Article
AI-900 Exam Guide 2026: FREE Azure AI Fundamentals Study Plan
Key Facts
- AI-901, the exam that replaced AI-900, costs $99 USD in the United States, the same fee AI-900 carried (Microsoft exam pricing).
- Microsoft Fundamentals exams such as AI-901 allow 45 minutes of exam time within 65 minutes of seat time (Microsoft Learn).
- The passing score for AI-900 and AI-901 is 700 on a scale of 1-1,000, a scaled score rather than a raw percentage.
- Microsoft retired AI-900 on June 30, 2026 and replaced it with Exam AI-901, announced in beta on April 15, 2026.
- The final AI-900 skills outline (May 2, 2025) weighted generative AI workloads heaviest at 20-25%.
- The other four AI-900 domains each carried 15-20%: AI workloads, machine learning fundamentals, computer vision, and NLP.
- Microsoft defines six Responsible AI principles: fairness, reliability & safety, privacy & security, inclusiveness, transparency, and accountability.
- The Azure AI Fundamentals certification has no expiration date once earned.
- AI-900 required no coding, but AI-901 expects knowledge of Python coding syntax and familiarity with Azure resources (Microsoft Learn).
- AI-901 uses two domains: Identify AI concepts and capabilities (40-45%) and Implement AI solutions by using Microsoft Foundry (55-60%).
Why the Azure AI Fundamentals Certification Is Worth $99 in 2026
Generative AI went from research curiosity to boardroom mandate in about 18 months. Every enterprise now has an "AI strategy," every product manager is being asked "how does this use AI," and every hiring manager is quietly scanning resumes for AI literacy. The Microsoft Certified: Azure AI Fundamentals certification is an entry-level way to show you can have that conversation. Until June 30, 2026 you earned it with Exam AI-900 , whose final outline gave generative AI and Azure OpenAI the largest share; you now earn it with Exam AI-901 .
Microsoft retired the AI-900 exam on June 30, 2026 ; new candidates now take AI-901 , which earns the same Azure AI Fundamentals certification. This guide covers the final AI-900 skills outline (the May 2025 skills refresh ), the AI-901 transition , the exam cost of $99 USD , and the six Responsible AI principles. By the end you will know what carries over to AI-901, what is new, and how to pace your study.
AI-900 At a Glance (2026)
- Attribute Detail
- Exam Code AI-900
- Status Retired June 30, 2026; replaced by AI-901
- Full Name Microsoft Azure AI Fundamentals
- Cost (USD) $99
- Exam Time 45 minutes
- Seat Time 65 minutes (includes instructions and the candidate agreement)
- Questions Not published (most Microsoft exams have 40-60)
- Passing Score 700 on a scale of 1-1,000 (scaled scoring, not raw 70%)
- Prerequisites None
Languages 13 (English, Japanese, Chinese Simplified/Traditional, Korean, German, French, Spanish, Portuguese-BR, Russian, Indonesian, Arabic, Italian)
- Delivery Pearson VUE test center or online proctored
- Validity Lifetime (no expiration)
Retirement Date June 30, 2026, 11:59 PM Central Standard Time (replaced by AI-901, announced in beta April 15, 2026)
Final Skills Version May 2, 2025
Preparing for the certification now? Practice with our free AI-901 questions , or review the AI-900 concepts with the free AI-900 practice questions . Every question includes an explanation.
What AI-900 Was and Why It Still Matters in 2026
AI-900 was Microsoft's fundamentals-level exam for Azure AI. It did not make you an AI engineer, and it did not test Python, notebooks, or deployment pipelines. It certified that you understood what AI workloads exist , which Azure service solves which problem , and how to apply AI responsibly . Those concepts now make up AI-901's "Identify AI concepts and capabilities" domain (40-45%), while AI-901 adds hands-on Python and Microsoft Foundry work.
That sounds narrow, and it is — but in 2026 that exact vocabulary is the difference between a product manager who can run an effective AI roadmap meeting and one who nods through it. The certification suits:
Product managers and business analysts who need to spec AI-enabled features without becoming data scientists.
Consultants and pre-sales engineers who pitch Azure AI to clients.
Career changers transitioning from non-AI backgrounds into AI-adjacent roles.
Developers and sysadmins who want the fundamentals credential before stacking AI-103 or AI-300.
Students exploring AI careers.
