Build a generative AI solution that's delivering value

Overview

This is not a session about how great AI or LLMs are or if they are good or not at counting the number of r's in "Strawberry". It's about how I'm leveraging agents and language models in my current project to deliver value for our customers. If there's one thing language models are good at, it's interpreting language and data. Especially if it's structured data with useful descriptions. In that sense, it's much like ourself. In my project, we have a lot of structured, financial, data. We now need to do analysis on the data and make sure calculations are deterministic and always correct. That means I can't rely on the nondeterministic nature of a language model. By combining the Microsoft Agent Framework, Foundry agents, and a mix of .NET and Python components, I’ve built a high‑performing solution we can trust. You'll leave the session knowing how to orchestrate agents and keep control over the accuracy & reliability. All based on my own experience working in the ecosystem.