AI-900 Certification: How to Pass in 30 Days
The AI-900 (Azure AI Fundamentals) is the entry-level Microsoft AI certification — the one most people consider when they want a credential that says "I understand AI" without committing to the deep technical track. It is not difficult. With 30 to 50 hours of focused study, anyone with a working knowledge of cloud services and basic IT terminology can pass on the first attempt. What follows is the 30-day study plan that consistently gets first-time candidates over the 700-mark threshold.
Table of contents
- What AI-900 actually tests
- 30-day study plan
- Free resources
- Practice exams
- Where this cert opens doors
- Where it doesn't
- Frequently asked questions
- The bottom line
What AI-900 actually tests
The AI-900 exam is 60 minutes, 40 to 60 multiple-choice and matching questions, with a pass mark of 700 out of 1000. The cost is around $99 USD (regional pricing applies). It can be taken online from home with a proctor or in person at a Pearson VUE test centre.
The exam is divided into five domains. As of the May 2025 exam refresh, the weights are roughly:
| Domain | Weight | What you need to know |
|---|---|---|
| Describe Artificial Intelligence workloads and considerations | 15-20% | Categories of AI workloads, responsible AI principles (fairness, reliability, privacy, inclusiveness, transparency, accountability) |
| Fundamentals of machine learning | 20-25% | Regression vs classification vs clustering, supervised vs unsupervised, training/validation/testing splits, Azure Machine Learning service |
| Computer vision on Azure | 15-20% | Image classification, object detection, OCR, face detection, custom vision, Azure AI Vision service capabilities |
| Natural language processing on Azure | 15-20% | Sentiment analysis, key phrase extraction, language detection, translation, Speech service, Azure AI Language service |
| Generative AI on Azure | 20-25% | What an LLM is, transformer basics, Azure OpenAI service, prompt engineering, Copilot relationship to Azure AI |
The largest expansion in the May 2025 update was generative AI — it went from a small bonus topic to a full domain. Most pre-2024 study guides under-emphasise this; check your materials are dated 2025 or later.
30-day study plan
The plan below assumes 60 to 90 minutes per day, five days a week. It works for someone with a basic IT background and no prior AI exposure.
Week 1 — Foundations. Watch the official Microsoft Learn AI-900 learning path end to end (about 8 hours of video and reading). Don't take notes; the goal is exposure to the vocabulary. Cover the AI workloads, responsible AI principles, and basic ML concepts modules.
Week 2 — Computer Vision and NLP. Re-watch the Computer Vision and NLP modules from Microsoft Learn, this time taking notes on each Azure service and what it does. Specifically, get comfortable with which service does what: Azure AI Vision (general image), Azure AI Custom Vision (your own classifier), Azure AI Document Intelligence (forms and OCR), Azure AI Language (text analysis), Azure AI Speech (audio).
Week 3 — Generative AI and Azure OpenAI. The newest and largest exam domain. Cover the Microsoft Learn modules on Azure OpenAI service, the difference between completion and chat models, prompt engineering basics, and how Copilot fits the Azure AI stack. This week alone is worth roughly 25% of exam questions.
Week 4 — Practice and weak spots. Take a full practice exam early in the week (MeasureUp or the official Microsoft practice assessment). Whatever you scored below 70% on becomes your target list. Re-study those modules and take a second practice test at the end of the week. If your second practice test is over 750, book the real exam.
The candidates who fail AI-900 typically failed because they did not work through Microsoft Learn end to end. The exam writers test exactly the vocabulary and conceptual model in those modules — independent study materials are useful as supplements, not as substitutes.
One specific study tip that has held up across thousands of candidate reports: write your own one-page cheat sheet by hand at the end of each week. Active recall is much more effective than re-reading. The cheat sheet should fit on a single side of paper and cover the highest-leverage facts: the six responsible AI principles, the difference between regression/classification/clustering, the mapping between Azure AI services and what they do, and the prompt engineering vocabulary (zero-shot, few-shot, chain-of-thought). If you can reproduce that cheat sheet from memory the day before the exam, you will pass.
Free resources
The honest answer is that the official Microsoft Learn path is enough on its own to pass AI-900. Free, self-paced, and consistently updated to match the exam. The full path is at learn.microsoft.com/training/paths/get-started-with-artificial-intelligence-on-azure/.
The supplements that add value: John Savill's free AI-900 video on YouTube (4-hour focused review, popular and accurate); the Microsoft Virtual Training Day for AI-900 (free live session run by Microsoft, includes a free exam voucher if you attend in full); and the official Microsoft Q&A site for any specific question you can't resolve.
The supplements that don't add value: most paid Udemy AI-900 courses repeat what Microsoft Learn already covers. If you're going to pay, pay for practice exam access (covered next), not video re-treads.
The one paid resource consistently worth the money is access to a Microsoft Learn sandbox subscription if you don't already have one. The hands-on labs in Microsoft Learn require an active Azure subscription; the free trial is enough to complete the AI-900 labs but expires quickly. For a one-off cost of around $50 in pay-as-you-go usage, you can complete the relevant labs at your own pace and avoid the trial-clock pressure that catches some candidates. The hands-on practice with the actual Azure portal is also genuinely useful for the exam — questions occasionally test whether you've actually navigated the AI services or just read about them.
Practice exams
The official Microsoft AI-900 practice assessment (free at learn.microsoft.com) is the single most useful study tool. It is not the same questions as the real exam, but it is the same style and depth. Score 80% or above on the official practice and you are very likely to pass the real thing.
Paid practice exam providers worth considering: MeasureUp (Microsoft's official practice partner — closest to real exam style), Whizlabs, and Tutorials Dojo. Avoid the very cheap braindumps; they are often outdated, frequently inaccurate, and using them violates the exam policy. Microsoft has revoked certifications from candidates caught using leaked questions.