AI-900's final refresh (May 2, 2025) raised the share of its generative AI domain to 20-25% of the exam — more weight than any other section. That domain covered generative AI models and scenarios, responsible generative AI, Azure OpenAI Service , and Azure AI Foundry and its model catalog. In AI-901, generative AI moves further to the center: you deploy models, write prompts, and build a simple agent in Microsoft Foundry.
The AI Skills Gap Is a Hiring Signal
The U.S. Bureau of Labor Statistics projects computer and information research scientists (which covers AI roles) to grow 22% from 2025 to 2035, much faster than the 3% average for all occupations, and reports a May 2025 median wage of $140,300. Even if you do not become an AI engineer, demonstrating that you understand AI concepts, terminology, and responsible deployment is increasingly non-negotiable. Azure AI Fundamentals, now earned through AI-901, is a low-cost way to put that signal on your resume.
- Who Should Earn Azure AI Fundamentals (And Who Can Skip It)
- Ideal Candidates
The Product Manager. You ship software features, your roadmap now requires AI components, and you cannot tell your eng lead whether "fine-tuning" or "RAG" is appropriate for your use case. The certification's concept domain covers exactly that vocabulary.
The Career Changer. You are pivoting from marketing, finance, operations, or teaching into a tech-adjacent role. You need a low-cost credential that says "I understand AI" to screening recruiters. AI-901 is $99 in the US; budget time to learn basic Python, which Microsoft expects.
The Business Analyst or Data Analyst. You already work with data but do not touch ML directly. Azure AI Fundamentals gives you the vocabulary to participate in AI project scoping and responsibly evaluate vendor AI claims.
The Consultant or Pre-Sales Engineer. You sell or implement Azure solutions. Clients now ask about AI on every call. The certification helps you credibly sketch Azure OpenAI, RAG, and Azure AI Search architectures on a whiteboard.
The Student. The Azure AI Fundamentals certification does not expire, which makes it a low-cost way to show AI literacy.
The Experienced Engineer Preparing for AI-103. Do not skip the fundamentals just because you are experienced. The conceptual scaffolding (responsible AI principles, prompt engineering vocabulary, service-to-workload mapping) makes AI-103 easier.
Candidates Who Can Skip It
Active AI/ML engineers who already ship LLM applications daily — jump to AI-103 or AI-300 instead.
Pure researchers working on model architecture — Azure AI Fundamentals is product-focused, not research-focused.
Candidates committed to AWS or Google Cloud — take AWS AI Practitioner or Google Cloud Generative AI Leader instead.
Prerequisites: AI-900 vs AI-901
AI-900 had no prerequisites : Microsoft said data science and software engineering experience were not required and recommended awareness of basic cloud concepts and client-server applications. AI-901 also has no prerequisite exam, but its bar is higher. Microsoft says AI-901 candidates need knowledge of Python coding syntax and programming techniques and should be familiar with Azure resources .
The Final AI-900 Skills Outline (May 2, 2025)
The skills measured on AI-900 as of May 2, 2025 (the final version before the June 30, 2026 retirement) were the following. Most of these concepts reappear in AI-901's concept domain; AI-900's separate machine learning fundamentals domain does not.
- Domain Weight What It Covers
- Describe AI workloads and considerations 15-20% Common AI workload types + six responsible AI principles
Fundamentals of machine learning on Azure 15-20% Regression, classification, clustering, deep learning, Azure Machine Learning
- Features of computer vision on Azure 15-20% Image classification, object detection, OCR, Azure AI Vision, Face
- Features of NLP on Azure 15-20% Text analytics, translation, speech, conversational language
Features of generative AI on Azure 20-25% Azure OpenAI , LLMs, prompt engineering, RAG, copilots, responsible GenAI
The generative AI section was the heaviest; Microsoft's change log shows its share increased in the May 2025 refresh while the machine learning share decreased. Let us take each domain in turn.
Domain 1: AI Workloads and Considerations (15-20%)
This domain has two halves: identifying workloads and responsible AI .
Common AI Workloads you must be able to identify by scenario:
Prediction/forecasting (supervised ML: will this customer churn?).
Anomaly detection (unsupervised: is this credit card charge fraud?).
Computer vision (image classification, object detection, face detection, OCR).
Natural language processing (sentiment, key phrases, entity recognition, translation).
Document intelligence (extracting structured fields from invoices, receipts, forms).
Knowledge mining (indexing unstructured data for search).
