Last updated: 2026. Microsoft Azure AI Fundamentals (AI-900) is an entry-level exam that has changed alongside Azure’s AI terminology over the past several years, including the move from the older Cognitive Services naming to Azure AI Services.
Microsoft Azure AI Fundamentals, known by exam code AI-900, is an entry-level Microsoft certification for people who need to understand common AI workloads, responsible AI principles, and the Azure services used to build AI-enabled solutions. It is designed for foundational understanding rather than software engineering depth, which makes it relevant to business users, technical beginners, students, project teams, and professionals who need credible AI vocabulary without becoming machine learning specialists.
The certification is often treated as a first step into Microsoft AI, but it is most useful when the candidate is clear about why they are taking it. Someone evaluating AI use cases for a business process will use AI-900 differently from a developer planning to move toward Azure AI Engineer Associate. The value of the exam lies in connecting AI concepts to real Azure capabilities, such as extracting text from documents, classifying images, building language-aware applications, or adding search and generative AI features to an internal knowledge system.
What AI-900 Measures
AI-900 assesses whether a candidate can describe AI workloads and identify the Azure services that support them. It does not require formal prerequisites, and it should not be approached as a coding, statistics, or model-training exam. A common preparation mistake is spending too much time on Python, deep mathematics, MLOps, or DP-100-style machine learning operations while overlooking Responsible AI and the service capabilities that AI-900 actually tests.
The current scope is centred on practical recognition: what kind of workload is being described, which Azure AI service is relevant, and what responsible design considerations matter. Microsoft’s terminology now groups many of these capabilities under Azure AI Services. Candidates may still see older references to Cognitive Services in articles or legacy materials, but current preparation should use the newer Azure AI Services terminology where Microsoft uses it.
| AI-900 area |
What candidates should understand |
Real Azure example |
| AI workloads and Responsible AI |
Common AI use cases, fairness, reliability, privacy, transparency, and accountability considerations. |
Reviewing whether an automated claims triage tool should include human review for sensitive decisions. |
| Machine learning fundamentals |
Basic supervised and unsupervised learning ideas, model evaluation concepts, and when machine learning is appropriate. |
Recognising that a product recommendation model learns from historical customer behaviour. |
| Computer vision |
Image analysis, object detection, optical character recognition, and document extraction concepts. |
Using Azure AI Document Intelligence to prototype invoice data extraction from scanned documents. |
| Natural language processing |
Language understanding, sentiment analysis, key phrase extraction, translation, and conversational AI concepts. |
Designing a basic support chatbot that classifies customer intent before routing a request. |
| Knowledge mining and search |
How search indexes and enrichment pipelines help users find information across content repositories. |
Creating a searchable internal policy portal using Azure AI Search concepts. |
| Generative AI fundamentals |
How Azure OpenAI Service fits into solution design at a conceptual level, including responsible use considerations. |
Identifying where a summarisation feature could help employees navigate long documents. |
Terminology note: Microsoft has revised product naming and exam language over time. Preparation materials should be checked against the official AI-900 skills outline so that older “Cognitive Services” wording is understood as legacy terminology rather than the preferred current label.
Who Should Take AI-900
AI-900 suits people who need to participate in AI conversations with accuracy. That includes business analysts shaping use cases, product owners evaluating AI features, service desk or operations staff supporting AI-enabled tools, students building a credential base, and technical professionals who want a low-risk introduction before taking a role-based certification.
It is also useful for managers and team leads who need a shared baseline across mixed technical and non-technical teams. In many organisations, the first AI implementation challenge is not model design. It is agreeing on what the technology can and cannot do, which services are appropriate, and where human oversight, data quality, and governance need attention. AI-900 gives that conversation a common structure.
AI-900 is not the right target for someone who already needs to design, implement, and maintain production Azure AI solutions in depth. That path points more naturally toward role-based credentials after the fundamentals are understood. Likewise, candidates whose main goal is cloud infrastructure literacy may be better served by Azure fundamentals first, while candidates focused on databases, analytics, and data workloads may need a data fundamentals route.
