An attendee-friendly schedule for the day

Have you heard the trend of people deploying code anywhere from checkout lines to doctors’ offices? In this talk, you will learn how to talk to your phone and have an app on Azure to show your friend 5 minutes later. The basic DevOps tools you already know, plus a little bit of AI magic, are all you need!

If you want to follow along, the demo will be in Azure with GitHub Actions and GitHub Copilot, and you can have it set up on your phone by the end of the session. Just come in with an Azure account and a GitHub repo.

AI agent architectures are quickly becoming a powerful pattern for building intelligent applications, but production systems must still handle failures, scaling events, and long-running execution.

In this session, we’ll explore how Durable Agents, built on Azure Durable Functions, provide a reliable execution model for these systems. After a brief introduction to the Microsoft Agent Framework, we’ll examine how the Durable Task extension enables resilient orchestration for distributed agent execution.

We’ll also demonstrate how Azure Durable Task Scheduler and Aspire can be used to develop, run, and observe production-ready agent systems on Azure.

Every new agent session is Frosty the Snowman putting on the hat — “Happy Birthday!” — cheerful, capable, and blissfully unaware of everything you’ve done together. Beads, Steve Yegge’s Dolt-backed execution tracker, solves this by giving agents structured external memory with dependency tracking, semantic compaction, and queryable state. But memory is just the foundation — the real value is in the workflow patterns Beads enables as you scale. This talk walks through those patterns at three levels: a single agent managing session-to-session continuity, a small team of parallel agents coordinating through dependency graphs and atomic claims, and full agent swarms where twenty-plus agents grind through epics with role hierarchies and convoy-based work distribution. You’ll leave with a practical understanding of which patterns matter at each scale, the architectural decisions that make them possible, and why execution-centric memory may be more important than the orchestration framework you pick.

For organizations looking to get actionable insights out of their documents, choosing the right solution is crucial to create a streamlined extraction process. Azure Content Understanding is a service that goes beyond traditional OCR services by providing analyzers for documents, audio and video as well as enrichment capabilities.

In this talk, we will evaluate how Azure Content Understanding compares with other document extraction solutions in the Azure ecosystem, we will then look at the different analyzers available as part of the service, and we will finish by looking at how this service can be used to implement a RAG solution.

By the end of this talk, attendees will understand when and why to use content understanding and how to apply its capabilities in different business scenarios.

Most teams use GitHub Copilot like a smarter autocomplete. The faster teams have reorganized how work gets done around agents that plan, code, deploy, and ship.

The agent isn’t the point. The system your team builds around it is: shared instructions, prompt libraries, agent definitions, and governance controls that encode how your team works into the tools themselves. Add Microsoft’s official Azure Skills and the Azure MCP Server, and those agents can provision resources, troubleshoot deployments, and operate your Azure environment — without leaving the editor.

I’ll show you how to build and govern that system. Then we’ll demo it live: starting from a GitHub issue, watch Copilot’s coding agent write the code, wire up the Azure Skills, deploy to Azure, and open a pull request — all while we narrate what’s happening and why.

Key Takeaways:

• How instructions, prompts, agents, skills, and MCP tools compose together in .github/
• Live demo: using Microsoft’s Azure Skills and @azure to deploy and operate Azure resources from Copilot
• Shared assets — copilot-instructions.md, AGENTS.md, PR templates — that raise the floor for every developer
• Enterprise governance: model policies, audit logs, and the agent control plane
• How to move your team from solo wins to agentic-first workflows

Coming soon!