Build Agents Once, Run Them Anywhere: AI Systems That Survive Change

Models change. Vendors change. Infrastructure requirements change. Your agent code shouldn't have to. This talk shows how open standards (MCP, Kubernetes, OpenTelemetry, and standardized agent protocols) let you build AI systems that survive disruption without rewrites. Build your agents once, then run them anywhere: across clouds, models, and the edge. Using real-world patterns and a reference architecture, we'll explore how to decouple agents from specific models and vendors, why protocols outlive products, what happens when your chosen infrastructure becomes unavailable and how to handle it gracefully, and why Microsoft is betting on open standards as the foundation for AI. You'll walk away with a mental model for architecture that outlasts hype cycles.