Microsoft retired AI-102 and its associated Azure AI Engineer Associate certification on June 30, 2026. Azure AI engineering skills remain relevant, but an AI-102 study plan is now historical and should not be presented as a bookable certification route. Learners comparing AI-103 or other current assessments should review the new role scope directly; the codes are not interchangeable.
The breadth of knowledge covered by AI-102 spans several distinct technical domains including natural language processing, computer vision, knowledge mining, conversational AI, and responsible AI principles. Each domain represents a cluster of Azure services and implementation patterns that the former exam assessed through scenario-based questions requiring practical judgment rather than simple fact recall. Candidates who approach this exam expecting a straightforward memorization exercise consistently underperform, while those who build genuine hands-on experience with the services covered tend to find the questions far more intuitive and manageable. The exam reflects real-world implementation challenges, and preparation that mirrors real-world practice is the most effective foundation for success.
Who Should Attempt AI-102
AI-102 is targeted at professionals who occupy or aspire to occupy the role of Azure AI Engineer, a position that bridges the gap between data science and software engineering with a specific focus on deploying pre-built and custom AI capabilities into production applications. Ideal candidates include software developers who have begun integrating Azure Cognitive Services into applications, solutions architects who design AI-enabled systems on the Azure platform, and data professionals who want to formalize their Azure AI implementation skills with a recognized industry credential. The certification is not suitable as a first Azure credential for someone entirely new to the platform.
A practical prerequisite for AI-102 success is comfort with at least one programming language supported by the Azure AI service SDKs, with Python and C# being the most commonly used in exam-aligned scenarios. Candidates should also have a working familiarity with REST API concepts, JSON data structures, and basic cloud computing concepts on Azure including resource groups, subscriptions, and service deployment patterns. Those who hold the AZ-900 Azure Fundamentals or AI-900 Azure AI Fundamentals certifications have a useful conceptual foundation but should not assume those credentials alone constitute sufficient technical preparation for the more demanding requirements of AI-102.
The AI-102 exam is organized around a set of functional skill areas that Microsoft publishes in a skills measured document updated periodically to reflect changes in Azure services and exam priorities. The major skill areas include planning and managing Azure AI solutions, implementing decision support solutions, implementing computer vision solutions, implementing natural language processing solutions, implementing knowledge mining and document intelligence solutions, and implementing generative AI solutions. Each area carries a specified percentage weight in the overall exam score, and candidates who are unaware of these weightings may allocate preparation time inefficiently.
The skill area weights reveal that no single domain dominates the exam to the exclusion of others, which means broad preparation across all areas is genuinely necessary. However, the combined weight of natural language processing and the planning and management of AI solutions typically represents a significant portion of the exam, making those areas particularly important to cover thoroughly. Candidates should download and carefully read the official skills measured document from Microsoft’s certification page at the beginning of their preparation because it serves as the most authoritative guide to exactly what will and will not be tested. Treating this document as the primary roadmap, rather than any third-party outline, ensures preparation is aligned with the actual exam rather than a potentially outdated approximation of it.
Azure Cognitive Services Overview
Azure Cognitive Services form the technical backbone of a large portion of the AI-102 exam, and developing a thorough working knowledge of these services is non-negotiable for candidates who want to perform well. These services provide pre-built AI capabilities through REST APIs and client SDKs that developers can integrate into applications without needing to build or train machine learning models from scratch. The major categories include vision services for image and video analysis, speech services for speech-to-text and text-to-speech conversion, language services for text analysis and translation, and decision services for content moderation and personalization.
Each individual service within these categories has its own specific capabilities, configuration options, deployment patterns, and SDK interaction methods that the exam may test. Azure Computer Vision, for instance, can analyze images to extract descriptions, detect objects, read text through optical character recognition, and identify faces, while Azure Custom Vision allows organizations to train custom image classification and object detection models using their own labeled data. Knowing not just that these services exist but precisely what each one can and cannot do, how they are configured and authenticated, and when one should be chosen over another is the level of detail the exam demands. Surface-level awareness of service names is insufficient preparation for the scenario-based questions that distinguish this exam.
Natural language processing represents one of the most heavily tested areas in AI-102 and encompasses a rich set of Azure services that enable applications to analyze, interpret, and generate human language. Azure AI Language, formerly known as Text Analytics, provides capabilities including sentiment analysis, key phrase extraction, named entity recognition, language detection, and personally identifiable information extraction. These capabilities are accessible through a unified API endpoint and are frequently combined in real-world applications that need to process large volumes of text data to extract actionable insights.
