Azure Decoded: The Developer's Guide to AI Agents on Microsoft Foundry

Ready to build AI that actually does things? In this hands-on session, Aaron Stark walks you through creating your first AI agent using Microsoft Foundry—Microsoft’s unified platform for building, deploying, and managing AI applications. You’ll learn how to move beyond simple chat completions to agents that can reason, plan, and take action. We’ll cover the core concepts of agentic AI, explore Microsoft Foundry’s agent capabilities, and build a working agent live—from prompt engineering to tool integration to deployment. Whether you’re an AI-curious developer or already building with Azure OpenAI, this session gives you the practical skills to start building agents that can transform how your applications interact with the world. What you’ll learn: What makes an AI agent different from a chatbot Microsoft Foundry’s agent development capabilities How to design agent workflows with reasoning and tool use Connecting agents to external data and APIs Best practices for testing and deploying agents in production Chapters: 0:08 Welcome & Housekeeping 0:21 Code of Conduct and Chat Guidelines 0:47 Session Recording and On‑Demand Info 0:59 Speaker Introductions 1:11 Overview: AI Agents on Microsoft Foundry 2:26 What Is an Agent? Reason, Act, Learn 3:26 Agent Feedback Loops and Tool Use 4:49 Evolution from Chatbots to Agents 5:19 Developer Perspective on Agent Reasoning 6:39 Industry Evolution and Agent Maturity 7:19 Friction in Scaling Agentic Systems 8:49 From Automation to Adaptive Agents 10:10 Why Build Agents in Code vs Low/No Code 11:01 Developer Experiences with Code‑First Agents 14:26 Benefits of Code‑First Agent Development 16:26 Challenges Building and Scaling Agents 18:29 Fragmented Frameworks and Tooling Gaps 20:24 Semantic Kernel vs Agent Framework 23:09 Human‑in‑the‑Loop and Risk Management 26:01 Practical Boundaries for Agent Automation 28:55 Introducing Microsoft Agent Framework 32:03 What a Modern Agent Platform Needs 34:19 Azure AI Foundry and Developer Options 38:32 Agent Framework Architecture Overview 41:18 Gaps Without an Agent Framework 44:19 Semantic Kernel and AutoGen Convergence 47:24 Why Agent Framework Feels Different 50:37 Unified SDK and Agent Primitives 52:11 First “Hello World” Agent Experience 54:01 From Proof of Concept to Production 55:02 Local‑First and Cloud‑Agnostic Development 56:28 Multi‑Agent Orchestration Patterns 58:18 Key Takeaways and Future Episodes 59:26 Closing Remarks and Series Wrap‑Up #MicrosoftReactor #learnconnectbuild [eventID:26778]