Microsoft · Practice Exam · Azure AI · Associate (retired)

Microsoft AI-102: Designing and Implementing an Azure AI Solution Practice Exam

AI-102 retired on June 30, 2026 and is succeeded by AI-103. This practice set still covers all six AI-102 skill areas — from planning Azure AI solutions to generative AI and agentic solutions — useful reference for the Azure AI service knowledge that carries directly into AI-103.

Start 24-hour free trial →

AI-102 has retired — the successor is AI-103

Microsoft retired AI-102 on June 30, 2026, and the Azure AI Engineer Associate certification can no longer be earned or renewed through it. If you already hold the certification, it remains on your Microsoft transcript until its expiration date.

The successor is AI-103: Developing AI Apps and Agents on Azure, leading to the new Microsoft Certified: Azure AI Apps and Agents Developer Associate. AI-103 shifts the focus decisively toward generative AI applications, autonomous agents, and Microsoft Foundry (RAG, agent orchestration, content safety). Your AI-102 service knowledge transfers, but the generative-AI and agent content is new ground. Confirm the current path on Microsoft Learn.

500+
Practice questions
6
Skill areas
2
Study modes
24h
Free trial

AI-102 exam at a glance

Vendor
Microsoft (Azure AI)
Exam code
AI-102 (retired June 30, 2026)
Certification
Microsoft Certified: Azure AI Engineer Associate
Successor
AI-103 (Developing AI Apps and Agents on Azure) → Azure AI Apps and Agents Developer Associate
Level
Associate (role-based; developer-focused)
Blueprint
Final edition: skills measured as of December 23, 2025
Format
Typically 40–60 questions; multiple choice, multiple response, drag-and-drop, and case studies
Duration
About 100–120 minutes
Passing score
700 of 1000 (scaled, not a raw percentage)
Recommended experience
Software development experience in Python or C#, using REST APIs and SDKs to build Azure AI solutions
Cost (USD)
Was $165 USD (associate tier) while active
Validity
One year; could be renewed free online while active — no longer renewable after retirement

Source: Microsoft Learn — AI-102 study guide (skills measured) and the Azure AI Engineer Associate certification page. AI-102 retired June 30, 2026; see AI-103 for the current path.

About the AI-102 Azure AI Engineer Associate certification

AI-102 validated that a developer could build, manage, and deploy AI solutions on Azure — selecting the right Azure AI service, applying responsible AI, and implementing computer vision, natural language processing, knowledge mining, generative AI, and (in its final edition) agentic solutions using REST APIs and SDKs in Python or C#. It earned the Microsoft Certified: Azure AI Engineer Associate certification with a single exam.

AI-102 retired on June 30, 2026 (see the notice above). Its successor, AI-103: Developing AI Apps and Agents on Azure, pivots toward generative AI applications, autonomous agents, and Microsoft Foundry. This practice set remains a useful reference for the Azure AI service fundamentals — vision, language, search, document intelligence, and Azure OpenAI — that carry directly into AI-103. If you are new to Azure AI, the AI-900 Azure AI Fundamentals exam is the conceptual starting point, and AZ-900 Azure Fundamentals covers the underlying cloud platform.

Every practice question maps to one of the six AI-102 skill areas and links to the specific Microsoft Learn page it was derived from, so your weak spots become a focused reading list rather than a guess.

AI-102 skill areas and weights

Six skill areas in the final edition, with planning and managing an Azure AI solution the heaviest at 20–25%. Microsoft published each area as a weighting range rather than an exact percentage; the ranges below are reproduced as Microsoft stated them. Note the newest area — agentic solutions — which foreshadows the AI-103 direction.

Plan and manage an Azure AI solution

Select the appropriate Azure AI (Microsoft Foundry) service for a business need; plan, create, deploy, secure, and monitor AI resources; manage cost, keys, and authentication; and implement responsible AI (content moderation, content safety, prompt shields, harm detection, governance).

20–25%Heaviest area
Implement generative AI solutions

Build generative AI solutions with Microsoft Foundry (hubs and projects, prompt flow, RAG grounding, evaluation, SDK integration, prompt templates); use Azure OpenAI models to generate content; and optimize and operationalize (parameters, monitoring, fine-tuning, orchestration, prompt engineering).