Beyond basic text analysis, the NLP domain in AI-102 covers Azure AI Language’s custom capabilities including custom named entity recognition and custom text classification, which allow organizations to train models on their own labeled datasets to recognize domain-specific entities or classify documents according to organization-specific categories. The Conversational Language Understanding service enables developers to build natural language understanding models that can identify intents and extract entities from user utterances, forming the cognitive core of conversational AI applications. Candidates must understand not only how to interact with these services through code but also how to design solutions that select the appropriate NLP capability for a given business requirement, configure the service correctly, and handle the outputs appropriately within a larger application architecture.
Computer Vision Implementation Skills
The computer vision domain in AI-102 covers a set of Azure services that enable applications to extract information from images and videos in ways that replicate and often exceed human visual perception for specific analytical tasks. Azure AI Vision provides image analysis capabilities including caption generation, object detection, tag generation, background removal, and spatial analysis of physical spaces using video feeds. The Document Intelligence service, formerly known as Form Recognizer, enables the extraction of structured data from documents including invoices, receipts, identity documents, and custom document types using pre-built and custom models.
Candidates must demonstrate the ability to implement solutions using these services, including provisioning the required Azure resources, configuring authentication using API keys or Azure Active Directory credentials, calling the appropriate API endpoints or SDK methods, and processing the response data to extract the needed information. The exam also tests knowledge of when to use pre-built models versus custom-trained models, how to train and evaluate custom models using labeled training data, and how to monitor model performance in production. Practical experience with the Azure portal, Azure AI Studio, and the relevant SDKs in Python or C# is the most effective way to build the implementation confidence that exam questions in this domain require.
Conversational AI is a distinct technical domain within AI-102 that covers the design and implementation of intelligent agents capable of conducting natural language conversations with users across multiple channels. Azure Bot Service provides the infrastructure for deploying conversational AI applications, while the Azure AI Language service’s question answering capability enables bots to respond to user questions by drawing on a knowledge base derived from documents, FAQ pages, or manually authored question-answer pairs. Together, these services form the foundation of most enterprise conversational AI implementations on the Azure platform.
the former exam assessed candidates’ ability to design multi-turn conversational flows, integrate language understanding models to handle complex user intents, connect bots to multiple communication channels including Microsoft Teams, web chat, and telephony interfaces, and implement authentication and security controls appropriate for conversational AI applications. Candidates should be familiar with the Bot Framework Composer as a visual tool for designing bot logic and with the Bot Framework SDK for programmatic bot development. The conversational AI domain rewards candidates who have actually built and deployed at least a basic bot implementation because the questions frequently involve recognizing the correct sequence of steps for accomplishing specific implementation tasks that are difficult to internalize from documentation alone.
Knowledge Mining With Azure Search
Azure AI Search, previously known as Azure Cognitive Search, is a cloud search service that incorporates AI capabilities to extract and index content from a wide variety of data sources, making that content discoverable through sophisticated query interfaces. Knowledge mining represents the application of AI to large document repositories to surface insights, connections, and structured information that would be impractical to extract manually. the retired AI-102 exam tested candidates’ ability to design and implement knowledge mining solutions that combine Azure AI Search with cognitive skills to enrich indexed content with AI-generated metadata.
An enrichment pipeline in Azure AI Search uses a skillset, which is a collection of cognitive skills that process documents during indexing to add derived information such as key phrases, entity mentions, sentiment scores, translated text, image descriptions, and custom model outputs to the search index. Candidates must understand how to define index schemas that capture both the original content and the enriched metadata, how to configure indexers that connect to data sources including Azure Blob Storage, Azure SQL Database, and Azure Cosmos DB, and how to design skillsets that apply the appropriate combination of built-in and custom skills to achieve the desired enrichment outcome. This domain rewards candidates who have worked through end-to-end knowledge mining implementation scenarios rather than those who have only read about the components in isolation.
Microsoft has embedded responsible AI principles throughout the Azure AI platform, and AI-102 explicitly tests candidates’ knowledge of these principles and their application to AI solution design and implementation. The six responsible AI principles that Microsoft promotes are fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability. Candidates are expected to understand not just the definitions of these principles but how they translate into specific design decisions, evaluation practices, and governance mechanisms in real AI implementations.
