The Last 5% Problem: Why Robots Still Need Humans to Scale

​Despite rapid advances in AI and robotics, most real-world deployments still break down in edge cases, the “last 5%” of situations that robots weren’t trained for. This talk explores why that gap persists and how human-in-the-loop systems turn these moments of failure into scalable infrastructure. We’ll show how combining autonomy with real-time human support unlocks reliable, cost-effective robotics beyond the lab.

​Ken Chester and Shefin Shajit will be speaking; the two (along with other team members) have built Anthropilot, a human-in-the-loop platform that provides on-demand, real-time guidance to help robots handle edge cases and operate reliably in the real world.