CertifyDeckAI-103 (Microsoft Certified: Azure AI Apps and Agents Developer Associate) lasts 120 minutes and requires 700 out of 1,000 points. It brings development, architecture and operations together: this plan turns its five official domains into eight weeks of hands-on preparation centered on Microsoft Foundry, generative AI and agents.
What you need before starting AI-103
Microsoft expects Python development experience and familiarity with Azure, APIs and generative AI. Before studying specific products, make sure you can consume a model from code, manage credentials and explain embeddings, vector search and RAG.
Create a Foundry project and deploy a small model. Connect with keyless identity where possible and record latency, usage and errors. Those foundations recur throughout the AI-103 skills outline.
Prioritize Foundry, RAG and agents
Planning and management plus generative AI and agents account for 55–65% of the exam. Practice model selection, deployment, quotas, private networking, managed identity, safety filters, evaluation and tracing.
Implement an agent with retrieval, memory and a custom function. Learn when to use a deterministic flow, an autonomous agent, specialized multi-agent orchestration or a hybrid model-and-rules engine.
A practical eight-week plan
Weeks one and two: infrastructure, models, security and governance. Weeks three to five: RAG, tools, agents, orchestration, evaluation and observability. Weeks six and seven: vision, multimodal content, language, speech and document extraction.
During the final week, take unseen mock exams and revisit only the mistakes you cannot justify. Reproduce authentication, indexing, filtering, human approval and telemetry decisions in a lab.
When are you ready for AI-103?
You are close when you can design an end-to-end solution and explain every service, model, index, identity and control. Keep results consistent across unseen mock exams.
Always check the official guide shortly before the exam: Foundry services and SDKs change quickly, and Microsoft can update objectives or terminology. The AI-103 page organizes practice around the five official domains; coming from AI-102, read what changes with AI-103.
Frequently asked questions
How long is the AI-103 exam?
120 minutes. The passing score is 700 out of 1,000, and the exam is available in English, Spanish and several other languages.
Does AI-103 replace AI-102?
AI-103 evolves the Azure AI Engineer path toward generative applications and agents built with Microsoft Foundry. If you are starting now, prepare AI-103; if AI-102 was already underway, check Microsoft's official transition dates.
Does AI-103 need to be renewed?
Yes, annually through Microsoft Learn's free renewal assessment, like other role-based certifications.
How much Python do I need for AI-103?
Enough to consume models and Azure SDKs from code, handle credentials and implement an agent with retrieval and a custom function. It is not an algorithms exam.