AI-900
AI-900
Microsoft Azure
AI Fundamentals
Microsoft Azure AI Fundamentals (AI‑900) teaches core AI concepts, machine‑learning basics, computer‑vision, NLP, and generative AI services on Azure, enabling learners to identify appropriate solutions and articulate their business value.
85
Minutes
50
Questions
700/1000
Passing Score
$99
Exam Cost
10
Languages
Who Should Take This
It is ideal for recent graduates, junior data analysts, or IT professionals who have limited exposure to AI and want to build a solid foundational knowledge of Azure AI services. These learners aim to validate their understanding through the AI‑900 certification, preparing them for roles that involve AI solution recommendation, data‑driven decision making, and cross‑functional collaboration.
What's Covered
1
Fundamental AI concepts including responsible AI principles, common AI workloads, and guiding principles for AI development on Azure.
2
Core ML concepts including supervised, unsupervised, and reinforcement learning; Azure Machine Learning capabilities and automated ML.
3
Computer vision solutions on Azure including image classification, object detection, optical character recognition, and Azure AI Vision service.
4
NLP capabilities on Azure including text analytics, language understanding, speech services, and Azure AI Language service.
5
Generative AI fundamentals including large language models, Azure OpenAI Service, prompt engineering, and responsible generative AI use.
Exam Structure
Question Types
- Multiple Choice
- Multiple Response
- Drag-And-Drop
Scoring Method
Scaled score 100-1000, passing score 700
Delivery Method
Proctored exam, 40-60 questions, 45 minutes
Recertification
Fundamentals certifications do not expire
What's Included in AccelaStudy® AI
Adaptive Knowledge Graph
Practice Questions
Lesson Modules
Console Simulator Labs
Exam Tips & Strategy
20 Activity Formats
Course Outline
72 learning goals
1
Describe AI workloads and considerations
2 topics
Identify common AI workloads
- Define artificial intelligence and describe how AI enables systems to perform tasks that typically require human intelligence including perception, reasoning, learning, and decision-making.
- Identify common AI workloads including prediction, anomaly detection, computer vision, natural language processing, and conversational AI and describe the purpose of each workload type.
- Identify generative AI workloads including text generation, image generation, and code generation and explain how they differ from traditional predictive AI workloads.
- Explain how document intelligence and knowledge mining workloads extract structured information from unstructured content using form recognition, data extraction, and cognitive search indexing.
- Apply knowledge of AI workload types to match business problem descriptions to the most appropriate AI workload category and justify the selection based on input data and desired output.
- Differentiate between prediction, classification, anomaly detection, and generative AI workloads by comparing their input requirements, output types, and suitable business scenarios.
Identify guiding principles for responsible AI
- Describe the six responsible AI principles (fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability) and state the purpose of each.
- Explain how fairness and inclusiveness principles guide the identification and mitigation of bias in AI systems to ensure equitable outcomes across diverse user populations.
- Explain how transparency and accountability principles require AI systems to provide understandable decisions, maintain audit trails, and establish human oversight governance structures.
- Explain how reliability, safety, privacy, and security principles guide the design of AI systems that perform consistently and protect user data throughout the AI lifecycle.
- Apply responsible AI principles to evaluate a given AI deployment scenario and identify which principles are at risk and what mitigation strategies should be implemented.
- Analyze tradeoffs between responsible AI principles when they conflict in a given scenario and determine which safeguards take priority based on the context, stakeholder impact, and risk level.
2
Describe fundamental principles of machine learning on Azure
4 topics
Identify common machine learning types
- Describe regression as a supervised learning technique that predicts continuous numeric values and identify appropriate use cases such as price prediction and demand forecasting.
- Describe classification as a supervised learning technique that predicts categorical labels and distinguish between binary classification and multiclass classification with relevant use cases.
- Describe clustering as an unsupervised learning technique that groups similar data points without predefined labels and identify appropriate use cases such as customer segmentation.
- Apply knowledge of ML types to select the appropriate learning approach (regression, classification, or clustering) for a given business scenario based on data characteristics and desired output.
- Analyze a dataset description to determine whether supervised or unsupervised learning is more appropriate and evaluate which specific ML technique best fits the prediction or grouping requirement.
Describe core machine learning concepts
- Define features and labels in machine learning and explain how they serve as inputs and outputs during model training and prediction workflows.
- Explain the purpose of training data and validation data and demonstrate how splitting datasets prevents overfitting and enables accurate model performance assessment.
- Explain the iterative model training process including data preparation, feature selection, algorithm selection, training, evaluation, and deployment as a cyclical workflow.
- Describe deep learning as a subset of machine learning that uses artificial neural networks with multiple layers to learn complex patterns from large datasets.
Describe training and evaluating models
- Identify evaluation metrics for regression models including mean absolute error, root mean squared error, and coefficient of determination and state what each metric measures.
- Identify evaluation metrics for classification models including accuracy, precision, recall, F1 score, and the confusion matrix and explain what each metric reveals about model performance.
- Explain overfitting and underfitting in machine learning models and describe strategies to mitigate each including cross-validation, regularization, and increasing training data volume.
