Abstract: When it comes to developing AI agents that can deliver reliable real-world results, the possibilities are broad and exciting… but the challenges can be profound and unexpected. Join Cai GoGwilt, Co-Founder and Chief Architect of Ironclad, as he shares the lessons learned in building the first-ever agent capable of performing complex legal contract analysis. This agent acts like a “talented intern,” capable of using tools independently, processing multiple logic loops, and applying imagination and creativity while still delivering precise results. It was designed to be immediately accepted by, and useful to, untrained business analysts and legal staff.
This session will help attendees understand how to:
- Make self-managing generative AI work reliably. AutoGPT-like agents have great potential, but typically fail because they hit a “hallucination wall” when using multiple tools across multiple steps. Learn how the team overcame this through careful balance and testing.
- Overcome the complexity of multi-loop agentic programming. When skyrocketing complexity threatened the project, the team designed an AI-specific visual IDE. Find out how to use this breakthrough tool for your own development.
- Design a user experience that works. AI agent design requires a very different mindset and approach. Learn how to think about your users and their needs in a new way.
This session is relevant for anyone interesting in the design and development of complex AI agents, particularly those meant to perform in a rigorous enterprise context.
Bio: Cai GoGwilt has always been passionate about building products that make people more effective at their jobs. He is an MIT graduate in Computer Science and Physics, as well as a former software engineer at Palantir Technologies. By day, he’s a celebrated software engineer and thoughtful leader; by night, he’s a concert cellist and karaoke enthusiast.