Azure AI-103 Certification: Complete Study Guide 2026
Everything you need to pass AI-103: Developing AI Apps and Agents on Azure — the successor to AI-102. Skills measured, Microsoft Foundry coverage, and expert strategy from a Microsoft Certified Trainer with 16 active certifications.
AI-102 retires June 30, 2026
If you are about to schedule the Azure AI Engineer exam, choose AI-103. AI-102 will sit closed after June 30, 2026 at 11:59 PM Central Time. AI-103 is the Microsoft Foundry-aligned exam your certification will renew against going forward.
Exam Overview
AI-103: Developing AI Apps and Agents on Azure validates that you can build, manage, and deploy AI apps and agents using Microsoft Foundry. As an Azure AI engineer, you collaborate with business stakeholders, solution architects, data scientists, DevOps engineers, and cloud security engineers to design, implement, and maintain AI solutions.
The exam expects practical experience developing apps in Python and familiarity with generative AI, agent frameworks, and the Microsoft Foundry stack — including the Foundry Agent Service, prompt flow, Foundry Tools, and Content Understanding.
Exam Details
- Exam code: AI-103
- Certification: Azure AI Engineer Associate
- Skills measured as of: April 16, 2026
- Duration: 120 minutes
- Passing score: 700 / 1000
- Cost: ~USD $165 (varies by region)
Who should take this exam
- Azure AI engineers building Foundry-based apps
- Engineers shipping agentic solutions in production
- Developers integrating generative AI into existing apps
- Solution architects validating their AI design skills
- AI-102 holders renewing into the current track
Skills Measured
Per the official Microsoft Learn study guide (skills measured as of April 16, 2026), the AI-103 exam covers five domains. The biggest weight shift from AI-102 is the new agentic solutions emphasis bundled with generative AI.
Plan and manage an Azure AI solution
25–30%- Choose appropriate Microsoft Foundry services for generative AI and agents
- Choose retrieval, indexing, memory, tool, and knowledge integration services
- Set up AI solutions in Foundry: infrastructure, deployment options, CI/CD
- Manage quotas, scaling, rate limits, and cost footprints for model and agent workloads
- Monitor performance, drift, safety events, grounding quality, search index health
- Configure security: managed identity, private networking, keyless credentials, role policies
- Implement responsible AI: safety filters, guardrails, content moderation, harm detection
Implement generative AI and agentic solutions
30–35%- Build generative AI solutions with Microsoft Foundry (hubs, projects, deployments)
- Implement prompt flow solutions and RAG patterns grounded in your data
- Use Azure OpenAI in Foundry Models for content generation, code, DALL-E, multimodal
- Configure parameters, monitoring, tracing, feedback, and model reflection
- Create custom agents with the Microsoft Foundry Agent Service
- Implement complex agents with Microsoft Agent Framework
- Orchestrate multi-agent solutions, multi-user workflows, and autonomous capabilities
Implement computer vision solutions
10–15%- Analyse images: tags, descriptions, objects, people, OCR
- Implement custom vision models for classification and object detection
- Analyse videos and video indexer scenarios
- Use Azure AI Vision and Foundry vision capabilities in production
Implement text analysis solutions
10–15%- Analyse text: sentiment, key phrases, entities, language detection
- Implement translation, including custom translation models
- Build NLP pipelines that route text through the right Foundry service
Implement information extraction solutions
10–15%- Implement Azure Document Intelligence in Foundry Tools
- Use prebuilt and custom document intelligence models, including composed models
- Extract information with Azure Content Understanding in Foundry Tools
- Process and ingest documents, images, video, and audio with Content Understanding
- Provision Azure AI Search: indexes, skillsets, custom skills, semantic and vector stores
Recommended Study Plan
Built from real candidate progress data across AI-102 and AI-103 cohorts. Allocate roughly 80 hours over 6-8 weeks; AI-102 holders typically need 25-30 hours to upskill.