- Analyze model evaluation results to determine whether a model is performing adequately and assess the most appropriate metric to prioritize based on the business cost of errors.
Describe Azure Machine Learning capabilities
- Describe Azure Machine Learning workspace capabilities including automated ML, designer (drag-and-drop), notebooks, and the model registry for the end-to-end ML lifecycle.
- Explain how Azure Machine Learning automated ML simplifies model building by automatically selecting algorithms, tuning hyperparameters, and evaluating model performance for given datasets.
- Explain how Azure Machine Learning data assets and compute assets are configured to provide training data access and scalable compute resources for model training pipelines.
- Determine the appropriate Azure Machine Learning tool (automated ML, designer, or notebooks) for a given user skill level, project complexity, and customization requirement.
3
Describe features of computer vision workloads on Azure
2 topics
Identify common types of computer vision solutions
- Describe image classification as a computer vision task that assigns a category label to an entire image and identify appropriate use cases such as product categorization and medical imaging.
- Describe object detection as a computer vision task that identifies and locates multiple objects within an image using bounding boxes and class labels.
- Explain how optical character recognition extracts printed and handwritten text from images and documents and demonstrate how OCR enables downstream document processing workflows.
- Explain facial detection and facial analysis capabilities including detecting faces in images, identifying facial attributes, and how detection differs from recognition and verification.
- Evaluate a business scenario involving visual data to select the most appropriate computer vision solution type by comparing image classification, object detection, OCR, and facial analysis capabilities.
Identify Azure tools and services for computer vision
- Describe Azure AI Vision service capabilities including image analysis, spatial analysis, image tagging, caption generation, and smart cropping for prebuilt computer vision tasks.
- Explain how Azure AI Custom Vision enables training custom image classification and object detection models with project-specific labeled training images for domain-specific visual analysis.
- Explain Azure AI Face service capabilities including face detection, verification, identification, and grouping and describe the ethical considerations and access restrictions for facial recognition.
- Explain how Azure AI Document Intelligence extracts text, key-value pairs, tables, and structure from forms and documents using prebuilt and custom models for automated data processing.
- Evaluate a business scenario to determine the appropriate Azure computer vision service (AI Vision, Custom Vision, Face, or Document Intelligence) based on task type, domain specificity, data availability, and customization needs.
4
Describe features of Natural Language Processing workloads on Azure
3 topics
Identify features of common NLP solutions
- Describe key phrase extraction and named entity recognition as NLP techniques that identify important terms and classify entities such as people, locations, and organizations in text.
- Describe sentiment analysis as an NLP technique that determines the emotional tone of text including positive, negative, neutral, and mixed sentiments for customer feedback and social media analysis.
- Explain language modeling concepts including tokenization, embeddings, and how language models represent and process text to enable downstream NLP tasks.
- Explain how speech recognition converts spoken language to text and how speech synthesis converts text to natural-sounding speech and describe common use cases for each capability.
- Explain how machine translation converts text between languages and describe the role of neural machine translation models in improving translation accuracy and fluency.
- Evaluate a business scenario involving unstructured text data to select the most appropriate NLP technique by comparing key phrase extraction, entity recognition, sentiment analysis, and translation capabilities.
Identify Azure tools and services for NLP workloads
- Describe Azure AI Language service capabilities including key phrase extraction, named entity recognition, sentiment analysis, language detection, and question answering features.
- Describe Azure AI Speech service capabilities including speech-to-text transcription, text-to-speech synthesis, speech translation, and speaker recognition features.
- Describe Azure AI Translator service capabilities including text translation, document translation, transliteration, and custom translator for domain-specific terminology.
- Apply knowledge of Azure NLP services to select the appropriate service (AI Language, AI Speech, or AI Translator) for a given natural language processing requirement and use case.
- Analyze a multilingual business scenario to determine the optimal combination of Azure AI Language, Speech, and Translator services for an end-to-end natural language processing solution.
Describe conversational AI capabilities on Azure
- Describe conversational AI concepts including conversational flow design, intents, entities, utterances, and the role of language understanding in building intelligent chat and voice bots.
- Explain how Azure AI Bot Service integrates with Azure AI Language question answering capabilities for building knowledge-base-powered conversational experiences with multi-turn dialog.
- Apply knowledge of conversational AI to design a basic bot solution by selecting appropriate dialog patterns, knowledge sources, and Azure services for a given customer interaction scenario.
- Compare traditional rule-based bot approaches with AI-powered conversational solutions to assess which approach better meets accuracy, flexibility, and maintenance requirements for a given scenario.
5
Describe features of generative AI workloads on Azure
4 topics
Identify features of generative AI solutions
- Define generative AI and explain how it differs from traditional AI by producing new content including text, images, and code rather than classifying or predicting from existing data.
- Describe large language models including how they are trained on large text corpora, how they generate text using probability distributions, and the concepts of tokens and context windows.
- Explain the capabilities and limitations of generative AI including content creation, summarization, and translation alongside risks such as hallucinations, factual inaccuracy, and non-deterministic behavior.