Generative AI (text, image, code, audio generation).
The question pattern is always the same: Microsoft describes a business scenario and asks which workload type it represents. Memorize one canonical example per workload and you will answer these in under 20 seconds.
Domain 2: Machine Learning Fundamentals on Azure (15-20%)
This domain tests your ability to identify ML techniques and describe Azure Machine Learning capabilities .
Techniques you must know:
Supervised learning — labeled data. Split into:
Regression → continuous numeric output (house price, temperature).
Classification → discrete category output (spam/not spam, disease/healthy).
Unsupervised learning — no labels.
Clustering → grouping similar items (customer segmentation).
Deep learning — neural networks, used under the hood for vision, speech, and language.
Azure Machine Learning (the service) capabilities you must recognize:
Azure ML Studio (the web workspace).
Automated ML (AutoML) — it tries many algorithms for you; tested as the "non-data-scientist" option.
Designer — drag-and-drop pipeline builder.
Notebooks — code-first experience.
Compute instances and compute clusters .
Endpoints for real-time and batch deployment.
Example test question pattern: "A business analyst with no coding background needs to build a classification model. Which Azure ML capability should they use?" Answer: Automated ML .
Domain 3: Computer Vision on Azure (15-20%)
Memorize the Azure AI Vision capabilities and the adjacent services:
- Capability Use Case Azure Service
- Image classification "What is in this image?" Azure AI Vision / Custom Vision
- Object detection "What objects, and WHERE?" Azure AI Vision / Custom Vision
- OCR / Read API Extract text from images Azure AI Vision
- Face detection/analysis Detect/analyze human faces Azure AI Face (gated access)
- Form/invoice extraction Structured field extraction from documents Azure AI Document Intelligence
- Video analysis Indexing and moderating video Azure AI Video Indexer
The single most common mistake: confusing classification (what) with detection (what + where via bounding boxes). Remember: detection includes location .
Microsoft also tests Custom Vision vs prebuilt Azure AI Vision. Custom Vision is when you train a model on your own labeled images (e.g., "classify defects on our specific widgets"). Use prebuilt Azure AI Vision when the general catalog of tags is enough.
Domain 4: Natural Language Processing on Azure (15-20%)
The NLP domain is about mapping text and speech tasks to Azure services.
Task Azure Service
Sentiment analysis, key phrase extraction, language detection, entity recognition, PII redaction Azure AI Language
Intent and entity extraction for chatbots (formerly LUIS) Azure AI Language (Conversational Language Understanding)
- Translation between 100+ languages Azure AI Translator
- Speech-to-text, text-to-speech, translation of speech Azure AI Speech
- Question answering over a knowledge base (formerly QnA Maker) Azure AI Language (Custom Question Answering)
- Orchestrating multi-turn conversations Azure AI Bot Service
Azure consolidated many of these under the "Azure AI Language" umbrella. If your study material refers to Text Analytics , LUIS , or QnA Maker as separate services, it is out of date — their capabilities now sit inside Azure AI Language.
Domain 5: Generative AI on Azure (20-25%) — The Biggest Domain
This was AI-900's heaviest-weighted domain, and these concepts underpin AI-901's Foundry objectives too. You must understand:
Core generative AI concepts.
Large Language Models (LLMs) — transformer models trained on massive text corpora.
Tokens — the chunks LLMs process (roughly 3-4 characters of English).
Prompts — your input to the model.
Completions — the model's output.
Temperature — a setting from 0 (deterministic) to 2 (creative/chaotic).
Grounding — anchoring a model's output in trusted external data.
Hallucination — when a model produces confident but false output.
Embeddings — numeric vector representations of text used for similarity search.
Azure OpenAI Service specifics.
It is the Azure-hosted, enterprise-governed version of OpenAI models (GPT-4o, GPT-4.1, o-series reasoning models, gpt-image models, Whisper).
Provides the same API as OpenAI but with Azure networking, identity, compliance, and private endpoints .
Supports deployments where you pick a model and scale.
Integrates with Azure AI Content Safety for input/output moderation.
Patterns you must recognize.
Zero-shot prompting — ask without examples.
Few-shot prompting — include examples in the prompt.
Chain-of-thought prompting — ask the model to reason step-by-step.
Retrieval-Augmented Generation (RAG) — retrieve relevant documents from Azure AI Search, inject them into the prompt as context, and let the LLM answer with grounding.
