Developing AI Apps and Agents on Azure Practice Exams

Exam AI-103, Developing AI Apps and Agents on Azure, is the successor to AI-102 and the current Azure AI engineering certification. Its skills measured took effect on 16 April 2026, and the rebuild is substantial rather than cosmetic: the exam is organised around Microsoft Foundry, and agents are a first-class subject rather than a late addition. The five domains are Plan and manage an Azure AI solution (25-30%), Implement generative AI and agentic solutions (30-35%), Implement computer vision solutions (10-15%), Implement text analysis solutions (10-15%), and Implement information extraction solutions (10-15%). Two thirds of the exam therefore sits in planning, generative AI, and agents. The computer vision domain has changed character completely — Custom Vision, the Face service, and spatial analysis are gone, replaced by image and video generation, inpainting and mask-based editing, multimodal understanding, and accessibility alt-text. Knowledge mining has become information extraction, centred on retrieval and grounding pipelines, hybrid and vector search, and Content Understanding analysers. Candidates are expected to develop in Python. The exam is scored out of 1000 with 700 to pass.

• 1The current Azure AI engineering certification — AI-102, its predecessor, was retired by Microsoft on 30 June 2026
• 2Agent development is a first-class exam domain, matching how AI engineering roles are actually being defined
• 3Two thirds of the exam covers planning, generative AI, and agents, so it validates the skills employers are hiring for now
• 4Built entirely around Microsoft Foundry, the platform Microsoft's AI tooling has consolidated onto
• 5Associate-level with no formal prerequisite, so it is reachable directly for developers moving into AI engineering

• ✓Choose appropriate models and Foundry services for generative, grounding, vector search, agent, and multimodal tasks
• ✓Design Azure infrastructure for AI apps and agents, and configure model and agent deployments
• ✓Integrate Foundry projects into CI/CD pipelines
• ✓Manage quotas, scaling, rate limits, and cost for model and agent workloads
• ✓Monitor model performance, drift, safety events, grounding quality, and search index health
• ✓Secure AI systems with managed identity, private networking, keyless credentials, and role policies
• ✓Configure safety filters, guardrails, risk detection, and content moderation, and govern agent behaviour
• ✓Build generative applications with RAG, tool-augmented flows, and multistep reasoning pipelines
• ✓Build agents with roles, goals, tool schemas, retrieval, function calling, and conversation memory
• ✓Implement orchestrated multi-agent solutions with approval flows and safeguards
• ✓Evaluate models and agents for fabrication, relevance, quality, and safety, and set up tracing and token analytics
• ✓Implement image and video generation, inpainting, and mask-based editing workflows
• ✓Build multimodal understanding solutions including captioning, visual question answering, and alt-text generation
• ✓Extract entities, topics, summaries, and structured JSON from text, and implement speech-to-text and text-to-speech for agents
• ✓Build retrieval and grounding pipelines with semantic, hybrid, and vector search, and extract document content with Content Understanding