LexisNexis builds agentic graph RAG and self-critiquing AI for legal accuracy

LexisNexis evolved beyond standard RAG to graph RAG and agentic graphs, building "planner" and "reflection" AI agents that parse legal requests and critique their own outputs. The approach uses sub-metrics including authority, citation accuracy, hallucination rates, and comprehensiveness. The company's SVP and chief AI officer acknowledged that "perfect AI" is impossible in high-stakes legal domains, but said the goal is managing uncertainty through continuous experimentation.

View full digest for February 19, 2026