Fine-tuning — retraining the model weights on your data (expensive, rare in fundamentals).
Agents/Copilots — LLMs that can call tools, execute code, browse, and act on behalf of the user.
Microsoft Foundry (formerly Azure AI Foundry, originally Azure AI Studio) is the unified development portal you are most likely to see referenced. The official AI-900 study guide still uses the Azure AI Foundry name, so treat the two as the same platform. It includes a model catalog , prompt flow (retiring April 20, 2027), evaluations , content safety , and deployment pipelines . If you see "Foundry" on the exam, think "the developer workspace for building GenAI apps on Azure."
A core GenAI concept is RAG , because it is Azure's answer to "how do I get a GPT model to answer from my own data without fine-tuning." If a question asks how to ground an LLM in private data cheaply, the answer is almost always RAG .
The Responsible AI Deep Dive (Tested Heavily, Misunderstood Often)
Microsoft's Responsible AI Standard defines six principles . AI-900 tested them, and AI-901's concept domain lists all six as objectives. Memorize these with one scenario each.
# Principle One-Sentence Definition Canonical Scenario
1 Fairness AI systems should treat people equitably and avoid bias. A hiring model rejects female candidates at a higher rate → fairness issue.
2 Reliability and safety Systems must perform consistently under expected and unexpected conditions. A self-driving car must handle rain, fog, and rare edge cases without dangerous behavior.
3 Privacy and security Personal data must be protected throughout the AI lifecycle. A model is trained on health records → data must be anonymized and access controlled.
4 Inclusiveness AI must empower people of all backgrounds and abilities. A speech-to-text service must work well for speakers of every dialect, including those with speech differences.
5 Transparency Users must understand how AI systems work and make decisions. A loan denial model must provide feature importance / reasons, not just a score.
6 Accountability Humans are responsible for AI; systems must be auditable and have human oversight. A hospital must retain logs of every AI-generated triage recommendation so reviewers can audit outcomes.
The test pattern: Microsoft gives you a scenario and asks which principle it violates or embodies. The most commonly confused pair is transparency vs accountability — transparency is about explaining the model, accountability is about who is responsible and how it is audited . Another frequently confused pair is fairness vs inclusiveness — fairness is about equitable outcomes once a system is deployed, inclusiveness is about making sure the system is accessible and usable by diverse populations in the first place.
Azure also ships practical Responsible AI tools you should recognize:
Responsible AI dashboard in Azure Machine Learning — fairness analysis, error analysis, counterfactuals, and interpretability.
Azure AI Content Safety — filters harmful content (hate, self-harm, sexual, violence) for prompts and completions.
Transparency notes — Microsoft's published documents that explain each service's intended uses and limitations.
How Hard Is Azure AI Fundamentals?
Microsoft does not publish pass rates for AI-900 or AI-901. What the official outlines do show: AI-900 asked you to describe and identify, while AI-901 puts 55-60% of the exam on implementing AI solutions in Microsoft Foundry, with Python expected. Practice questions matter more than more videos, and for AI-901 so does hands-on time in Foundry.
Concepts that are easy to confuse:
Generative AI questions about Azure OpenAI patterns (RAG vs fine-tuning vs zero-shot).
Responsible AI principle identification (especially transparency vs accountability).
Azure AI service mapping post-rebrand (confusing Text Analytics/LUIS with Azure AI Language).
Classification vs detection in computer vision.
Supervised vs unsupervised ML task identification.
Ready to benchmark yourself? Try our free AI-901 practice questions , or review the five AI-900 domains with the free AI-900 practice questions . Every question is explained.
Study Plans for AI-901 (After AI-900's Retirement)
AI-900 can no longer be booked, so these plans prepare you for AI-901 . Use the AI-900 material in this guide for AI-901's concept domain (40-45%), and spend at least half your time on its implementation domain (55-60%), which asks you to build small solutions in Microsoft Foundry. Finish either plan with Microsoft's free practice assessment and a review of every wrong answer before you book.
4-Week Beginner Plan
Target: 8-10 hours per week.
Week Focus Activities
1 AI concepts + Responsible AI Read the AI-901 study guide. Memorize the 6 responsible AI principles with one scenario each. Learn how generative AI models work, how to choose a model by capability, and the deployment options and parameters. End of week: 25 practice questions.
