Legora's Blind AI Agent

🔥 What happened Legora, one of the fastest-growing legal tech companies, builds an AI agent for lawyers but can't see its work because the data belongs to law firm clients. CTO Jacob Lauritzen explains how his team tests the agent in simulated environments modeled on customer workflows, much like the “worlds” used in reinforcement learning. 💡 Why it matters Models already handle about 60% of legal work well, and the rest needs smarter models. One release alone lifted citation accuracy from 70% to 99% without Legora changing anything – proof that model upgrades can dramatically move product quality. ⚡ Our take Building agents for regulated industries means testing without seeing real data. Legora's “worlds” approach is a blueprint for any sensitive domain – and a wake-up call for teams still treating evals as an afterthought.