What's the deal with multi-agency?! – Architectures and frameworks for building multi-agent systems

Ved Henri Schulte, Microsoft

In the world of AI, 2024 has undoubtedly been the year of AI agents. 2025, on the other hand, looks to be the year where we push AI agents to their limit. Where do we go from here? Add even more agents?! 

In this session, we'll demystify Agentic AI, from the first AI agents to the rise of multi-agency. Along the way, we will discover how AI agents work under the hood, which components are needed to build effective AI agents, and how we can orchestrate them in multi-agent systems to solve complex problems. I will share my perspective on why single agents are no longer enough, what different architectures we can use to remedy that, as well as some of the tools and open-source frameworks that Microsoft itself uses to build AI applications, and how you can use them in your own projects. If all goes well, we might even be able to observe a multi-agent system in action! 

Bio: With a background in Software Development and Natural Language Processing, I am now working as Microsoft Denmark's Partner Solution Architect for AI. For the past five years, I have for been advising public and private organizations across Denmark on building AI systems in the cloud. Throughout this time, I have been working hands-on with Large Language Models since their inception and count myself among one of the earliest Generative AI advocates in Denmark.

Optagelsen stammer fra dette års Driving AI afholdt i København den 20. maj 2025, hvor 250 AI-professionelle og entusiaster mødtes for at høre om nye tendenser og teknologier inden for AI.