While the world is busy chasing the latest LLM release, architects and engineers are solving a very different problem: how to make AI actually do work.

This session explores the shift from prompt-driven AI to agent-driven systems—where AI doesn’t just respond, but orchestrates workflows, maintains state, invokes tools, enforces policies, and operates under enterprise-grade observability.

Using Microsoft Agent Framework as the backbone, we’ll dive into how Agent Hubs are designed in real-world enterprises—connecting models, tools, data, and governance into a single, operational system. You’ll see how vibe coding accelerates agent creation using tools like GitHub Copilot, Antigravity, and Kiro, without sacrificing architecture discipline, security, or control.

Through engaging examples and practical demos, we’ll break down what architects and senior developers must think about when building agentic AI systems—from policy enforcement and logging to scalability and cost control.

This is the difference between talking to AI and having AI work for you.

What You’ll Learn

✔ Real-world use cases built using agent frameworks
✔ What it takes to build enterprise-ready AI agent applications
✔ How to design and operate an AI Hub / Agent Hub
✔ Microsoft products that support this architecture end-to-end
✔ How to create AI agents using Microsoft Agent Framework + vibe coding tools