Architecting AI Agents with Microsoft Agent Framework
summary
Shows a proof-of-concept refund agent built with Microsoft Agent Framework and C# that uses Azure AI Foundry (GPT 5.4 mini) plus MCP/function tools, memory and workflow orchestration. Demo runs a customer lookup, checks payments, and refuses automatic refunds above $100 — a $150 case triggers human-in-the-loop approval. Framework supplies harness, middleware and observability; implementation is a POC console app, not yet production, and demonstrates tool-calling, durable state and workflow checkpointing in practice.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:29From deterministic to agentic
- 2:33Traditional vs agentic dev
- 5:20What defines AI agents
- 6:29Agent anatomy and flow
- 9:00Need for robust architecture
- 11:28Microsoft Agent Framework
- 15:25Key framework capabilities
- 17:16Single vs multi-agent choices
- 18:22Human-in-the-loop workflows
- 19:27Refund agent demo
- 27:31Q&A
related talks
05h 17mBuilding AI Applications
36mBuilding AI Agents in .NET with Azure AI Foundry
58mAdvanced Autonomous AI Agents: Memory, Collaboration and Evolution in Action
39mEvolving Intelligence: Designing Autonomous AI Agents
56mBuilding AI Agents and Multi-Agent Systems with Azure AI
01h 54mBuilding Code First AI Agents with Azure AI Agent Service
26mAI powered automation & multi-agent orchestration
22mMaking AI Accountable: Azure AI Foundry