Microsoft · AI-900
Validates foundational knowledge of machine learning and AI concepts and related Microsoft Azure services. Designed for candidates with both technical and non-technical backgrounds.
Practice Questions
464
≈ 9 practice exams
Duration
45 minutes
Passing Score
700/1000
Difficulty
FundamentalsLast Updated
Jan 2025
Use this AI-900 practice exam to prepare for Microsoft Certified: Azure AI Fundamentals (AI-900) with realistic questions, detailed explanations, and focused study modes. The practice bank includes 464 questions for Microsoft AI-900, so you can review the exam steadily instead of relying on one long cram session.
As you practice, pay extra attention to recurring topics such as Artificial Intelligence workloads and considerations, Machine learning principles on Azure, Computer vision workloads on Azure, Natural Language Processing workloads on Azure, and Generative AI workloads on Azure. Start with short sessions to identify weak areas, then move into timed quizzes once your accuracy is consistent.
The explanations are especially useful when you want to connect exam wording to the responsibilities and scenarios described in the official certification guidance. Use the free preview first, then unlock the full question bank when you are ready to build a complete study routine.
The Microsoft Certified: Azure AI Fundamentals certification (AI-900) validates foundational knowledge of machine learning and artificial intelligence concepts, along with the Microsoft Azure services that support them. It covers the full breadth of modern AI workloads — from classical machine learning techniques such as regression, classification, and clustering, to computer vision capabilities like image classification and optical character recognition, to natural language processing services including sentiment analysis and speech recognition, and through to generative AI powered by large language models via Azure OpenAI Service and Azure AI Foundry. The certification was most recently updated on May 2, 2025, with an increased emphasis on generative AI (now weighted at 20–25%), reflecting industry demand. Note: Azure AI Foundry is in the process of being rebranded to Microsoft Foundry, and associated exam materials are being updated accordingly.
This is a fundamentals-level, non-expiring certification priced at approximately $99 USD (varies by region). It is delivered through Pearson VUE for general candidates, or through Certiport for students and educators. The certification serves as a stepping stone toward more advanced role-based credentials such as Azure AI Engineer Associate (AI-102) and Azure Data Scientist Associate (DP-100), though it is not a formal prerequisite for either.
AI-900 is explicitly designed for candidates from both technical and non-technical backgrounds, requiring no prior data science or software engineering experience. It is well-suited for IT professionals, developers, and cloud practitioners who want to establish a formal AI foundation, as well as business stakeholders — project managers, sales professionals, marketing specialists, and executives — who need to understand AI capabilities in order to collaborate with technical teams or identify AI opportunities within their organizations.
Job roles that commonly pursue this certification include aspiring AI Engineers, Data Analysts, Cloud Support Associates, Solutions Architects, and students entering technology fields. It is also valuable for professionals currently working in Azure environments who want to formalize their understanding of Azure's AI and ML service portfolio before pursuing the AI-102 or DP-100 credentials.
There are no mandatory prerequisites for AI-900. Microsoft does not require any prior certification or formal coursework before scheduling the exam. However, candidates benefit from a basic awareness of cloud computing concepts (ideally familiarity with Azure fundamentals) and a general understanding of how client-server applications function. These are not hard requirements, but they help contextualize the Azure-specific content on the exam.
For practical preparation, Microsoft recommends completing the official AI-900 self-paced learning path on Microsoft Learn, which is free and structured around the five exam domains. Candidates with no prior cloud exposure may wish to complete the AZ-900 (Azure Fundamentals) content first, though this is optional. Most candidates with basic technology literacy can prepare adequately in one to two weeks using the official materials.
The AI-900 exam consists of a variable number of questions (Microsoft does not publish a fixed count; typical sittings include approximately 40–60 questions) and must be completed within a 45-minute time limit. The exam may include multiple question types such as multiple choice, drag-and-drop, matching, and interactive scenario-based items. Candidates should expect some questions on Preview Azure features that are in common use, in addition to Generally Available (GA) features.
The exam is delivered online through Pearson VUE (with remote proctoring available) or at a Pearson VUE testing center. Students and educators may schedule through Certiport. A score of 700 or higher on a scale of 100–1000 is required to pass. If a candidate fails, a retake can be scheduled 24 hours after the first attempt; subsequent retake waiting periods vary per Microsoft's exam retake policy. The AI-900 is a fundamentals-level certification and does not expire, so no annual renewal is required.