15–20%
Implement natural language processing solutions

Analyze and translate text (key phrases, entities, sentiment, language detection, PII, translation); process and translate speech (text-to-speech, speech-to-text, SSML, custom speech); and implement custom language models and question answering.

15–20%
Implement knowledge mining and information extraction solutions

Implement Azure AI Search (indexes, indexers, skillsets, custom skills, querying, semantic and vector search); Azure Document Intelligence (prebuilt and custom models, composed models); and Azure Content Understanding (OCR pipelines, classification, entity and table extraction).

15–20%
Implement computer vision solutions

Analyze images (visual features, object detection, tagging, OCR, handwriting); implement custom vision models (classification versus object detection, labeling, training, evaluation, publishing); and analyze videos with Azure AI Video Indexer and spatial analysis.

10–15%
Implement an agentic solution

The newest area, foreshadowing AI-103: understand the role and use cases of an agent; configure resources and create an agent with Microsoft Foundry Agent Service; implement complex agents with the Microsoft Agent Framework; and orchestrate multi-agent workflows, then test, optimize, and deploy.

5–10%Newest area

Source: Microsoft Learn — AI-102 study guide (final edition, skills measured as of December 23, 2025). Microsoft published weightings as ranges; planning and managing an Azure AI solution was the heaviest area.

Who AI-102 was for

The Azure AI Engineer Associate certification targeted developers building AI into applications on Azure:

  • AI and software engineers building vision, language, search, and generative AI features using Azure AI services, REST APIs, and SDKs.
  • Application developers proficient in Python or C# integrating AI capabilities into production solutions.
  • Cloud developers working with solution architects and data professionals to deliver end-to-end AI solutions.
  • Practitioners moving toward agentic AI — the area AI-103 now expands into.

AI-102 assumed real development experience — it was not an entry point. If you are new to Azure AI concepts, the AI-900 Azure AI Fundamentals exam is the conceptual starting point, and AZ-900 Azure Fundamentals covers the cloud platform. Because AI-102 has retired, plan toward AI-103 for the current credential. For the roles this skill set supports — with skills, tools, and salary ranges — see the AI Engineer career path.

What this AI-102 practice exam delivers

Learn mode

Get the correct answer, the reasoning, and a direct link to the exact Microsoft Learn page each question was derived from — immediately after each question. Built for the service-selection and implementation scenarios across Azure AI vision, language, search, and generative AI.

Exam mode

A timed run in the AI-102 shape — roughly 40 to 60 questions in about 100 to 120 minutes, including case studies — useful pacing practice for AI-103, which keeps a similar format.

Source-linked explanations

Every answer cites the exact Microsoft Learn documentation page it was built from — so you can verify Azure AI service behavior, much of which carries directly into AI-103.

Score by skill area

Results break down by the six AI-102 areas — planning, generative AI, NLP, knowledge mining, computer vision, and agentic solutions — so practice tells you exactly which area to revisit.

Sample AI-102 practice questions

Ten free questions across the six AI-102 skill areas, with full explanations and source links to the Microsoft Learn pages each is derived from. The complete bank is available with the 24-hour trial.

Question 1 · Plan and manage (20–25%)

A team needs a single multi-service Azure AI resource so several capabilities — vision, language, and speech — can be accessed through one endpoint and key, with unified billing.

Which approach should you recommend?

  1. Provision a multi-service Azure AI services resource
  2. Provision a separate single-service resource for each capability only
  3. Use a storage account with static files
  4. Deploy an Azure Kubernetes Service cluster
Show answer & explanation

Correct: A — provision a multi-service Azure AI services resource. A multi-service resource exposes multiple Azure AI capabilities through one endpoint and key with unified billing — exactly the requirement. Selecting the right resource shape is a core plan-and-manage skill.

Why not the others: separate single-service resources (B) fragment endpoints, keys, and billing — the opposite of the ask; a storage account (C) holds data, not AI capabilities; an AKS cluster (D) is container orchestration, not an AI service resource.

