Kibeom Lee
Head of AI/ML at delight.ai
The failure that scares us most looks exactly like success.
Most AI agent failures don't look dramatic in the moment. An agent gets talked into a conversation it should have shut down, or quietly invents a policy that was never real and a customer acts on it before anyone notices. The customer usually finds the problem before the company does.
Join Kibeom Lee, Head of AI/ML at delight.ai, as he breaks down the Trust Stack, the four-layer framework delight.ai uses in production to catch failures before your customers do. The goal isn't a smarter model. It's a system that catches itself, and that earns AI agents more autonomy without losing control of what they do.
What you'll walk away with
- How to tell if an agent's "success" is actually hiding a failure
- The four-layer framework for earning AI agents more autonomy, one layer at a time
- A simple way to check whether any workflow you're automating is ready for more trust
Save your seat and see exactly where your own AI agents stand on the trust ladder.
