Microsoft · AI-102
Microsoft Certified: Azure AI Engineer Associate (AI-102) Practice Test
Validates expertise in designing and implementing Azure AI solutions using Azure AI services, Azure AI Search, and Azure OpenAI. Covers computer vision, natural language processing, knowledge mining, generative AI, and agentic solutions.
Exam Details
Questions
1392
Duration
100 minutes
Passing Score
700/1000
Difficulty
AssociateLast Updated
Jan 2026
Topics Covered
Azure AI ServicesComputer VisionNatural Language ProcessingKnowledge MiningGenerative AIAzure OpenAIAzure AI SearchResponsible AI
Microsoft Certified: Azure AI Engineer Associate (AI-102) Practice Exam Preparation
Use this AI-102 practice exam to prepare for Microsoft Certified: Azure AI Engineer Associate (AI-102) with realistic questions, detailed explanations, and focused study modes. The practice bank includes 1,392 questions for Microsoft AI-102, 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 Azure AI Services, Computer Vision, Natural Language Processing, Knowledge Mining, and Generative AI. 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.
Exam Domain Breakdown
Plan and manage an Azure AI solution23%
Implement generative AI solutions18%
Implement an agentic solution8%
Implement computer vision solutions13%
Implement natural language processing solutions18%
Implement knowledge mining and information extraction solutions18%
Exam Overview
The Microsoft Certified: Azure AI Engineer Associate certification, validated through exam AI-102 (Designing and Implementing a Microsoft Azure AI Solution), demonstrates expertise in building, managing, and deploying production-grade AI solutions on the Azure platform. The exam covers the full breadth of Azure AI capabilities including Azure AI Services, Azure AI Search, Azure OpenAI in Foundry Models, Azure AI Vision, Azure AI Language, Azure AI Speech, Azure AI Document Intelligence, and Azure Content Understanding. Candidates are assessed on their ability to architect secure, end-to-end AI solutions that span computer vision, natural language processing, knowledge mining, generative AI, and autonomous agentic workflows.
Last updated on December 23, 2025, the exam reflects Microsoft's shift toward the Microsoft Foundry platform and its emphasis on generative AI and agentic solutions. It aligns with the real-world responsibilities of an Azure AI engineer, including requirements definition, development, deployment, integration, performance tuning, and monitoring. The certification also requires demonstrating the ability to apply Responsible AI principles, configure content safety filters, and design governance frameworks—reflecting increasing enterprise demand for trustworthy AI systems.
Who Should Take This Exam
This certification is designed for software engineers and developers who build AI-integrated applications and want to formalize their expertise on the Azure platform. Ideal candidates work in roles such as AI Engineer, Machine Learning Engineer, Cloud Developer, Solutions Architect, or Data Engineer with a focus on AI. Candidates should have hands-on development experience using Python or C# and be comfortable consuming REST APIs and SDKs across the Azure AI portfolio.
The certification is most relevant to professionals who collaborate with data scientists, data engineers, IoT specialists, and infrastructure teams to integrate AI capabilities into broader enterprise solutions. It suits those with at least one to two years of experience building cloud-based applications on Azure and some familiarity with AI/ML concepts, though it does not require deep data science or model-training expertise.
Prerequisites
There are no mandatory prerequisites to sit for the exam, but Microsoft recommends foundational Azure knowledge equivalent to the AZ-900 (Azure Fundamentals) or AI-900 (Azure AI Fundamentals) certifications before attempting AI-102. Candidates should be proficient in Python or C# and comfortable with REST API consumption and SDK usage in at least one of those languages.
Practical experience is strongly recommended: candidates should have hands-on familiarity with Azure AI Services (including Vision, Language, Speech, and Document Intelligence), Azure AI Search, and Azure OpenAI. Understanding of cloud architecture concepts—such as resource provisioning, identity and access management, containerization, and CI/CD pipelines—will be essential for the planning and management domain. Exposure to prompt engineering, retrieval-augmented generation (RAG) patterns, and responsible AI principles is increasingly important given the exam's updated 2025 objectives.
Exam Format
Exam AI-102 is a proctored assessment delivered through Pearson VUE, available in both online proctored and in-person testing center formats. Candidates are given 100 minutes to complete the exam. The question count typically ranges from 40 to 60 items and may include multiple-choice, multiple-select, drag-and-drop, case studies, and interactive lab-style components. Microsoft does not publish a fixed question count, as it varies per exam form.
