Beyond RAG: How LexisNexis is Building the Next Generation of AI for Legal and Business

As artificial intelligence rapidly evolves, enterprises are prioritizing accuracy in AI model outputs. However, in high-stakes industries like law, accuracy is only the beginning. Trustworthy AI requires relevancy, authority, accurate citations, and a low rate of hallucinations. To address these challenges, LexisNexis has moved beyond standard Retrieval-Augmented Generation (RAG) to incorporate graph RAG and agentic graphs, alongside “planner” and “reflection” AI agents designed to critically evaluate their own outputs.

The Limitations of Accuracy Alone

“There’s no such [thing] as ‘perfect AI’ because you never get 100% accuracy or 100% relevancy, especially in complex, high stake domains like legal,” acknowledges Min Chen, LexisNexis’ SVP and chief AI officer LexisNexis. The focus, shifts to managing uncertainty and consistently delivering customer value. “At the end of the day, what matters most for us is the quality of the AI outcome and that is a continuous journey of experimentation, iteration and improvement,” Chen stated.

Evaluating AI Usefulness: Beyond Relevancy

LexisNexis evaluates its models using several “sub metrics” to measure “usefulness,” focusing on factors like authority, citation accuracy, hallucination rates, and “comprehensiveness.” Comprehensiveness assesses whether a generative AI response fully addresses all aspects of a user’s legal question. A response addressing only three out of five legal considerations, while relevant, is considered incomplete and potentially misleading.

the utility of citations is paramount. Citations must be legally valid; arguments or instances overruled by courts are deemed “not citable” and not useful LexisNexis.

From RAG to Graph RAG and Agentic Graphs

LexisNexis launched Lexis+ AI, a legal AI tool for drafting, research, and analysis, in 2023, initially built on a standard RAG framework and hybrid vector search leveraging LexisNexis’ knowledge base. The company then introduced Protégé in 2024, a personal legal assistant that incorporates a knowledge graph layer on top of vector search.

While semantic search excels at retrieving contextually relevant content, it doesn’t guarantee authoritative answers. LexisNexis’ approach involves initial semantic search followed by traversing results across a “point of law” graph to filter for the most authoritative documents. The company is further developing agentic graphs to automate complex, multi-step tasks.

AI Agents for Planning and Reflection

“Planner agents” break down user questions into sub-questions, allowing for human review and refinement. A “reflection agent” dynamically critiques and refines draft documents. However, LexisNexis emphasizes that these advancements are designed to augment, not replace, human expertise. Chen envisions a future of “deeper collaboration between humans and AI.”

Key Components for AI Success

LexisNexis highlights the importance of balancing cost, speed, and quality when implementing AI solutions. Enterprises should define Key Performance Indicators (KPIs) and success metrics before beginning experimentation.