The metric to watch on practice exams: not your overall score but your score on the generative AI domain. This is the newest and most heavily weighted area, and the area where most outdated study guides fall short.
Where this cert opens doors
AI-900 is genuinely useful in three contexts. For non-technical roles in AI-adjacent jobs — sales, customer success, product management, marketing — it is a credible signal that you understand the vocabulary well enough to have intelligent conversations with technical teams. Recruiters at Microsoft partner organisations specifically look for it on CVs.
For internal IT staff being repositioned toward AI work, AI-900 is the foundation cert. Pair it with the AI-102 (Azure AI Engineer Associate) within 6-12 months and you have a credible AI engineering profile, which is genuinely in demand at $100K+ salary levels in 2026 markets like the US, UK, and Western Europe.
For trainers, consultants, and technical educators, AI-900 is the table-stakes credential to teach AI fundamentals workshops. Most corporate training contracts require at least one trainer per cohort to hold relevant certifications; AI-900 satisfies this for most "intro to AI" mandates. For more on the broader career pathway, see our AI careers hub.
The honest signal AI-900 sends to a hiring manager is "this person took the time to learn the basics formally." That signal is more valuable in mid-career professionals pivoting toward AI than in junior candidates, where it can look like resume padding. For a marketing manager moving into product marketing for an AI tool, AI-900 is genuinely useful. For a fresh CS graduate, the cert is below the bar; the time is better spent building something. The role you are targeting matters more than the cert itself.
Where it doesn't
AI-900 will not make you a machine learning engineer or data scientist. It is not a credential that justifies a senior salary on its own. It will not impress recruiters at AI-first companies (OpenAI, Anthropic, DeepMind hire on demonstrated work, not certifications). It will not get you past a technical interview at any company that takes AI seriously as a discipline.
It is an entry-level credential. Treat it as such. The right framing on a CV is "AI-900 (Azure AI Fundamentals), [date]" listed alongside other foundational certs, not as the headline credential. The credentials that carry weight in AI hiring in 2026 — production model deployments, contributions to open-source ML libraries, papers, real-world AI products you've shipped — none of these are replaced by a multiple-choice exam.
That said, the cost-benefit is favourable. $99 and 30 hours for a credible signal that you understand the vocabulary is a fair trade in most professional contexts.
One more honest point about where AI-900 falls short: it does not cover the actual operational realities of running AI in production. Cost optimisation, model evaluation, hallucination management, prompt versioning, evals, RAG architecture, vector database selection — none of these are on the AI-900 syllabus. They are on AI-102 in part, but the rigorous treatment is on the AWS, Google, and OpenAI certifications, not the Microsoft fundamentals track. If your role involves shipping AI features rather than discussing them, AI-900 is not enough; pair it with practical experience or stack on AI-102 within 6 to 12 months.
The final framing that matters: certifications are a substitute for nothing, including for understanding how to use Copilot effectively in your own work. The cert proves you can pass an exam about AI vocabulary; it does not prove you can write a useful prompt or judge when AI output is wrong. Both skills matter. Don't conflate them.
Frequently asked questions
Is AI-900 hard?
No. With 30 to 50 hours of focused study working through Microsoft Learn and one or two practice exams, the typical first-time pass rate is well above 70%. The exam tests vocabulary recognition and conceptual understanding, not technical depth. There is no coding, no math, and no live Azure portal navigation — purely multiple choice and matching. People who fail typically did not complete Microsoft Learn modules end to end.
How long does the AI-900 certification last?
Microsoft fundamentals certifications (the AI-900, MS-900, AZ-900 family) do not expire. Once earned, AI-900 is yours permanently. This contrasts with associate-level and expert-level Microsoft certifications, which require annual renewal via free online assessment. The fundamentals don't.
Can I retake AI-900 if I fail?
Yes. If you fail, you can retake the exam after 24 hours. After a second failure, you must wait 14 days. After a third or subsequent failure, the wait is also 14 days, with a maximum of five attempts per year. Each attempt costs the full exam fee — there is no retake voucher. Most candidates who fail the first attempt and then study weak areas pass on the second try.
Should I take AI-900 or AI-102?
AI-900 if you have no prior AI background or if you're in a non-technical role. AI-102 if you're a developer or engineer planning to build AI services and you already have working Azure familiarity (AZ-204 level or equivalent). AI-102 is significantly harder, requires Python coding experience, and tests hands-on Azure AI service implementation. Most career paths benefit from passing AI-900 first as a stepping stone, then AI-102 for the technical credibility.
Does AI-900 cover Microsoft Copilot?
Yes — the May 2025 exam refresh added explicit coverage of Microsoft Copilot, Copilot Studio, and Azure OpenAI service. About 20-25% of exam questions touch generative AI topics that include Copilot architecture and use cases. If your study guide is dated 2024 or earlier, it is missing this material; refer to current Microsoft Learn modules.
The bottom line
AI-900 is a 30-hour, $99 credential that signals you understand AI vocabulary. It is genuinely useful for non-technical professionals moving toward AI-adjacent roles, internal IT staff repositioning, and trainers/consultants who need a defensible credential to deliver AI workshops. It is not a substitute for technical depth or hands-on experience.
The 30-day plan above — Microsoft Learn end to end in week 1, deep dives on vision/NLP in week 2, generative AI in week 3, practice exams in week 4 — works reliably. Stick to the official material, prioritise the generative AI domain, and book the exam as soon as you score over 80% on a practice test.
For the broader Microsoft Copilot picture beyond certification, see our complete training guide. For other AI career pathways, the AI careers hub covers learning routes for data scientists, ML engineers, and AI product managers.
Last updated: January 2026