The legal field is not known for its collaborative environment. The adversarial nature of litigation—coupled with fierce competition for top corporate clients—often results in a siloed approach to client service and solution development.
That isolative culture will need to evolve as the practice of law becomes increasingly data driven. While the legal industry changes at a glacial pace, technology changes overnight. Its rapid evolution creates novel data challenges for legal teams that are best tackled through collaboration and sharing cross-functional expertise.
This is especially true within e-discovery. Innovations in AI can help corporate legal teams manage the tidal wave of e-discovery data challenges they face (skyrocketing data volumes, continuously emerging data sources, etc.). And there is no more fertile ground for AI innovation than strategic collaboration, where each partner can bring their own unique AI experience and expertise to the table and, together, assemble some pretty incredible things.
The partnership between Lighthouse and Cleary Gottlieb Steen & Hamilton LLP (Cleary) is a marquee example of the pragmatic innovation that this type of partnership breeds. The combination of advanced AI development resources and diverse legal and technological expertise has resulted in AI technology and workflows that overcome persistent data challenges for clients.
While the partnership’s innovative environment could be described as “lightning in a bottle,” the alignment was not formed by happenstance. Rather, it was forged around each partner’s individual strengths and a shared appetite for innovative solutions to client problems.
CJ Mahoney, counsel and head of the e-discovery and litigation technology group at Cleary explains it best: “It’s a true collaboration between Lighthouse experts and our experts. We make each other better.”
Here’s a look at the origin story of this partnership, and how relationships like this one help make our industry better year after year.
A Shared Appetite for Innovation
Mahoney was integral to the partnership’s beginnings. He was first drawn to Lighthouse years ago, as he searched for review solutions that allowed him to customize advanced technology for the benefit of his clients. Like many law firms at the time, Cleary used an array of solutions for different review tasks. But Mahoney often found their workflows rigid, limiting each tool’s ability to substantially affect the cost and burden of review.
With Lighthouse’s AI applications, Mahoney found the flexibility he was looking for. Large language models (LLMs)—which have begun making headlines in the legal industry—have been a part of Lighthouse’s approach to creating and deploying predictive AI solutions for the last four years. Lighthouse solutions in this vein had proven capable of making reliable classifications and were designed to build upon RelativityOne’s already robust review and analytics capabilities.
Speaking with the team at Lighthouse, Mahoney was thrilled to find the AI-backed talent and experience he needed from a collaborator. “I can be ambitious because Lighthouse has true data scientists, AI deployment experts, and former attorneys backing me up,” he says. The combination of creativity and expertise made for smart engineering and a lot of opportunity to move the needle on particularly cumbersome reviews and build unique solutions for clients.
A Partnership Founded on Innovating with AI
One of the first big wins using AI for the Cleary + Lighthouse partnership happened on a challenging merger control matter with a government regulator. The legacy technology-assisted review workflows Cleary was using to identify a responsive set struggled with the data set, which featured very low richness (i.e., very low rates of responsiveness across the overall document population). As a result, the project’s precision languished in single digits.
Out of curiosity, Mahoney proposed to modify Lighthouse’s predictive AI technology—which had originally been developed to identify privilege—so that it could be used to accurately identify responsiveness. Mahoney had helped develop the algorithms that made Lighthouse AI so accurate at identifying privilege, and he had seen first-hand how well the tool worked when deployed in Relativity. He was confident that, if thoughtfully applied for responsiveness, Lighthouse AI could help improve review efficiency and accuracy.
He was right.
“I ran the same training and controls through Lighthouse AI, and precision immediately improved tenfold,” says Mahoney.
For Cleary’s next HSR Second Request, Mahoney got the Lighthouse AI solution approved by the Department of Justice for a responsiveness review within Relativity. To get that approval, Lighthouse had to test and validate new workflow strategies and models. In the process, the Lighthouse tool saved Cleary’s client millions on review and achieved 75 percent recall.
From there, Lighthouse and Cleary continued to innovate with Lighthouse’s predictive AI solutions—achieving even better results on privilege and responsiveness with precision and recall rates well above 80 percent and above 90 percent for PHI, with minimal training.
A Shared Vision for Using AI to Innovate in e-Discovery
Since then, the Cleary + Lighthouse partnership has continued to thrive based on a shared ethos around innovation using AI. Both partners understand that to take advantage of the huge opportunities that AI presents in e-discovery, it is critical to identify the ideal use cases—the ones that have the most practical and beneficial impacts on clients’ existing Relativity workflows and are appropriate for purpose-built AI. The partners also share a commitment to relentlessly honing the accuracy of results with rigorous testing and continuous development. With those principles in mind, both Cleary and Lighthouse have continued to focus on testing and applying Lighthouse AI capabilities against high-volume, routine tasks for clients, effectively filtering out "noise" in information to home in on the nuanced legal needs that require human expertise—with great success.
Over the last three years, Lighthouse and Cleary have continued to expand the use cases and team together to build new workflows for Lighthouse’s groundbreaking predictive AI models to identify PII/PHI, sensitive language, IP source code, and “junk” documents with greater accuracy and speed than manual review. The ability to build upon Relativity’s platform made these achievements possible and more impactful to clients, who depend on Relativity’s advanced capabilities and workflows.
While the impact for clients has been groundbreaking, the results are also a clear measure of how successful the partnership has been for stakeholders at both Lighthouse and Cleary.
“Our success using AI is evidence that Lighthouse’s relationship with Cleary has been a critical driver for innovation. I am incredibly proud to be a part of such an impactful partnership,” says Karl Sobylak, Director of Product Management at Lighthouse.
A Future Using Generative AI
As new capabilities arise with Generative Pre-trained Transformers (GPT), Cleary and Lighthouse have expanded AI solution development and testing work to include generative AI use cases. The partnership’s approach, vision, and methodology for developing and using LLMs within predictive AI models carries over to their work using generative AI. The partners continue to work together to identify the use cases that are most ripe for the application of generative AI and are working hand-in-hand to develop and rigorously test generative AI solutions in real-world scenarios.
“With the capabilities provided by Generative Pre-trained Transformers (GPT), we are expanding to completely new methods of data extraction, linking, summarization, and reasoning,” says Sobylak.
If their past success with predictive AI is any indicator, the e-discovery industry can expect to see continuous innovation on the generative AI-front from this partnership.
Graphics for this article were created by Sarah Vachlon.