The passing score is 700 on a scale of 100–1000. Scoring is scaled and not a simple percentage of correct answers. The exam is available in English, Japanese, Chinese (Simplified and Traditional), Korean, German, French, Spanish, Portuguese (Brazil), and Italian. Candidates who test in a non-English language can request 30 additional minutes. If a candidate fails, they may retake the exam after 24 hours; subsequent retakes require a 14-day waiting period, and no more than five attempts are allowed within a 12-month period. The resulting certification is valid for one year and can be renewed at no cost via an online assessment on Microsoft Learn.
Skills Measured
- 1.Plan and manage an Azure AI solution (20–25%): Selecting appropriate Microsoft Foundry services for generative AI, vision, NLP, speech, and information extraction scenarios; provisioning and deploying Azure AI resources and models; integrating services into CI/CD pipelines; container deployment; monitoring costs and performance; managing authentication and account keys; implementing responsible AI governance including content filters, blocklists, prompt shields, and harm detection.
- 2.Implement generative AI solutions (15–20%): Building solutions with Microsoft Foundry including deploying hubs, projects, and models; implementing prompt flow and retrieval-augmented generation (RAG); evaluating models and flows; integrating with the Microsoft Foundry SDK; using Azure OpenAI in Foundry Models to generate text, code, and images (including DALL-E and large multimodal models); applying prompt engineering techniques; fine-tuning models; configuring model monitoring, tracing, and feedback; and orchestrating multiple generative AI models.
- 3.Implement an agentic solution (5–10%): Understanding agent roles and use cases; configuring resources and creating agents with the Microsoft Foundry Agent Service; implementing complex agents and multi-agent orchestration workflows using Microsoft Agent Framework; enabling autonomous capabilities; and testing, optimizing, and deploying agents.
- 4.Implement computer vision solutions (10–15%): Analyzing images with Azure Vision in Foundry Tools including object detection, tagging, and text extraction (OCR and handwriting); building and consuming custom image classification and object detection models; evaluating and publishing custom vision models; and analyzing videos with Azure AI Video Indexer and Spatial Analysis for detecting people presence and movement.
- 5.Implement natural language processing solutions (15–20%): Extracting key phrases, entities, sentiment, language, and PII from text; translating text and documents with Azure Translator in Foundry Tools; implementing text-to-speech and speech-to-text with Azure Speech in Foundry Tools including SSML, custom speech, intent recognition, and speech translation; and building custom language understanding models and question answering (QnA) solutions including multi-turn conversations and multi-language support.
- 6.Implement knowledge mining and information extraction solutions (15–20%): Provisioning Azure AI Search resources, defining indexes, skillsets, and indexers; implementing custom skills; querying with full-text, semantic, and vector search; managing Knowledge Store projections; using Azure Document Intelligence in Foundry Tools with prebuilt and custom models; training and publishing custom document intelligence models; and using Azure Content Understanding in Foundry Tools to extract text, entities, tables, images, and insights from documents, images, video, and audio.
Study Tips
- Work through the official Microsoft Learn learning path for AI-102, which is structured by exam domain and includes hands-on sandbox exercises. Access it via the 'Two ways to prepare' section on the official certification page.
- Use the free official Practice Assessment (Assessment ID 61) on Microsoft Learn before scheduling your exam. It mirrors the wording, style, and difficulty of real questions and identifies specific knowledge gaps by domain.
- Provision a free-tier or pay-as-you-go Azure subscription and build small end-to-end projects covering each domain: deploy an Azure OpenAI resource and build a RAG pipeline, create a custom vision model, build a QnA solution, and index documents with Azure AI Search. Hands-on lab time is essential.
- Study the official AI-102 Study Guide at aka.ms/ai102-StudyGuide carefully. The December 23, 2025 version reflects current objectives. Pay special attention to the 'Implement an agentic solution' domain (5–10%), which is new and covers Microsoft Foundry Agent Service and multi-agent orchestration with Microsoft Agent Framework.
- Explore the Azure AI Foundry portal (ai.azure.com) hands-on. The exam heavily tests Microsoft Foundry workflows including hub/project deployment, prompt flow, model evaluation, and tracing—capabilities that are best learned by actually using the portal rather than reading about them.
