Imagine typing a message in Microsoft Teams and kicking off an AI workflow that automatically spreads across Azure Functions, Container Apps, and Web Apps all stitched together as a single application through Durable Task Scheduler. It ties these services into one coordinated workflow, scheduling activities to the right machine, fanning out to leverage each service's scale, and orchestrating the results back together.

Each AI agent checkpoints its progress along the way, so if anything fails, the system picks up right where it left off. And when it's done, the result lands right back in the Teams channel where it all began.

In this talk, we'll show how Durable Task Scheduler lets you compose an app from your existing Azure services, now enhanced with AI-native capabilities: natural language agents on Functions equipped with skills to react to Teams messages and schedule orchestrations, durable Copilot agents with MCP tools and GPU inference on Container Apps, and front-end experiences on Web Apps. You'll see how agents built with the GitHub Copilot SDK become durable and resilient with automatic checkpointing, and how Durable Task Scheduler connects all of this without needing a separate message broker or custom routing logic.

Walk away with a clear, reusable pattern for building AI workflows where every service contributes what it's best at and meets your users where they already work.