2 AI workloads + Python basics Text analysis (keyword extraction, entity detection, sentiment, summarization), speech recognition and synthesis, computer vision and image generation, and information extraction. Practice basic Python syntax every day. End of week: 50 practice questions.
3 Generative AI apps and agents in Foundry Deploy a model and chat with it in the Foundry portal, write system and user prompts, build a lightweight chat client with the Foundry SDK, then create and test a single agent and a small client for it.
4 Text, speech, vision and extraction apps + review Build small apps for text analysis, Azure Speech in Foundry Tools, visual input and image generation, and Content Understanding extraction from documents, images, audio and video. Take the Microsoft practice assessment, review every wrong answer, then book AI-901.
2-Week Plan (Some Python, Azure or AI Experience)
Target: 10-12 hours per week.
Week Focus Activities
1 Concept domain sweep Read the AI-901 study guide and this guide's domain notes. Take a 50-question diagnostic. Flag weak areas.
2 Foundry implementation Work through each implementation objective in Foundry: model deployment, an SDK chat client, a single agent, speech, vision, and Content Understanding. Do 100+ practice questions, take the Microsoft practice assessment, then book AI-901.
- Recommended Resources (Free First, Then Paid)
- Free Resources (Start Here)
Microsoft Learn AI-901 study guide — the official skills list for the current exam. Always the first stop.
Microsoft Learn AI-900 study guide (May 2, 2025) — still a useful concept checklist for the retired exam.
Microsoft practice assessment — for AI-901, Microsoft hosts its free practice assessment on AI Skills Navigator (sign-in required). Use it as a diagnostic and again before exam day.
Instructor-led course — Microsoft's AI-901 exam page lists "Introduction to AI in Azure".
ExamFinch free practice banks — AI-901 practice questions and AI-900 practice questions , each with an explanation.
Microsoft Azure free account — $200 of credit plus always-free tiers to try Azure AI services hands-on.
Paid Resources (If Budget Allows)
Third-party practice tests and courses can help, but many were written for AI-900. Before you buy, check that the product names AI-901 and covers its Microsoft Foundry implementation domain.
- Exam-Day Strategy (Pearson VUE + Online Proctoring)
- Registration and Setup
AI-900 can no longer be booked; register for AI-901 on Microsoft Learn.
Before Exam Day
Run the Pearson VUE system test at least 24 hours in advance (check bandwidth, camera, microphone).
Clear your exam room: no phone, no notebooks, no second monitor, no water bottles with labels, no smartwatches.
Have a valid government-issued photo ID ready.
During the Exam
Read every question twice. Microsoft's question writers are precise; a single word ("NOT", "EXCEPT", "MOST") flips the answer.
Use the Mark for Review flag. If you are unsure, flag and move on. AI-901 gives you 45 minutes, and Microsoft does not publish the question count, so keep an eye on the clock.
Eliminate wrong answers first. Most multiple-choice questions have one obviously wrong distractor. Crossing it out improves your odds even when you are guessing.
Do not overthink Responsible AI questions. The obvious answer is usually correct. Microsoft is testing recognition, not subtle philosophical distinctions.
After the Exam
For most exams, Microsoft says you have results within minutes of finishing, along with a score report, and your score is available in your Microsoft Learn profile within 24 hours. If you pass, you can share your Microsoft Certified: Azure AI Fundamentals badge on LinkedIn and your résumé.
- Cost, Retakes, Vouchers, and Student Discounts in 2026
- Pricing
The exam fee in the United States was $99 USD until Microsoft retired AI-900 on June 30, 2026; prices varied by country.
Retake Policy
If you failed a Microsoft exam:
Retake #1: 24-hour wait.
Retakes #2-#5: 14-day wait between attempts.
You can take the same exam a maximum of 5 times in 12 months from your first attempt.
Exam Replay Bundle
Microsoft's Exam Replay offer bundles an exam with a retake.
Student Discounts and Free Vouchers
Microsoft offers academic exam pricing for verified students in most countries and regions (except India and China); check Microsoft Learn for current offers.
Career and Salary Impact of Azure AI Fundamentals
Azure AI Fundamentals is not a job on a resume — it is a credibility signal. It does not guarantee any salary bump on its own. What it does is move you past the "does this candidate understand AI" filter that more recruiters are applying in 2026.
For the full AI-engineering track, Azure AI Fundamentals is the first rung. The ladder is: AI-901 → AI-103 (Azure AI Apps and Agents Developer Associate) or AI-300 (Machine Learning Operations Engineer Associate) .