- Review responsible AI documentation including Azure AI Content Safety, prompt shields, content filters, and blocklists. The 'Plan and manage' domain (20–25%) includes a responsible AI governance section that is frequently tested.
- Use the Exam Sandbox at https://aka.ms/examdemo to familiarize yourself with interactive question types including drag-and-drop and case study formats before exam day, so you don't lose time getting oriented during the real exam.
Career Benefits
Earning the AI-102 certification positions professionals for high-demand roles including Azure AI Engineer, Machine Learning Engineer, Cloud AI Developer, AI Solutions Architect, and Cognitive Services Developer. According to ZipRecruiter, the average annual salary for Azure AI Engineers in the United States was approximately $158,110 as of early 2026, with senior roles at top technology companies exceeding $200,000. The certification is particularly valuable as enterprises scale generative AI adoption—engineers who can implement production-ready RAG pipelines, deploy Azure OpenAI solutions, and build agentic workflows are commanding significant salary premiums.
The AI-102 is the primary associate-level AI engineering credential in the Microsoft ecosystem, sitting above AI-900 (Azure AI Fundamentals) and complementing DP-100 (Azure Data Scientist Associate) for teams building end-to-end ML pipelines. Industry surveys indicate that 26% of certified professionals report job promotions and 35% report salary increases following certification. The exam's 2025 update to include agentic AI and Microsoft Foundry aligns the credential with current enterprise priorities, making it more relevant than its predecessors to organizations deploying Azure OpenAI at scale.
Sample Questions
5 sample questions with answers and explanations. Start a practice session to test yourself across all 1392 questions.
Preview — answers shown1. A financial institution, 'SecureBank,' plans to use the containerized Anomaly Detector API in its on-premises data center. To comply with strict security policies, they must ensure that API keys and billing endpoints are never stored in command-line histories, and access to pull the container image is tightly controlled via Azure Role-Based Access Control (RBAC). What is the correct sequence of actions to achieve this secure deployment?
A1. Pull image from Microsoft Registry. 2. Push image to Azure Container Registry (ACR). 3. Store secrets in Azure Key Vault. 4. Use 'docker run' with a configuration file mount.
B1. Create Azure Key Vault. 2. Push image to Azure Container Registry (ACR). 3. Use 'docker run' with secrets in the command line. 4. Pull image from Microsoft Registry.
C1. Store secrets in environment variables. 2. Pull image from Microsoft Registry. 3. Push image to Azure Container Registry (ACR). 4. Use 'docker run'.
D1. Use 'docker run' with a configuration file mount. 2. Store secrets in Azure Key Vault. 3. Pull image from Microsoft Registry. 4. Push image to Azure Container Registry (ACR).
Explanation
The correct and most secure sequence of actions is to pull the image, push it to a private registry, store secrets in Key Vault, and then run the container using a mounted configuration file. Why this is correct: 1. Pull image from Microsoft Registry: First, you get the official container image from Microsoft's public container registry (MCR). 2. Push image to Azure Container Registry (ACR): To control access, you push this image into your own private ACR. You can then apply Azure RBAC policies to the ACR to define which users or service principals are allowed to pull the image. 3. Store secrets in Azure Key Vault: The sensitive API key and billing endpoint should be stored securely in an Azure Key Vault, not in code or scripts. 4. Use 'docker run' with a configuration file mount: To run the container, you use the 'docker run' command with the '--mount' option to attach a configuration file (like an environment file) that contains the secrets retrieved from Key Vault. This prevents the sensitive values from ever appearing in the shell's command-line history. Why the others are incorrect: - Any option that involves putting secrets directly in the command line is insecure and violates the requirements. - Storing secrets in plain-text environment variables can also be a security risk. Using a Key Vault and a mounted file is the recommended best practice for this scenario. - The order is crucial; you cannot push an image to ACR before you have pulled it from the source registry.
2. Fabrikam must retain diagnostic logs from their Azure OpenAI resource for 5 years to satisfy a regulatory compliance requirement. Their security team also needs to run KQL queries against recent logs for daily troubleshooting. They want to minimize the number of diagnostic setting destinations while meeting both requirements. Which combination of destinations should they configure? (Select two!)