Source: Microsoft Learn — What are Azure AI services? → Further reading: PowerKram — AI-900 Azure AI Fundamentals →
Question 2 · Plan and manage / responsible AI (20–25%)

A generative AI chat application must block user prompts that attempt to jailbreak the model or inject malicious instructions, before they reach the model.

Which Azure AI Content Safety capability should you implement?

  1. Azure AI Search semantic ranking
  2. Custom Vision object detection
  3. Prompt Shields
  4. Speech synthesis (SSML)
Show answer & explanation

Correct: C — Prompt Shields. Prompt Shields in Azure AI Content Safety detect and block jailbreak and prompt-injection attacks in user input before it reaches the model — exactly the responsible-AI control described.

Why not the others: semantic ranking (A) improves search relevance; Custom Vision object detection (B) is an image task; SSML (D) shapes synthesized speech. None guards against prompt injection.

Source: Microsoft Learn — Azure AI Content Safety →
Question 3 · Generative AI (15–20%)

A generative AI assistant must answer questions using a company’s own private documents so responses are grounded in that content rather than only the model’s training data.

Which pattern should you implement?

  1. Retrieval-augmented generation (RAG)
  2. Image classification
  3. Speech-to-text transcription
  4. Anomaly detection
Show answer & explanation

Correct: A — retrieval-augmented generation (RAG). RAG grounds a model in your own data by retrieving relevant content (often via Azure AI Search) and supplying it as context, so answers reflect private documents — exactly the requirement, and a central AI-102/AI-103 pattern.

Why not the others: image classification (B) labels images; speech-to-text (C) transcribes audio; anomaly detection (D) flags outliers. None grounds a language model in private content.

Source: Microsoft Learn — Retrieval-augmented generation in Microsoft Foundry → Further reading: PowerKram — AI-900 Azure AI Fundamentals →
Question 4 · Generative AI (15–20%)

You are building generative AI applications and want a unified platform to deploy hubs and projects, evaluate models and prompt flows, and integrate them into an app via an SDK.

Which Azure platform should you use?

  1. Azure Blob Storage
  2. Microsoft Foundry (Azure AI Foundry)
  3. Azure Virtual Machines
  4. Azure DNS
Show answer & explanation

Correct: B — Microsoft Foundry (Azure AI Foundry). Foundry is the unified platform for building generative AI solutions — hubs and projects, model and prompt-flow evaluation, and SDK integration — exactly the platform described. It is central to the final AI-102 blueprint and to AI-103.

Why not the others: Blob Storage (A) stores data; Virtual Machines (C) are raw compute; Azure DNS (D) resolves names. None is the generative AI development platform.

Source: Microsoft Learn — What is Microsoft Foundry? →
Question 5 · Natural language processing (15–20%)

An application must analyze customer feedback to determine whether each comment is positive, negative, or neutral, with confidence scores.

Which Azure AI Language feature should you use?

  1. Key phrase extraction
  2. Language detection
  3. Sentiment analysis
  4. Optical character recognition
Show answer & explanation

Correct: C — sentiment analysis. Sentiment analysis in Azure AI Language classifies text as positive, negative, or neutral with confidence scores — exactly the requirement.

Why not the others: key phrase extraction (A) pulls main topics; language detection (B) identifies the language; OCR (D) extracts text from images, not sentiment.

Source: Microsoft Learn — Azure AI Language →
Question 6 · Natural language processing (15–20%)

A support application must extract people, organizations, and locations mentioned in incoming text so they can be tagged and routed automatically.

Which Azure AI Language capability should you use?

  1. Text-to-speech
  2. Named entity recognition (NER)
  3. Custom Vision
  4. Semantic ranking
Show answer & explanation

Correct: B — named entity recognition (NER). NER identifies and categorizes entities such as people, organizations, and locations in text — exactly the extraction described.

Why not the others: text-to-speech (A) synthesizes audio; Custom Vision (C) is an image service; semantic ranking (D) reorders search results. None extracts named entities from text.