- Three Common Mistakes
- Mistake 1: Studying Only AI-900 Material for AI-901
AI-900 material covers the concepts, but 55-60% of AI-901 is the Microsoft Foundry implementation domain: deploying models, writing prompts, building lightweight apps with the Foundry SDK, a single-agent solution, speech, vision, and Content Understanding. If your prep has no hands-on Foundry or Python work, it is incomplete.
Mistake 2: Skimming Responsible AI
The six principles are finite and testable, and AI-901 lists every one of them as an objective. Candidates who skim the concepts instead of memorizing the distinctions lose easy points. Make flashcards. Know the difference between transparency and accountability. Know when "inclusiveness" is the right answer.
Mistake 3: Never Practicing Under Time Pressure
The real exam is 45 minutes under proctored conditions, and time pressure changes your accuracy. Before exam day, set a 45-minute timer and work through a full set of practice questions in one sitting.
AI-900 vs AI-901: What Changes After June 30, 2026
Microsoft's "Evolving the Microsoft Certified: Azure AI Fundamentals Certification" blog post (Tech Community, April 15, 2026) and the AI-901 study guide confirm the details below. The certification name stays the same — only the exam code and content shift. Candidates who passed AI-900 before it retired hold the Microsoft Certified: Azure AI Fundamentals credential permanently; since June 30, 2026, AI-901 is the exam that earns it.
- Dimension AI-900 (retired June 30, 2026) AI-901 (replacement, announced in beta April 15, 2026)
- Credential earned Microsoft Certified: Azure AI Fundamentals Microsoft Certified: Azure AI Fundamentals (same)
- Level Beginner Beginner
Focus Understand AI concepts; identify which Azure service to use Understand AI concepts; implement basic AI solutions with Microsoft Foundry
- Hands-on required No Yes — light Python + Foundry portal work
- Python required No Yes (knowledge of Python coding syntax expected)
Skills structure Five describe-only domains (15-25% each) Two domains: Identify AI concepts and capabilities (40-45%); Implement AI solutions by using Microsoft Foundry (55-60%)
GenAI emphasis 20-25% of the exam Generative AI apps and a single-agent solution built in Foundry
Microsoft Foundry coverage Describe Azure AI Foundry and its model catalog The whole implementation domain (55-60%)
Multimodal and extraction Describe vision and speech workloads Multimodal prompts, image generation, and Content Understanding extraction from documents, images, audio and video
- Still counts toward the credential? Yes, through June 30, 2026 Yes, from launch onward
- Validity after passing Lifetime Lifetime
Which should you sit now? AI-900 is no longer available, so new candidates take AI-901 . If you already hold AI-900, do nothing — your certification remains valid indefinitely; there is no need to re-certify.
The biggest structural change: AI-900 spread across five describe-only domains, but AI-901 (objectives published April 15, 2026) collapses to just two — Identify AI concepts and capabilities (40-45%) and Implement AI solutions by using Microsoft Foundry (55-60%). Over half of AI-901 is now hands-on implementation inside Microsoft Foundry, which is why a light Python and portal-skills layer is expected.
What carries over: responsible AI principles, AI workload identification, text analysis, speech, computer vision, and core generative AI concepts all appear in AI-901's concept domain. What is new: the Foundry implementation domain (model deployment, prompts, the Foundry SDK, a single-agent solution, Azure Speech in Foundry Tools, image generation, and Content Understanding) and Python. AI-900's separate machine learning fundamentals domain (regression, classification, clustering, Azure Machine Learning) has no matching AI-901 objective.
AI-901 vs AWS AI Practitioner vs Google Cloud Generative AI Leader
All three exams are cloud-provider entry-level AI credentials, and AI-901 is now the Azure option. Pick by the cloud your employer uses or the one you want to build your career on. Here is the head-to-head:
- Dimension AI-901 (Microsoft Azure) AWS AI Practitioner (AIF-C01) Google Cloud Generative AI Leader
- Cost (USD) $99 $100 $99
- Duration 45 min exam / 65 min seat 90 minutes 90 minutes
- Questions Not published 65 (50 scored) 50-60
- Passing Score 700/1000 (scaled) 700/1000 (scaled) Not published
- Prereqs None (Python syntax knowledge expected) None None
- Validity Does not expire 3 years 3 years
GenAI coverage Generative AI concepts plus building apps and an agent in Foundry Fundamentals of GenAI 24%; Applications of Foundation Models 28% The whole exam is about generative AI
Hands-on building Yes (55-60% implementation in Foundry) No (coding AI/ML models is out of scope) No (business-level; open to non-technical roles)
Our take: Pick by the cloud your target employer uses. If you are on Azure, AI-901 is the only way to earn Azure AI Fundamentals now; expect basic Python and hands-on Foundry work, and note that Fundamentals certifications do not expire.