Concrete responsible AI capabilities in Azure include Content Safety, which provides tools for detecting and filtering harmful content across text and image modalities, and the Transparency Notes published by Microsoft for each cognitive service that explain how the underlying models work, what their intended uses are, and what limitations and potential harms they carry. The exam may present scenarios in which a candidate must identify which responsible AI principle is being addressed by a specific design choice, recommend appropriate safeguards for a given deployment context, or select the Azure tool most suited to addressing a particular responsible AI concern. This domain is increasingly important as organizations face regulatory and reputational pressure to deploy AI systems that can be audited, explained, and corrected when they produce harmful outcomes.
Generative AI On Azure Platform
The inclusion of generative AI in the AI-102 exam reflects the rapid growth of large language model capabilities and their integration into Azure AI services. Azure OpenAI Service provides access to powerful generative AI models including GPT-4 and other models from OpenAI, integrated within the Azure security and compliance framework. Candidates must understand how to provision Azure OpenAI resources, deploy specific model versions, interact with the completions and chat completions APIs, and implement responsible use patterns appropriate for generative AI applications including content filtering and usage monitoring.
The former exam also covered prompt engineering principles that guide the design of effective inputs to generative AI models, including the use of system messages to establish model behavior, few-shot examples to guide response format, and retrieval augmented generation patterns that combine generative models with external knowledge sources to produce more accurate and grounded responses. Azure AI Studio provides an integrated development environment for experimenting with and deploying generative AI solutions, and familiarity with its interface and workflow is beneficial for candidates tackling questions in this domain. The generative AI domain is among the most recently added to the exam and reflects the direction in which Azure AI engineering is rapidly moving, making it a high-priority area for any candidate preparing today.
Hands-On Lab Practice Importance
No amount of reading documentation or watching instructional videos can fully substitute for the learning that comes from actually provisioning Azure AI services, writing code that calls their APIs, and troubleshooting the errors and unexpected behaviors that arise during real implementation work. Hands-on lab practice is the single most effective preparation activity for AI-102 because the exam’s scenario-based questions are designed to test the kind of practical judgment that only develops through direct experience with the services. Candidates who have personally built working implementations of the key services covered by the exam answer these questions with a confidence and accuracy that purely theoretical preparation cannot produce.
Microsoft Learn, the official learning platform for Microsoft certifications, provides a structured set of learning paths and modules specifically aligned with AI-102 that include interactive sandbox environments where candidates can complete hands-on exercises without needing their own Azure subscription. These sandboxes provide temporary access to real Azure resources within a constrained environment, allowing candidates to follow guided implementation exercises across all the major service areas covered by the exam. Supplementing the Microsoft Learn exercises with independent practice projects, where the candidate designs and implements an AI solution without step-by-step guidance, significantly deepens the learning and builds the problem-solving confidence that translates most directly into exam performance.
The landscape of available study resources for AI-102 preparation includes both official Microsoft materials and high-quality third-party offerings that candidates can combine according to their learning preferences and preparation timeline. The official Microsoft Learn learning paths for AI-102 represent the most authoritative and up-to-date content because they are maintained by Microsoft and updated when services change or exam objectives are revised. These learning paths are free, comprehensive, and directly aligned with the skills measured document, making them the foundation that every preparation strategy should include regardless of what additional resources are used.
Beyond Microsoft Learn, several well-regarded third-party platforms offer AI-102 preparation courses that provide alternative explanations, additional practice questions, and instructor-guided walkthroughs of complex topics. Platforms like Pluralsight, Udemy, and A Cloud Guru have courses specifically designed for AI-102 that many candidates find valuable for their explanatory clarity and structured curriculum. Official Microsoft documentation for each Azure AI service, accessible through the Azure documentation portal, serves as the definitive technical reference for the specific capabilities, configuration parameters, and API behaviors that the exam may test. The combination of structured learning paths, hands-on lab work, third-party courses for alternative perspectives, and official documentation for deep dives on specific services creates a preparation ecosystem that covers all the dimensions the exam measures.
The skills that survive the AI-102 retirement
AI-102 retired in June 2026, but it still documents a useful earlier phase of Azure AI engineering. Skills such as choosing retrieval strategies, evaluating extracted text, protecting personally identifiable information, testing conversational responses, and monitoring latency remain essential. The retirement means the old exam blueprint is no longer a sufficient study plan for an active certification.
