CertifyDeckAI-103 evolves the Azure AI Engineer path (AI-102) toward generative applications and agents built with Microsoft Foundry. If you are starting now, prepare AI-103; if you were studying for AI-102, much of the foundation still helps, but the focus shifts from consuming isolated cognitive APIs to designing and operating generative AI and agent solutions.
What changes with AI-103
AI-103 places far greater emphasis on Foundry, model selection and operations, RAG, agents with memory and tools, multi-agent orchestration, evaluation, security and observability. Its outline puts 55–65% of the exam on planning and management plus generative AI and agents.
It also includes image and video generation and editing, multimodal reasoning, and Content Understanding across documents and media.
Which AI-102 skills still transfer
Experience with Azure, Python, managed identity, networking, Azure AI Search, language, speech, vision and document extraction remains valuable. Security, cost, scaling and monitoring foundations transfer too.
The difference is applying those capabilities within generative and agent solutions instead of consuming separate cognitive APIs alone.
Which exam to prepare
If you are starting now, follow the current AI-103 credential and study guide. If you already booked or completed AI-102, check Microsoft's official dates and transition terms before changing your plan.
Do not assume an older course covers the new taxonomy. Map every module against AI-103's five published domains.
How to adapt an AI-102 plan to AI-103
Keep your search, language, speech, vision, security and identity labs. Add a Foundry project with a deployed model, hybrid index, RAG, agent, custom tool, evaluation, filters, tracing and deployment automation.
Use new AI-103-aligned questions, such as those on the AI-103 page; repeating old material can conceal gaps in agents and generative AI operations. The eight-week plan organizes that shift in focus.
Frequently asked questions
Is AI-102 still available?
Check Microsoft's official dates: AI-103 is the current Azure AI Engineer path. If you already booked or passed AI-102, review the transition terms before changing plans.
Is AI-103 harder than AI-102?
It is broader in generative AI and agents and demands more architecture and operations decisions. Anyone already working with Foundry and RAG will find it natural; anyone who only consumed cognitive APIs will need to strengthen agents and evaluation.
Can an AI-102 course prepare me for AI-103?
Partly: it covers search, language, speech and vision. It misses Foundry, agents, orchestration, evaluation and observability, which carry the most weight in AI-103.