LexisNexis Redefines Legal AI with Focus on Small, Specialized Models

In the world of legal research and practice, the adage “bigger is better” has long been the guiding principle for technology adoption. Yet, in a surprising twist, LexisNexis is championing a different narrative, one that echoes the wisdom of “less is more.”

With the development of their AI assistant, Protégé, LexisNexis is not simply playing with the latest technology but redefining its potential by embracing the power of small, fine-tuned models.

When LexisNexis embarked on creating Protégé, they sought to sidestep the colossal complexity (and cost) associated with deploying large language models (LLMs). Rather than unleashing a massive, generalized AI juggernaut, they opted for a more nuanced approach—distilling larger models into smaller, specialized ones.

This strategic pivot not only aligns with the company’s commitment to efficiency but also illustrates a growing trend in AI development: the rise of small language models (SLMs).

Small models, often overshadowed by their larger counterparts, are proving to be surprisingly effective for specific applications.

LexisNexis’s decision to use these models is not a concession to limitations but a conscious choice to harness precision and speed.

Jeff Riehl, CTO of LexisNexis Legal and Professional, notes that by adopting a multi-model approach, the company can cherry-pick the best model for each task.

This ensures that the AI is not just a one-size-fits-all solution but a tailored assistant that understands the nuances of legal workflows.

Protégé is designed to be a paralegal’s best friend—an AI that helps draft and proof legal documents, ensuring every citation is accurate and every brief is bulletproof.

But its capabilities go beyond mere automation.

Protégé learns a firm’s unique workflow, offering suggestions for legal strategies and even drafting questions for depositions.

It’s a tool that doesn’t just perform tasks but anticipates needs, reflecting LexisNexis’s vision of creating personalized AI experiences for lawyers across various specialties.

The decision to focus on small models is not without its challenges.

It requires a meticulous architecture where tasks are broken down, and the appropriate model is summoned to perform specific functions.

For instance, a fine-tuned Mistral model might assess a query’s intent, while a more sophisticated LLM is reserved for generating complex legal arguments.

This dynamic switching ensures that Protégé is both fast and accurate—qualities that are non-negotiable in the legal realm.

LexisNexis’s journey with AI is not new, but the landscape drastically changed with the advent of generative AI technologies, such as OpenAI’s ChatGPT.

The conversational prowess and generative capabilities of these models have opened new doors, prompting LexisNexis to explore innovative ways to integrate these features into their platforms.

Protégé now stands as a testament to the company’s forward-thinking approach, not just following trends but setting them.

In the competitive arena of legal technology, LexisNexis faces formidable challengers.

Companies like Thomson Reuters and Harvey are also making strides with their AI assistants.

Yet, Protégé’s emphasis on customization and efficiency might just give it an edge.

By focusing on small yet mighty models, LexisNexis is not just building an AI but crafting a legal companion that promises to revolutionize how law firms operate.

In an industry where precision is paramount, LexisNexis’s Protégé exemplifies how innovation does not always require scaling up.

Sometimes, the most profound advancements come from scaling down, refining, and focusing on what truly matters.

As Protégé continues to evolve, it may well become not only a paralegal’s ally but a beacon for the future of AI in law.