Source: Microsoft Learn — Azure AI Language (NER) → Further reading: PowerKram — AZ-104 Azure Administrator →
Question 7 · Knowledge mining (15–20%)

A company wants to make thousands of internal documents searchable with an index, enrichment pipeline, and support for semantic and vector queries.

Which Azure service should you use?

  1. Azure Blob Storage alone
  2. Azure Speech
  3. Azure AI Search
  4. Azure Bastion
Show answer & explanation

Correct: C — Azure AI Search. Azure AI Search builds indexes with enrichment skillsets and supports semantic and vector queries over large document sets — exactly the knowledge-mining requirement, and the retrieval engine behind most RAG solutions.

Why not the others: Blob Storage alone (A) stores files without search; Azure Speech (B) handles audio; Azure Bastion (D) provides secure VM access. None is a search-and-enrichment service.

Source: Microsoft Learn — Azure AI Search → Further reading: PowerKram — AI-900 Azure AI Fundamentals →
Question 8 · Knowledge mining / document intelligence (15–20%)

An accounts-payable process must automatically extract fields — vendor, total, line items — from scanned invoices in varied layouts.

Which Azure service is designed for this?

  1. Azure AI Translator
  2. Azure AI Document Intelligence
  3. Azure Traffic Manager
  4. Azure Key Vault
Show answer & explanation

Correct: B — Azure AI Document Intelligence. Document Intelligence extracts structured fields and tables from documents such as invoices, using prebuilt or custom models across varied layouts — exactly the requirement.

Why not the others: Translator (A) converts languages; Traffic Manager (C) is DNS-based routing; Key Vault (D) stores secrets. None extracts fields from documents.

Source: Microsoft Learn — Azure AI Document Intelligence →
Question 9 · Computer vision (10–15%)

An app must read printed and handwritten text from photographed documents and return the text with its location on the page.

Which Azure AI Vision capability should you use?

  1. Sentiment analysis
  2. Azure AI Search indexing
  3. Prompt flow
  4. Optical character recognition (OCR / Read)
Show answer & explanation

Correct: D — optical character recognition (OCR / Read). Azure AI Vision OCR (the Read capability) extracts printed and handwritten text from images and returns the text with bounding-box locations — exactly the requirement.

Why not the others: sentiment analysis (A) scores text tone; Azure AI Search indexing (B) makes content searchable; prompt flow (C) orchestrates generative AI. None reads text from images.

Source: Microsoft Learn — Azure AI Vision →
Question 10 · Agentic solution (5–10%)

You need to build an AI agent on Azure that can reason over a goal, call tools and functions, and complete multi-step tasks with minimal human intervention.

Which Azure capability should you use?

  1. Azure AI Search indexers
  2. Custom Vision training
  3. Speech Synthesis Markup Language
  4. Microsoft Foundry Agent Service
Show answer & explanation

Correct: D — Microsoft Foundry Agent Service. Foundry Agent Service is the platform for building AI agents that reason over goals, call tools and functions, and complete multi-step tasks — exactly the agentic scenario. This newest AI-102 area is the direct bridge to AI-103’s agent focus.

Why not the others: AI Search indexers (A) populate a search index; Custom Vision training (B) builds image models; SSML (C) shapes synthesized speech. None builds an autonomous agent.

Source: Microsoft Learn — Microsoft Foundry Agent Service → Further reading: PowerKram — AZ-104 Azure Administrator →

Keep going: study & career resources

AI-102 has retired, but the Azure AI service knowledge behind it carries directly into AI-103. Below it, the fundamentals exam builds the conceptual base — and the AI-engineer career track shows where these skills lead.

Deep dive: AI-102 retirement, the AI-103 successor, the six areas, and study value

Retirement and the AI-103 successor

AI-102 retired on June 30, 2026, and the Azure AI Engineer Associate certification can no longer be earned or renewed through it. Its successor is AI-103: Developing AI Apps and Agents on Azure, leading to the Microsoft Certified: Azure AI Apps and Agents Developer Associate. AI-103 was in beta from April 2026 and reached general availability around the retirement date. If you already hold the AI-102 certification, it stays on your transcript until it expires, but you cannot renew it. See the AI-900 Fundamentals path →