- Assess the suitability of generative AI for a given business scenario by evaluating risk tolerance, output quality needs, content accuracy requirements, and human oversight considerations.
Identify capabilities of Azure OpenAI Service
- Describe Azure OpenAI Service capabilities including access to GPT, DALL-E, and embeddings models with enterprise security, compliance, and regional data residency guarantees.
- Explain Azure OpenAI Service deployment options including model deployment, Azure OpenAI Studio playground, and integration with Azure AI services for end-to-end generative AI solutions.
- Explain how Azure AI Search integrates with Azure OpenAI Service to implement retrieval-augmented generation that grounds model responses in enterprise knowledge bases and documents.
- Apply knowledge of Azure OpenAI Service features to select the appropriate model type (GPT for text, DALL-E for images, embeddings for search) for a given generative AI task.
- Analyze a business scenario to determine the optimal Azure OpenAI configuration by evaluating model selection, grounding strategy, content filtering, and integration architecture requirements.
Describe prompt engineering concepts
- Define prompt engineering and describe the role of system messages, user prompts, and assistant responses in structuring effective interactions with large language models.
- Explain prompt engineering techniques including zero-shot, few-shot, and chain-of-thought prompting and demonstrate how clear instructions and context improve response quality.
- Explain how model parameters including temperature, top-p, max tokens, and frequency penalty influence the creativity, determinism, and length of generated outputs.
- Apply prompt engineering techniques to construct effective prompts that produce accurate, relevant, and well-formatted responses for a given text generation task in Azure OpenAI Studio.
- Analyze prompt effectiveness by evaluating response quality issues, identifying prompt failure modes, and recommending iterative prompt design improvements to increase output accuracy and relevance.
Describe responsible generative AI practices
- Identify risks specific to generative AI including hallucinations, harmful content generation, copyright concerns, data privacy leakage, and potential for misuse in deception or manipulation.
- Explain content filtering and safety features in Azure OpenAI Service including configurable content filters for hate, violence, sexual content, and self-harm categories and how to configure severity levels.
- Apply responsible AI practices to generative AI solutions by implementing content moderation, grounding strategies, human oversight, and transparent disclosure of AI-generated content.
- Analyze a generative AI deployment scenario to evaluate responsible AI risks and recommend a layered mitigation strategy addressing content safety, grounding, monitoring, and user feedback mechanisms.
Hands-On Labs
15 labs
~200 min total
Console Simulator
Practice in a simulated cloud console or Python code sandbox — no account needed. Each lab runs entirely in your browser.
Certification Benefits
Salary Impact
$92,000
Average Salary
Related Job Roles
AI Business Analyst
Data Analyst
IT Professional
AI Solutions Consultant
Industry Recognition
Microsoft Azure certifications are among the most valued in enterprise IT, with Microsoft holding the second-largest cloud market share globally and serving as the dominant platform in enterprise and hybrid cloud environments.
Scope
Included Topics
- All domains in the Microsoft Azure AI Fundamentals (AI-900) exam: Domain 1 Describe AI workloads and considerations (15-20%), Domain 2 Describe fundamental principles of machine learning on Azure (20-25%), Domain 3 Describe features of computer vision workloads on Azure (15-20%), Domain 4 Describe features of Natural Language Processing workloads on Azure (15-20%), Domain 5 Describe features of generative AI workloads on Azure (15-20%).
- Common AI workloads including prediction, anomaly detection, natural language processing, conversational AI, computer vision, and generative AI.
- Responsible AI principles including fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability.
- Core machine learning concepts including features, labels, training data, validation data, supervised learning (regression, classification), unsupervised learning (clustering), and deep learning fundamentals.
- Azure AI services including Azure Machine Learning (automated ML, designer, data and compute assets), Azure AI Vision, Azure AI Custom Vision, Azure AI Face, Azure AI Language, Azure AI Speech, Azure AI Translator, Azure AI Document Intelligence, Azure AI Search, and Azure OpenAI Service.
- Computer vision tasks including image classification, object detection, optical character recognition, and facial detection and analysis.
- Natural language processing capabilities including key phrase extraction, named entity recognition, sentiment analysis, language modeling, speech recognition, speech synthesis, and translation.
- Generative AI concepts including natural language generation, image generation, code generation, large language models, Azure OpenAI Service models, prompt engineering fundamentals, and responsible generative AI with content filters.
Not Covered
- Advanced machine learning mathematics, gradient descent derivations, neural network architecture internals, and deep learning implementation details beyond foundational conceptual understanding.
- Azure infrastructure administration, networking, identity management, and non-AI Azure services not directly related to AI-900 exam objectives.
- Hands-on SDK code implementations, REST API call syntax, CLI commands, and infrastructure-as-code templates beyond conceptual service understanding.
- Transient service pricing details, rapidly changing benchmark values, and region-specific availability information that is not stable for a long-lived domain specification.
- Third-party AI/ML frameworks, non-Azure cloud AI services, and open-source model training workflows not covered in the AI-900 exam guide.
Official Exam Page
Learn more at Microsoft Azure
Ready to master AI-900?
Adaptive learning that maps your knowledge and closes your gaps.
Subscribe to Access