LexisNexis, a leading legal research company, has introduced Protégé, an AI assistant designed to support lawyers, paralegals, and legal associates in drafting and proofreading legal documents, ensuring citation accuracy, and streamlining workflows. Unlike generic AI tools, Protégé is tailored to adapt to a law firm’s specific processes, offering a more personalized and efficient solution for legal professionals.
The Power of Small, Fine-Tuned Models
To build Protégé, LexisNexis adopted a multi-model approach, leveraging both large language models (LLMs) and smaller, fine-tuned models. Jeff Reihl, Chief Technology Officer of LexisNexis Legal and Professional, explained that the company selects the best model for each specific task, balancing performance and response time.
“We use the model that provides the best result with the fastest response time,” Reihl told VentureBeat. “For some use cases, that will be a small language model like Mistral, or we perform distillation to improve performance and reduce costs.”
Distillation, a process where a larger model “teaches” a smaller one, has become a popular method for optimizing AI applications. Small models are particularly effective for tasks like query assessment, code completion, and chatbot interactions, making them ideal for Protégé’s functionality.
A Multi-Model Architecture
LexisNexis employs a variety of models from leading providers, including Anthropic’s Claude, OpenAI’s GPT, and Mistral. The company’s platform is designed to switch between models based on the task at hand. For example, Protégé uses a fine-tuned Mistral model to assess user queries and determine their intent before routing the task to the most suitable model for completion.
“You don’t need GPT-4 to assess a query,” Reihl said. “We use smaller models for simpler tasks and reserve larger models for more complex work.”
LexisNexis also integrates its own knowledge graph to enhance retrieval-augmented generation (RAG) capabilities, ensuring that Protégé delivers accurate and contextually relevant results.
Protégé’s Role in Legal Workflows
Protégé is part of the LexisNexis + AI platform, which consolidates the company’s AI-driven legal tools. The assistant helps legal professionals with a range of tasks, including:
- Drafting legal briefs and complaints
- Suggesting next steps in workflows
- Generating deposition and discovery questions
- Linking quotes in filings for accuracy
- Creating timelines and summarizing complex documents
Reihl envisions Protégé evolving into a highly personalized assistant for legal professionals. “Our vision is that every lawyer will have a personal assistant tailored to their specific practice area, whether it’s M&A, litigation, or real estate,” he said.
Competition in Legal AI
Protégé enters a competitive landscape, with rivals like Thomson Reuters’ CoCounsel and Harvey, a legal AI startup backed by LexisNexis, offering similar solutions. These tools aim to revolutionize legal research and workflow automation, empowering legal professionals to work more efficiently.
As LexisNexis continues to refine Protégé, the company is exploring additional models, including OpenAI’s reasoning models and Google’s Gemini, to further enhance its capabilities.
The Future of Legal AI
Protégé represents a significant step toward integrating AI into the legal industry, offering tools that not only streamline tasks but also adapt to the unique needs of individual lawyers and firms. By combining small, fine-tuned models with advanced AI architectures, LexisNexis is paving the way for a new era of legal technology.