Microsoft Learn AI-103 learning paths
Start with the official AI-103 collection on Microsoft Learn. Cover the Foundry, agents, generative AI, vision, NLP, and information extraction paths. Free, sandbox-enabled.
Hands-on with Microsoft Foundry
Provision a hub and project. Deploy a model, build a prompt flow, ground it with RAG, instrument it with tracing. Then build a custom agent with the Foundry Agent Service and add tools.
Build a multi-agent solution
Use the Microsoft Agent Framework to compose a planner + specialist pattern. Implement evaluation, content safety, and a CI/CD pipeline. This single project covers the heaviest exam weight.
Review the under-weighted but tricky areas
Computer vision, text analysis, and information extraction together account for ~35% of the exam. Don't ignore them — Document Intelligence and Content Understanding in particular have subtle Foundry Tools positioning you should know.
Practice assessment + mock
Use the official Microsoft Learn AI-103 practice assessment (free). Rerun until you consistently score 80%+. Time-box your attempts to the real 120-minute window.
Tips From The Trainer
- Learn the Foundry vocabulary precisely — "Foundry Tools", "Foundry Agent Service", "Microsoft Foundry" replace older "Azure AI Studio" wording and the exam tests recognition.
- Practise writing safety filters and guardrails as code, not just clicking through Studio. The exam includes scenario-based responsible AI questions.
- Memorise when to choose Document Intelligence vs Content Understanding — the boundaries are subtle and frequently tested.
- For agent orchestration, know the difference between planner-driven, autonomous, and multi-agent patterns and when each is appropriate.
- Don't skip the cost, scaling, and security domain — it's 25-30% of the exam but candidates routinely under-prepare it.
- AI-102 holders: focus your 30 hours on the agentic + Foundry-positioning material. The vision/NLP/extraction topics overlap heavily with what you already know.
Frequently Asked Questions
What is the AI-103 certification and how does it replace AI-102?
AI-103: Developing AI Apps and Agents on Azure is the successor to AI-102. It validates your ability to build, manage, and deploy AI apps and agents that take advantage of Microsoft Foundry — including generative AI, agentic solutions, computer vision, text analysis, and information extraction. AI-102 retires on June 30, 2026 at 11:59 PM Central Time.
What is the difference between AI-102 and AI-103?
AI-103 reflects the move to Microsoft Foundry and adds significant weight on agentic solutions and generative AI (30-35% of the exam). Skills are now grouped into five domains: Plan and manage an Azure AI solution (25-30%), Implement generative AI and agentic solutions (30-35%), Implement computer vision (10-15%), Implement text analysis (10-15%), and Implement information extraction (10-15%). The exam expects working knowledge of Python.
What are the prerequisites for AI-103?
Microsoft expects you to be an Azure AI engineer with experience developing apps in Python and familiarity with general AI, generative AI, and Azure services. While not required, AI-900 fundamentals or prior AI-102 experience provides a strong foundation. Practical experience with Microsoft Foundry, prompt flow, and agent frameworks is highly beneficial.
What does the AI-103 exam cover specifically?
Five domains: (1) planning and managing Azure AI solutions including responsible AI across generative and agentic systems, (2) implementing generative AI and agentic solutions including the Microsoft Foundry Agent Service and multi-agent orchestration, (3) computer vision, (4) text analysis and translation, and (5) information extraction using Document Intelligence and Content Understanding in Foundry Tools.
How long does it take to prepare for AI-103?
For someone with Azure fundamentals and Python experience, 6-8 weeks of focused study is typical. Engineers transitioning from AI-102 typically need 3-4 weeks to upskill on Microsoft Foundry, agent frameworks, and the expanded agentic domain. Structured training with hands-on Foundry labs accelerates readiness significantly.
Ready to Pass the AI-103 Exam?
Get expert-led AI-103 preparation from a Microsoft Certified Trainer with 16 active certifications. Hands-on Microsoft Foundry labs, agent-building practice, and the exam strategy that has worked for hundreds of candidates.