Modern Azure AI projects need an evaluation plan before they need a complicated prompt. Define measurable task success, retrieval relevance, human-review thresholds, cost per request, safety failure scenarios, and logging boundaries. Compare currently offered Microsoft learning paths and assessments by role and the actual services used; do not assume that a nearby exam code represents a direct replacement for AI-102.
Practice Exams And Mock Tests
The retired AI-102 question bank may still illuminate some older Azure AI scenarios, but a practice score against that blueprint cannot predict readiness for an active Microsoft certification. For professional development, turn each scenario into a testable engineering decision: decide how to evaluate model outputs, secure service credentials, detect retrieval failures, or control latency and cost. Compare the design with current Microsoft documentation and record what has changed since older examples were written.
Use practical reviews as diagnostic exercises rather than proof of certification progress. If a question exposes a weak understanding of Azure AI Search, grounding, or model evaluation, create a small nonproduction test case, inspect logs and outputs, and document the result. A mistaken assumption that is discovered and corrected in a lab can be more valuable than a high score on an outdated practice assessment. When seeking an active exam, use its current official skills outline and the practice materials intended for that exam.
Managing Exam Day Preparation
A modern project-ready learning plan should end with consolidation of documented results, not with a countdown to an exam that no longer exists. Revisit weak service areas and verify whether current interfaces, SDKs, quotas, and security controls match the older AI-102 course. Turn the lessons into a portfolio note describing the architecture, evaluation metrics, security boundaries, and unresolved risks. This makes progress understandable to an engineering reviewer.
Demonstrate hands-on competence with a small end-to-end AI application: ingest source content, establish a retrieval strategy, generate responses with grounded citations, test unacceptable responses, measure latency and cost, and protect secrets. Record failures in each stage, not merely successful screenshots. The resulting evidence can guide conversations about current Microsoft AI roles and credentials without implying that AI-102 remains bookable.
Post-Certification Career Opportunities
Earning the AI-102 certification opens tangible career opportunities in a technology sector where demand for qualified Azure AI practitioners consistently exceeds the available supply of certified professionals. Organizations across industries including healthcare, finance, retail, manufacturing, and government are actively investing in Azure AI implementations to automate processes, extract insights from unstructured data, improve customer experiences through conversational AI, and build intelligent applications that leverage the capabilities of cloud-based AI services. Certified AI engineers who can bridge the gap between business requirements and Azure AI implementation are valuable contributors to these initiatives.
The certification also serves as a credible signal of technical capability in competitive job markets where employers use credentials to filter large applicant pools. Roles such as Azure AI Engineer, AI Solutions Architect, Cognitive Services Developer, and AI Platform Specialist frequently list AI-102 as a preferred or required qualification. The certification can also accelerate advancement within existing roles by demonstrating a level of Azure AI expertise that justifies expanded responsibilities and compensation. Many professionals find that the process of earning the certification opens conversations with colleagues and managers about AI implementation projects that were not previously accessible to them, creating practical opportunities to apply and deepen the skills developed during preparation.
The Azure AI Engineer Associate certification associated with AI-102 was retired in June 2026. That historical credential may appear in older job listings and learning material, but it is no longer a current certification target. Its useful legacy is the emphasis on services, implementation trade-offs, and responsible AI practices; contemporary candidates should evaluate Microsoft’s available assessment choices against their role and actual technology stack.
A practical roadmap starts by defining the project problem and the criteria for a successful solution. It continues with secure service provisioning, retrieval or document processing where needed, reliable API integration, and realistic evaluation. The final stage is operational readiness: observe quality drift, log failures without exposing sensitive data, set service budgets, and rehearse a rollback path. Each stage gives a reviewer something concrete to inspect instead of a study-hours tally.
Azure AI projects span infrastructure, security, data engineering, model integration, and application delivery. Professionals need enough cross-functional understanding to recognize when a data quality defect looks like a model defect, when a permission error prevents retrieval, or when prompt changes hide a monitoring gap. AI-102 introduced many practitioners to that landscape, but current capabilities and controls should be the point of reference for production decisions.
The strongest outcome from studying the retired AI-102 material is not a new badge; it is the ability to justify technical choices and demonstrate safe, measurable AI behavior. Use supported services, consult current first-party instructions, and choose any future certification because its role profile matches the work to be done. Retain historical learning where it still teaches a mechanism, and discard exam-day advice that ceased to apply when AI-102 retired.