Multiple correct answersAAzure Storage account with lifecycle management policies for long-term archival
BAzure Event Hub to stream logs to an external SIEM platform
CLog Analytics workspace for interactive KQL querying of recent logs
DAzure Data Lake Storage Gen2 with hierarchical namespace enabled
EApplication Insights for application performance monitoring
Explanation
Configuring both Azure Storage and a Log Analytics workspace addresses both requirements with the fewest destinations. Azure Storage with lifecycle management policies provides cost-effective long-term archival well beyond 5 years, making it ideal for compliance retention. Log Analytics workspace enables full KQL querying capabilities against recent logs, supporting daily troubleshooting workflows. Log Analytics interactive retention is configurable up to 730 days (2 years), which is sufficient for active troubleshooting while Azure Storage handles the full 5-year compliance window. Event Hub is designed for streaming logs to external systems like SIEMs in real time and does not provide persistent storage or native query capabilities. Azure Data Lake Storage Gen2 can store logs but lacks built-in KQL query support and requires additional tooling. Application Insights is for application performance telemetry and is not a diagnostic settings destination for Azure AI service resource logs.
3. Litware needs to implement prompt shields for their Azure OpenAI chat application that allows users to upload reference documents. They want to protect against jailbreak attacks in user messages and attacks embedded in uploaded documents. Which prompt shields configuration should you enable? (Select one!)
AEnable user prompt attacks only, as document attacks are enabled by default
BEnable both user prompt attacks and document attacks explicitly
CEnable document attacks only, as user prompt attacks are always active
DUse the default configuration which includes both attack types
Explanation
Both user prompt attacks and document attacks must be explicitly enabled. Document attacks (also called indirect attacks) are OFF by default and must be specifically enabled to protect against attacks embedded in uploaded documents. User prompt attacks protect against jailbreak attempts in direct user messages and are ON by default. Since Litware accepts document uploads and needs protection against both attack vectors, both shields must be explicitly configured. The default configuration only protects against user prompt attacks, not document attacks.
4. ContractAnalysis Corp has a large collection of legal agreements in various formats and wants to optimize their custom NER model training approach. They have 500 contracts from different law firms, each with unique formatting styles. Which data selection strategy would best serve their model training?
ASelect the 50 most recent contracts to ensure current legal language
BChoose a diverse subset representing different firms, formats, and contract types while maintaining realistic distribution
CUse all 500 contracts to maximize training data volume
DFocus only on contracts from the largest law firm for consistency
Explanation
ContractAnalysis should choose a diverse subset that represents different law firms, formatting styles, and contract types while maintaining realistic distribution of their actual contract mix. This approach ensures their model can handle the variety they'll encounter in production while preventing overfitting to specific firm styles or overrepresenting less common contract types in training data.
5. Fabrikam is building a document processing solution using Azure Document Intelligence v4.0. They need to process invoices, receipts, and purchase orders that arrive in various formats. The team wants to create a composed model that can automatically route documents to the appropriate custom model for extraction. Which TWO requirements must be met to successfully create this composed model? (Select two!)
Multiple correct answersAAll component models must be custom models trained with labeled data
BAll component models must be trained using the same API version
CA separate document classifier model must be trained to handle routing between component models
DAll component models must be created in the same Azure Document Intelligence resource
EThe composed model can contain a maximum of 100 component models
Explanation
In Azure Document Intelligence v4.0, composed models require that all component models be custom models trained with labeled data. Additionally, v4.0 requires training an explicit document classifier model to handle routing between the component models, unlike earlier versions that used confidence-based routing. The API version does not need to be the same across component models (v3.0, v3.1, and v4.0 models can be mixed), and the maximum limit for component models in v4.0 is 500, not 100.
More Microsoft Practice Exams
Microsoft Dynamics 365 Supply Chain Management Functional Consultant Expert (MB-335)
MB-335 · 2039 questions
Microsoft Dynamics 365 Business Central Functional Consultant (MB-800)
MB-800 · 1899 questions
Microsoft Certified: Windows Server Hybrid Administrator Associate (AZ-801)
AZ-801 · 1376 questions
Microsoft Dynamics 365 Finance Functional Consultant (MB-310)
MB-310 · 1299 questions
Microsoft 365 Certified: Fundamentals (MS-900)
MS-900 · 1201 questions
Microsoft Certified: Power Platform Functional Consultant Associate (PL-200)
PL-200 · 1120 questions
$17.99
One-time access to this exam
Full access to all 1392 questions
Or $15/mo for all 253 exams
Detailed explanations
Free preview stays available