Case Study: DraftWise boosts legal search quality by 30% with Cohere's models
Key results
The challenge
DraftWise builds an AI platform for drafting, reviewing, and negotiating contracts, aiming to reduce the time lawyers spend on routine, repetitive drafting tasks. Surfacing the most relevant clauses and precedents from past agreements required high-quality semantic search across large document sets.
The solution
DraftWise adopted Cohere's Command model for text generation and the Embed and Rerank models for search and retrieval, using a retrieval-augmented generation system with fine-tuned Cohere models. The stacked models return relevant clauses and examples with citations to source documents.
“Cohere boosted search quality by 30%, giving lawyers the right proprietary data at their fingertips.”
JDJames DingCo-Founder and CEO, DraftWise
The results, in context
DraftWise's internal benchmarks showed a 30% improvement in search-result quality after incorporating the fine-tuned Cohere models, and the backend saw a threefold increase in API calls to Cohere. The company also reported a sharp rise in usage.