, vision, language, and generative AI. |
| Responsible AI | Create flashcards for each responsible AI principle and practice scenario recognition. |
| Basic ML vocabulary | Learn features, labels, datasets, training, validation, testing, inference, and evaluation. |
| First diagnostic | Take a low-pressure diagnostic near the end of Week 2 to guide the next phase. |
Weeks 3-4: Microsoft Azure AI services overview
| Focus area | Study actions |
|---|---|
| Azure Machine Learning | Understand when a team would use Azure Machine Learning instead of a prebuilt AI service. |
| Azure AI services | Map common tasks to service categories: vision, language, speech, translation, document, search, and generative AI. |
| Computer vision and documents | Practice OCR, image analysis, object detection, and document extraction scenarios. |
| Language and speech | Practice sentiment, entity recognition, key phrases, translation, speech-to-text, and text-to-speech scenarios. |
Weeks 5-6: Scenario drills
| Focus area | Study actions |
|---|---|
| Service selection | Given a business problem, choose the most appropriate Azure AI capability. |
| Combined scenarios | Practice cases that combine language, speech, translation, search, or document processing. |
| Responsible AI in design | Identify privacy, fairness, transparency, safety, and accountability concerns in solution descriptions. |
| Generative AI | Review prompts, grounding, content safety, copilots, and responsible generative AI patterns. |
Weeks 7-8 or final month: Mock exams and repair
| Task | Frequency | How to use it |
|---|---|---|
| Timed mock | 1 per week | Simulate exam pacing and identify weak areas |
| Targeted drill | 3-4 times per week | Repair one weak topic at a time |
| Missed-question review | After every practice session | Record the reason for each miss |
| Final mixed set | Final 2-3 days | Keep it light and confidence-building |
Missed-question review method
A missed question is useful only if you convert it into a rule. Use this simple log.
| Field | What to write |
|---|---|
| Topic | Example: responsible AI, NLP, vision, generative AI, ML concepts |
| Why I missed it | Misread, did not know term, confused two services, guessed, rushed |
| Correct rule | A short statement that would help you answer a similar question next time |
| Review action | Re-read objective, drill 10 questions, update service map, make flashcard |
| Retest date | Schedule a quick retest within 2-4 days |
Common AI-900 miss patterns
| Miss pattern | Fix |
|---|---|
| Confusing workload type with service name | First identify the task, then choose the Azure capability. |
| Mixing classification and regression | Classification predicts a category; regression predicts a numeric value. |
| Treating all text workloads as generative AI | Some text tasks are extraction, sentiment, translation, or speech-related. |
| Ignoring responsible AI language | Look for fairness, safety, privacy, transparency, inclusiveness, and accountability clues. |
| Memorizing service names without scenarios | Practice “requirement to service” mapping every day. |
| Overstudying implementation details | AI-900 is fundamentals-focused; prioritize concepts and service selection. |
Service-selection practice table
Build your own version of this table as you study.
| Scenario clue | Think about |
|---|---|
| Predict a numeric value | Regression-style machine learning |
| Predict a category or label | Classification-style machine learning |
| Group similar items without predefined labels | Clustering-style machine learning |
| Extract text from images or scanned documents | OCR or document processing capability |
| Identify objects or describe image content | Computer vision capability |
| Detect sentiment or key phrases in text | Language capability |
| Convert spoken audio to text | Speech-to-text capability |
| Convert text into spoken audio | Text-to-speech capability |
| Translate text or speech between languages | Translation capability |
| Retrieve relevant information from indexed content | Search or retrieval capability |
| Generate text, summarize, or assist with content creation | Generative AI capability |
| Reduce harmful, biased, unsafe, or private-data risks | Responsible AI and safety controls |
When to use timed mock exams
Timed mocks are most useful after you have reviewed the main objective areas once. Taking too many full mocks too early can waste good practice material.
| Preparation length | Best mock schedule |
|---|---|
| 7 days | Diagnostic on Day 1, timed mock on Day 6 |
| 14 days | Diagnostic on Day 1, timed mock on Day 12 |
| 30 days | Diagnostic in Week 1, timed mocks around Days 20 and 26 |
| 60 days | Diagnostic in Week 2, mocks in Weeks 6, 7, and 8 |
| 90 days | Light diagnostics during the first two months, mocks in final month |
After each timed mock:
- Review every incorrect answer.
- Review every correct answer you guessed.
- Mark questions that took too long.
- Group misses by topic.
- Re-study the top two weak topics before taking another mock.
Final-week rules
During the final week, your job is to reduce uncertainty, not collect more resources.
Do
- Use the current Microsoft AI-900 skills outline as your checklist.
- Review your missed-question log daily.
- Drill service-selection scenarios.
- Revisit responsible AI and generative AI concepts.
- Take one timed mock if you have not done so recently.
- Keep final review sessions short and focused.
Avoid
- Starting a long new course.
- Reading documentation without practice questions.
- Memorizing implementation details that are beyond fundamentals level.
- Taking multiple full mocks back-to-back without review.
- Studying heavily late the night before the exam.
Exam-readiness checks
You are likely ready to sit for Microsoft AI-900 when you can do the following without notes:
| Readiness check | Can you do it? |
|---|---|
| Explain common AI workload types in plain language | Yes / No |
| Distinguish classification, regression, and clustering | Yes / No |
| Identify responsible AI concerns from a scenario | Yes / No |
| Choose between ML, vision, language, speech, translation, document, search, and generative AI capabilities | Yes / No |
| Explain basic training, evaluation, and inference concepts | Yes / No |
| Handle mixed practice questions within a timed session | Yes / No |
| Review a missed question and state the rule you missed | Yes / No |
If you answer “No” to more than two items, spend your next study block on targeted review instead of another full mock.
Practical next step
Start with a mixed diagnostic set. Then choose the 7-day, 14-day, 30-day, or 60/90-day path based on your exam date. Keep a missed-question log from the first session, and let that log decide what you study next.