NetDocuments today (18 August) published a new Legal Context Engineering Benchmark (LCEB) report, which measures how structured legal context affects both the quality and cost of AI answers. The report focuses at its core on the performance of NetDocuments’ Legal Context Graph, which gives AI agents organisation-wide context, with permissioning and walls preserved.
The benchmark report asks what changes when the model, agent, questions and documents stay the same, but the AI has structured context to help it understand and navigate those documents?
The results suggest that the Legal Context Graph cut the cost of a correct AI answer by 48% while maintaining essentially the same answer quality. For a 2,000-person firm asking four million AI questions a year, the report estimates that could translate to nearly $1 million in annual savings. The report is timely, given the focus on costs now that tech companies are shifting from a subscription to consumption pricing model.
This is the point made by NetDocuments’ chief executive officer Josh Baxter, who said: “As AI shifts to consumption pricing, the industry is swinging between two extremes: spending without limits and cutting without strategy. We’ve believed from the beginning that context is the lever that changes that equation. This benchmark is the first proof. Get the context right, and firms spend less to get better answers. That’s not a feature. That’s a new foundation for how legal AI should work.”
The benchmark covered 300 questions across ten real matters: 874 documents and about 60 million characters drawn from public regulatory filings and court dockets. You can read it HERE