What changed: from services to agents

AI-102 centered on wiring up pre-built Azure AI services — vision, language, search, document intelligence, and Azure OpenAI — into applications. AI-103 pivots to building generative AI applications and multi-agent solutions with Microsoft Foundry: RAG patterns, agent orchestration, the Semantic Kernel and Agent frameworks, content safety, and model monitoring. Microsoft signaled this shift within AI-102 itself, adding an agentic-solution area to the final blueprint. Career Hub — AI Engineer →

The six skill areas

The final AI-102 blueprint covered planning and managing an Azure AI solution (heaviest, including responsible AI), implementing generative AI, NLP, and knowledge-mining solutions (each a middle-weight band), computer vision, and the newest area, agentic solutions. Much of this — service selection, responsible AI, RAG, Azure AI Search, and Foundry — maps directly onto AI-103, which is why AI-102 practice remains useful reference material. See the AZ-104 Administrator path →

Why this practice set is still worth using

The exam is retired, but the underlying Azure AI service knowledge is not. If you are preparing for AI-103, working through AI-102-style scenarios on Azure AI services, responsible AI, RAG, and Azure AI Search builds the foundation the new exam assumes — you then layer the newer agent and Foundry-app content on top. Treat this set as a service-fundamentals warm-up, not a substitute for AI-103-specific study. See the AZ-900 Azure Fundamentals path →

Format and scoring (as it was)

AI-102 delivered roughly 40 to 60 questions in about 100 to 120 minutes, scored 700 out of 1000 to pass, with multiple choice, multiple response, drag-and-drop, and case studies. AI-103 keeps a similar shape ($165, 700/1000, about 120 minutes), so the pacing practice transfers even though the content emphasis shifts. Azure AI Engineer Associate certification page →

Career outlook

Azure AI engineering skills remain in strong demand — the credential simply moved to AI-103. Roles include AI engineer, generative-AI application developer, and machine-learning engineer. Employers increasingly expect familiarity with generative AI, RAG, and agentic patterns, which is exactly where the certification path has moved. Career Hub — AI Engineer →

Microsoft AI-102 (Azure AI Engineer Associate) exam FAQ

Is AI-102 retired, and what replaces it?
Yes. AI-102 retired on June 30, 2026, and the Azure AI Engineer Associate certification can no longer be earned or renewed through it. The successor is AI-103: Developing AI Apps and Agents on Azure, leading to the Microsoft Certified: Azure AI Apps and Agents Developer Associate. AI-103 focuses on generative AI applications, autonomous agents, and Microsoft Foundry. If you already hold the AI-102 certification, it stays on your transcript until it expires.
What were the AI-102 skill areas and weights?
Per Microsoft’s final AI-102 study guide (skills measured as of December 23, 2025): Plan and manage an Azure AI solution (20–25%, heaviest), Implement generative AI solutions (15–20%), Implement natural language processing solutions (15–20%), Implement knowledge mining and information extraction solutions (15–20%), Implement computer vision solutions (10–15%), and Implement an agentic solution (5–10%). Microsoft published these as ranges.
Should I still study AI-102 material?
For the certification itself, no — take AI-103 instead. But the Azure AI service knowledge AI-102 covered (service selection, responsible AI, RAG, Azure AI Search, vision, language, document intelligence, and Foundry) transfers directly into AI-103, so AI-102-style practice is useful foundational reference before you layer on the newer agent and generative-app content.
What was the AI-102 passing score and format?
A passing score of 700 out of 1000 (scaled), with roughly 40 to 60 questions in about 100 to 120 minutes, including one or more case studies. Formats included multiple choice, multiple response, and drag-and-drop. AI-103 keeps a similar format.
What experience did AI-102 assume?
AI-102 was a developer exam. Microsoft expected software development experience in Python or C#, using REST APIs and SDKs to build Azure AI solutions. It was not an entry point; candidates newer to Azure AI concepts should start with AI-900 Azure AI Fundamentals.

Start your free 24-hour AI-102 practice trial

Full access to the question bank, both study modes, source-linked explanations, and score-by-skill-area — a solid Azure AI service foundation as you move toward AI-103. No credit card required.

Start free trial →