Legal research company LexisNexis recently launched Protégé, an AI assistant designed to support paralegals, assistants, and legal professionals. Protégé's primary function is to help users draft and proofread legal documents, ensuring accurate citations in pleadings and briefs. LexisNexis aims to create an intelligent assistant that goes beyond a generic legal AI tool, learning the workflows of individual law firms and providing customized services.
In developing Protégé, LexisNexis leveraged advanced large language models (LLMs) from Anthropic and Mistral, experimenting to find the models best suited to user needs. LexisNexis CTO Jeff Riehl stated, "We select the optimal model based on the specific use case to achieve the best performance and fastest response times." In some instances, LexisNexis opts for smaller language models (SLMs) or distills larger models to enhance performance and reduce costs.
Distillation, where a larger model "teaches" a smaller one, is a growing practice among organizations. Smaller models often perform better in applications like chatbots or simple code completion, a goal LexisNexis sought for Protégé.
LexisNexis has a history of AI application, exploring natural language processing, deep learning, and machine learning before launching the LexisNexis + AI Legal Research Center in July 2024. Riehl noted that the capabilities of generative and conversational AI, since the advent of ChatGPT in November 2022, have been particularly exciting.
When building its AI platform, LexisNexis adopted a multi-model approach, utilizing models from various providers such as Claude, OpenAI's GPT, and Mistral. Riehl explained, "We break down each user's task into individual components and find the best large language model to support each component." For example, when a user inputs a query, Protégé first uses a fine-tuned Mistral model to analyze the query before switching to the model best suited for completing the task.
Currently, LexisNexis primarily relies on a fine-tuned Mistral model, with plans to incorporate the Claude model in other areas. The company is also considering OpenAI's newer models and Google's Gemini model to further enhance its AI platform's capabilities.
Protégé is more than just a legal assistant; it helps draft legal documents, suggest next steps in legal workflows, generate questions, link citations for accuracy, generate timelines, and summarize complex legal documents. Riehl points out that Protégé represents an initial foray into personalized and intelligent legal assistance, aiming to provide each legal professional with a personalized assistant to support their specific work needs.
Key Highlights:
📜 Protégé is LexisNexis' AI assistant, designed to provide personalized support for legal professionals.
⚙️ LexisNexis utilizes a variety of large and small language models to improve the AI assistant's speed and accuracy.
🔍 Protégé assists in drafting legal documents, generating questions, and summarizing complex content, driving the intelligent evolution of legal work.