AI-900, AZ-900, or DP-900: Choosing the Right Starting Point
The three Microsoft fundamentals exams are often confused because they all sit near the beginning of a learning path. The distinction is straightforward: AI-900 validates foundational AI concepts and awareness of Azure AI Services such as Vision, Language, Search, Azure OpenAI, and Responsible AI. AZ-900 targets cloud concepts and core Azure services. DP-900 targets core data concepts and Azure data services.
AI-900 is the better first step when the outcome is to understand AI workloads, evaluate AI use cases, or communicate with teams building AI-enabled applications. AZ-900 is usually a stronger starting point when the candidate lacks cloud vocabulary altogether and needs to understand regions, subscriptions, compute, networking, identity, pricing concepts, and governance. DP-900 is more relevant when the immediate work involves relational data, analytics platforms, data storage, or reporting foundations.
From a practical perspective, a business analyst working on chatbot requirements can start with AI-900. A support technician moving into Azure administration should consider Azure fundamentals through a broader Microsoft Azure route, such as Microsoft Azure training options. A reporting analyst who needs to understand data stores and analytics concepts before AI should prioritise data foundations before moving deeper into AI services.
Exam Day: What to Expect
Microsoft certification exams are delivered through Microsoft’s exam delivery partners, with options that may include test centre and online proctored delivery depending on location and current availability. Candidates should confirm the current delivery options, exam language, rescheduling rules, identification requirements, and accessibility accommodations on the official Microsoft exam page before booking. These policies can change and may vary by region.
The AI-900 exam uses Microsoft’s standard certification scoring approach, with a scaled passing score of 700. A scaled score means the result is not a simple percentage of correct answers. Microsoft may use different question formats, including multiple choice, matching, case-style scenarios, and tasks that ask candidates to select the most appropriate service or concept for a described business need.
Remote proctoring requires more preparation than many first-time candidates expect. The candidate should plan for a quiet room, a stable internet connection, a clear desk, valid identification, and enough time for check-in. During the exam, reading the scenario carefully matters because AI-900 questions often test the difference between similar services. For example, a question about extracting fields from invoices points in a different direction from a question about detecting objects in images or analysing customer sentiment.
Retake rules, waiting periods, and accommodation processes should be reviewed on Microsoft’s official certification pages rather than inferred from older blog posts. The safest approach is to treat the exam booking page as the source of record for current policies and to prepare ethically using official skills outlines, Microsoft Learn content, practice assessments, and structured review.
How to Prepare Without Overstudying the Wrong Material
The most efficient preparation starts with the skills measured document and then uses Microsoft Learn modules to fill knowledge gaps. Candidates should be able to describe each workload in plain English, identify which Azure service category fits a scenario, and explain why Responsible AI is part of the design conversation. Memorising product names without understanding use cases is a weak strategy because many questions are scenario-led.
A useful milestone is the ability to explain an AI use case to a non-specialist. For instance, a candidate should be able to say why Document Intelligence is relevant to OCR and form extraction, why Azure AI Language is relevant to sentiment or key phrase analysis, why Azure AI Search helps with knowledge discovery, and why Azure OpenAI Service introduces additional responsible use considerations. That level of explanation is closer to the spirit of AI-900 than writing code against an API.
| Candidate background |
Preparation approach |
Readiness check |
| Non-technical or business-focused |
Start with AI vocabulary, Responsible AI, and service use cases before looking at product details. |
Can match a business scenario to Vision, Language, Search, Azure OpenAI, or machine learning concepts. |
| Technical but new to AI |
Move faster through cloud basics and spend more time distinguishing similar AI services and workloads. |
Can explain why one Azure AI service is more suitable than another for a given requirement. |
| Student or early-career learner |
Combine Microsoft Learn with short notes, practice questions, and examples from common workplace use cases. |
Can answer scenario questions without relying on memorised definitions alone. |
Structured training can help when the candidate wants a guided route through the blueprint rather than assembling resources independently. Readynez offers an AI-900 Azure AI Fundamentals course that focuses on the certification scope, but candidates should still compare any course content with the current Microsoft skills outline before sitting the exam.
