8/22/26

Build AI Agents in Azure Logic Apps (Conversational + Autonomous)

For a long time most of us treated Azure Logic Apps as the integration layer. It moved data between systems, it handled retries and connections, and it stayed in the background. That is no longer the whole picture.

In this session Kent Weare, David Lorenzo and I go through what Logic Apps can do now. You can build conversational agents that talk to people and autonomous agents that run on their own. You can expose your workflows as MCP tools so that other agents can call them. You can use the A2A protocol so your agent can work together with agents built on completely different platforms. We explain the agent loop, how tools and knowledge are added, and what actually changes between a conversational agent and an autonomous one. We show Logic Apps working as an orchestrator, connecting Azure Document Intelligence, Blob Storage and Microsoft Foundry. We build a conversational agent, then an autonomous one, and we look at where each of them makes sense in real work. We also demo a claims processing agent from end to end, where Logic Apps does the orchestration and the final output is generated by the new Copilot Studio workflows. The last part is about auto.azure.com, and for me this is the most interesting change. The Azure portal was always a barrier for business users, and for good reasons around access and governance. Logic Apps Automation gives those users their own experience where they can build automations without entering Azure, with scoped permissions. It is still in preview, but the direction is clear.

The session is long because we wanted to show real things instead of slides. All data used in the demos is synthetic.

Next

How to Run Discovery Sessions for Copilot Studio Agents | EP 04