AI-900 functions primarily as a credential signaling foundational AI literacy, making it valuable both as a standalone certification for non-technical professionals and as a launch point for deeper technical paths. For business-oriented roles — product managers, consultants, sales engineers, and executives — it provides the vocabulary and conceptual framework to evaluate AI solutions, communicate with technical teams, and contribute to AI strategy. For technical candidates, it establishes a documented baseline before pursuing the Azure AI Engineer Associate (AI-102), which commands salaries in the $120,000–$175,000+ range for experienced practitioners.
Job roles commonly associated with AI certifications on the Azure track include AI Engineer, Machine Learning Engineer, Data Scientist, Cloud Solutions Architect, and Data Analyst. Salary ranges for Azure AI-related roles broadly span $96,900 to over $200,000 depending on seniority, specialization, and geography. While AI-900 alone does not qualify candidates for senior technical positions, it is a recognized signal of initiative and foundational knowledge in a hiring market where 81% of hiring managers, according to recent industry surveys, report prioritizing demonstrated AI skills during candidate screening. The certification does not expire, so it retains its value without requiring periodic renewal exams.
5 sample questions with answers and explanations. The full bank has 464 questions, enough for 9 full-length practice exams.
Preview — answers shown1. When analyzing a typical grocery store receipt using Azure AI Document Intelligence's prebuilt receipt model, which of the following pieces of information is the service well-equipped to recognize and extract as distinct fields?
Explanation
The prebuilt receipt model in Azure AI Document Intelligence is trained to recognize common fields found on receipts, such as the merchant's name and their physical address. It also extracts items, totals, dates, etc. Merchant retail type might be inferred but isn't a primary extracted field. Date of incorporation is generally not on a sales receipt.
2. The Azure AI Language service's 'Personally Identifiable Information (PII) detection' feature is used for what primary purpose?
Explanation
The PII detection feature is a critical tool for privacy and compliance. It is designed to scan text and identify, categorize, and even redact various types of personally identifiable information (PII) to help organizations protect sensitive data and comply with privacy regulations like GDPR.
3. OnlineGuardian Corp provides comment moderation services for news websites. They need to automatically detect and flag user comments that contain offensive language, threats, or incite violence. Which specific capability of Azure AI Content Safety is designed to analyze textual input and identify such potentially harmful content based on categories like 'Hate' or 'Violence'?
Explanation
Azure AI Content Safety's text analysis API is specifically built to detect harmful content within text across various categories, such as hate speech, violence, self-harm, and sexual content. It provides severity scores for detected content, enabling automated flagging or moderation actions. Image analysis is for visuals. Jailbreak detection is for LLM misuse. Translation and copyright detection are different functionalities.
4. HealthMonitor Corp develops wearable devices that continuously track heart rate, sleep patterns, and activity levels to detect early signs of health issues. What type of AI application is this primarily?
Explanation
Real-time anomaly detection and health monitoring is the primary application because wearable devices need to continuously analyze streaming physiological data to identify unusual patterns that might indicate health issues. This requires real-time processing capabilities and anomaly detection algorithms that can work with continuous sensor data streams.
5. GadgetMart is preparing a dataset of smartphone features and their price ranges to train a classification model. One of the features in their dataset is 'Color', but they observe that every single phone in their dataset is listed as 'black'. During the feature selection phase of data pre-processing, what is the most appropriate action to take regarding the 'Color' feature?
Explanation
If a feature, like 'Color' in this case, has the exact same value (e.g., 'black') for every single instance in the dataset, it provides no information to help distinguish between different price ranges or outcomes. Such a constant feature has no predictive power. Therefore, the most appropriate action during feature selection is to discard the 'Color' feature from the dataset used for training the model. Prioritizing a constant feature is incorrect because it offers no predictive value—consistency across all samples means there's no variation to learn from. Converting the feature to a numerical value or using one-hot encoding are standard methods for handling categorical data, but they are irrelevant if the feature is constant and offers no variability. These encoding techniques are designed to represent different categories, but when only one category exists, they provide no discriminative information. Heavily weighting a non-informative feature would be detrimental to model performance, as it would give undue importance to a feature that cannot contribute to distinguishing between different outcomes or classes.
Microsoft Certified: Agentic AI Business Solutions Architect (AB-100)
AB-100 · 700 questions
Microsoft Certified: Azure Administrator Associate (AZ-104)
AZ-104 · 757 questions
Microsoft Certified: Azure AI Engineer Associate (AI-102)
AI-102 · 1392 questions
Microsoft Azure AI Fundamentals (AI-901)
AI-901 · 600 questions
Microsoft Certified: Azure Cosmos DB Developer Specialty (DP-420)
DP-420 · 677 questions
Microsoft Certified: Azure Data Fundamentals (DP-900)
DP-900 · 381 questions
$17.99
One-time access to this exam