Sample Question Styles
AI-900 questions are usually less about performing technical configuration and more about recognising the right concept or service from a scenario. The following examples are illustrative study prompts rather than exam questions.
Scenario: A company wants to extract supplier names, invoice numbers, and totals from scanned invoices. Which Azure AI capability is most relevant?
What this tests: The candidate should recognise document extraction and OCR-style processing rather than general chatbot or translation functionality.
Scenario: A retailer wants to understand whether recent customer reviews are positive, neutral, or negative. Which AI workload is being described?
What this tests: The candidate should identify natural language processing and sentiment analysis concepts.
Scenario: A team is planning an AI system that could affect customer eligibility decisions. Which concern should be considered early?
What this tests: The candidate should connect the scenario to Responsible AI principles such as fairness, transparency, reliability, and human oversight.
What AI-900 Can Lead To
After AI-900, the next step depends on the learner’s role rather than on a fixed certification ladder. Someone who discovered gaps in cloud fundamentals may benefit from studying broader Azure concepts. Someone whose work is data-heavy may need data fundamentals before moving further into analytics or AI. A developer, consultant, or engineer who wants to build Azure AI solutions may eventually look toward role-based AI certification after gaining hands-on experience.
The key is to avoid using AI-900 as proof of implementation depth. It is a foundation credential. It can support career conversations, project participation, and further learning, but it does not replace experience with solution design, identity and security, deployment practices, monitoring, data governance, or production operations.
FAQ
Are there prerequisites for AI-900?
There are no formal prerequisites for AI-900. Basic familiarity with technology and cloud concepts can help, but programming, advanced statistics, and machine learning engineering experience are not required for the fundamentals exam.
Does AI-900 require coding?
No. AI-900 focuses on describing AI workloads, Responsible AI principles, and Azure AI service capabilities. Candidates may benefit from seeing examples of how services are used, but the exam is not designed to test coding skill.
What topics are covered in AI-900?
The exam covers foundational AI workloads, machine learning concepts, computer vision, natural language processing, knowledge mining and search, generative AI concepts, Azure AI Services, and Responsible AI. Candidates should always check Microsoft’s current skills outline for the latest wording and scope.
Is AI-900 suitable for business professionals?
Yes. AI-900 is particularly relevant for business professionals who need to evaluate AI use cases, work with technical teams, or understand how Azure AI Services support common workplace scenarios.
How should someone prepare for AI-900?
A sound preparation route is to read the official skills measured outline, complete the relevant Microsoft Learn modules, practise scenario-style questions, and review weak areas against real examples such as document extraction, sentiment analysis, search enrichment, and responsible AI design choices.
References and Current Sources
The most reliable sources for exam details are Microsoft’s official Exam AI-900 page, the current AI-900 skills measured outline, Microsoft Learn modules for Azure AI Fundamentals, and Microsoft certification policy pages covering exam delivery, scoring, retakes, identification, and accommodations. These sources should be checked close to the booking date because Microsoft can revise terminology, policies, and exam scope.
Building the Right AI Foundation
AI-900 is a useful credential when it is treated as a foundation for better judgement about AI, not as a shortcut to advanced engineering work. The candidate who prepares well will understand the main AI workload types, recognise the Azure services associated with them, and explain why Responsible AI is part of every credible AI discussion.
A practical next step is to compare the AI-900 skills outline with current job or project responsibilities and close the most relevant gaps first. Learners planning a broader Microsoft certification path can also explore Unlimited Microsoft Training, while those who want guidance on the right route can contact Readynez with questions about preparing for Azure AI Fundamentals.