What To Do After Azure AI Fundamentals
Azure AI Fundamentals (AI-900 before June 30, 2026, AI-901 since) is a springboard, not a destination. Here are the natural next steps for most candidates:
AI-103: Azure AI Apps and Agents Developer Associate (role-based). The successor to AI-102 (Azure AI Engineer Associate), which Microsoft retired on June 30, 2026. It shifts the role-based track toward Microsoft Foundry, multi-agent orchestration, and production-grade AI application patterns.
AI-300: Machine Learning Operations Engineer Associate. The replacement for DP-100 (Azure Data Scientist Associate), which Microsoft retired on June 1, 2026. Best if you are moving toward ML engineering and ML operations.
AZ-900: Azure Fundamentals ($99). If you do not already have it, grab it — the fundamentals trio (AZ-900 + AI-901 + DP-900) is a classic resume combo for cloud-adjacent roles.
DP-900: Azure Data Fundamentals ($99). Closes the trio. Data concepts, Azure SQL, Cosmos DB, Synapse, Fabric.
If you do not hold Azure AI Fundamentals yet, start with AI-901 : since AI-900 retired on June 30, 2026, it is the only exam that earns the credential, and it adds light Python and Microsoft Foundry implementation skills to AI-900's conceptual coverage.
For career-switchers targeting AI product or consulting roles, the pragmatic path is AI-901 → portfolio project (build a RAG chatbot on Azure) → AI-103 → apply for roles . The portfolio project matters more than any additional certification.
Final Checklist (AI-901)
Read the Microsoft Learn AI-901 study guide (the skills measured document) end to end.
Memorized the six Responsible AI principles with one scenario each.
Memorized the Azure AI service → workload mapping table (Vision/Language/Speech/Translator/Document Intelligence/OpenAI).
Deployed a model in Microsoft Foundry and built a lightweight chat client with the Foundry SDK in Python.
Created and tested a single-agent solution in the Foundry portal.
Tried speech, vision, and Content Understanding extraction hands-on.
Comfortable explaining RAG , prompt engineering , and fine-tuning in plain English.
Completed Microsoft's free practice assessment and reviewed every wrong answer.
System-tested your environment if taking the exam online (webcam, bandwidth, quiet room).
Registered with a personal Microsoft Account , not a work account.
If you can tick every box, you are ready.
One more nudge: Try the free ExamFinch AI-901 practice questions — every question comes with an explanation, and you can ask the AI tutor follow-up questions until the concept clicks.
Official Sources and Further Reading
Microsoft Learn — Exam AI-900: Microsoft Azure AI Fundamentals (official exam page; retirement notice, final skills outline, languages)
- Microsoft Learn — AI-900 Study Guide (as of May 2, 2025) (canonical skills document used by exam writers)
- Microsoft Learn — Azure AI Fundamentals certification (certification overview and AI-901 transition)
- Microsoft Learn — Exam duration and exam experience (official 45-minute / 65-minute seat time source)
- Microsoft Learn — Exam scoring reports (700/1000 passing score reference)
- Microsoft — Responsible AI Standard (six principles canonical reference)
- Microsoft Tech Community — The AI Job Boom and AI-901 (AI-900 retirement and AI-901 rollout)
Microsoft Tech Community — Evolving the Azure AI Fundamentals Certification (official AI-901 beta announcement — April 15, 2026)
- Microsoft Learn — AI-901 Study Guide (current Azure AI Fundamentals skills outline, as of April 15, 2026)
- Microsoft Tech Community — New Azure AI Apps and Agents Developer Associate (AI-103) (AI-102 successor exam)
- Pearson VUE — Microsoft Certifications (exam delivery platform)
U.S. Bureau of Labor Statistics — Computer and Information Research Scientists (AI occupation outlook and salary)
- ExamFinch — Free AI-900 Practice Questions
- ExamFinch — AI-900 Study Guide