The meeting that changed US legal history happened over frozen yogurt in Palo Alto. Daniel Lewis, then finishing his JD at Stanford Law School, sat down with Jonathan Zittrain, director of Harvard's law library, and pitched something audacious: digitize every official US court decision ever printed - all 40 million pages, spanning 360 years - and give it away for free. Zittrain said yes. The Caselaw Access Project was born. And Lewis walked out with a partnership that would set the tone for everything he'd build afterward.
That was 2015. The startup Lewis had co-founded in 2012 - Ravel Law, built with classmate Nik Reed while they were still in law school - was already doing what most lawyers considered impossible: using machine learning to map judicial behavior, predict outcomes, and help lawyers build sharper arguments. LexisNexis and Westlaw had owned legal research for decades. Ravel walked in with algorithms. The incumbents weren't amused. Lewis didn't care.
From Stanford to LexisNexis - Then Out Again
LexisNexis acquired Ravel Law in 2017. Lewis stayed - not as a reluctant holdover, but as VP and General Manager of Practical Guidance and Analytical products, overseeing tools that 100,000+ attorneys across the US, UK, and Canada used every day. Five years at one of the most established names in legal information. He saw what worked, what couldn't be fixed from the inside, and what the next generation of legal tech needed to look like.
There are some really tedious parts of lawyering that few people happily sign up for, and I want to change that.
- Daniel Lewis, Global CEO, LegalOn TechnologiesIn December 2022, LegalOn Technologies - a Tokyo-founded legal AI company with 3,000+ customers and $101 million in funding - announced its US expansion. They hired Lewis to lead it. He didn't inherit a working playbook. He built one: restructuring go-to-market, building the US team, and repositioning a Japanese-market leader for the largest and most demanding legal market in the world.
The Attorney-Playbook Bet
When Lewis talks about LegalOn's technology, he uses a word that most AI companies avoid: consistency. Raw language models are powerful, but they hallucinate. They produce outputs that look authoritative and are legally wrong. LegalOn's answer is to ground its AI in attorney-built playbooks - over 50 of them, developed by practicing lawyers - that define what good contract review actually looks like across deal types, risk categories, and jurisdictions.
The platform lives inside Microsoft Word, where most legal work already happens. In-house teams can review contracts, flag risks, redline clauses, manage matters, and compare documents against custom playbooks - without switching tools or re-training staff. Lewis doesn't describe this as AI replacing lawyers. He describes it as AI handling the part of the job that nobody got into law to do: the high-volume, time-pressured, often urgent review work that lands on an in-house team's desk with a 24-hour turnaround demand.
Our approach ensures contract reviews are aligned with real legal standards, making the output more accurate, consistent, and practical for legal teams.
- Daniel LewisGoldman Sachs, SoftBank, and OpenAI Walk Into a Term Sheet
In July 2025, LegalOn closed a $50 million Series E led by Goldman Sachs' growth equity fund, with participation from World Innovation Lab (WiL), Mori Hamada and Matsumoto law firm, Mizuho Bank, and Shoko Chukin Bank. The round brought total funding to $200 million, making LegalOn the most well-funded AI platform specifically built for in-house contract review.
Alongside the funding, LegalOn announced a strategic partnership with OpenAI - not a co-marketing arrangement, but an engineering collaboration. Lewis explained the logic plainly: "The partnership gives us early access to their latest models, and it positions our engineers to work side-by-side with engineers from OpenAI - building cutting-edge agents using great technology, grounded in our proprietary legal content and expertise."
The phrase "grounded in proprietary legal content" is doing significant work there. LegalOn's thesis is that frontier AI models plus attorney expertise produces something qualitatively different from either alone. Hallucination rates drop. Playbook alignment goes up. The output is something a general counsel can actually sign off on - not just a draft that still needs a full human review to be usable.
Six Million Cases, One Frozen Yogurt Conversation
The Caselaw Access Project deserves more recognition than it usually gets in coverage of Lewis's career. Between 2013 and 2018, Harvard Law School's Library Innovation Lab - working with Ravel Law's technology and funding - digitized the entire corpus of official US court decisions: 6.7 million cases, over 360 years of legal history, 40+ million pages. All of it made freely available to researchers, lawyers, and the public.
That project required convincing Harvard to partner with a startup that was less than two years old. It required building digitization infrastructure for a collection that predated the internet by centuries. And it started because Lewis showed up to a meeting in Palo Alto, made his case, and the answer was yes. The pattern - spotting an absurd constraint on legal information, building the thing to remove it - runs straight through from the Caselaw Access Project to LegalOn's contract review platform.
Who He Is When Not Running a 600-Person Company
Lewis is a father of two young daughters and an active mountain biker who rides the trails in the Marin Hills north of San Francisco. He runs. He hikes. He cooks. He describes himself as someone who has genuine difficulty staying sedentary - which may explain both the energy of his career trajectory and his patience, or lack thereof, with legal technology that moves slowly.
He's also a communicator who has invested in how his company talks about AI internally. When employee concerns about automation surfaced, Lewis didn't default to reassurance. He built workshops. He explains AI's role as expansion rather than replacement. Whether that's strategy or genuine belief is hard to say from the outside - but the distinction matters less than the fact that LegalOn's team is 600 people and growing.
The Road From Here
Lewis is building in a market that's moving fast. Every major AI company has some version of a legal AI offering. Microsoft Copilot touches legal work. Harvey has significant backing. Ironclad operates in adjacent contract management territory. The differentiation Lewis leans into is depth: attorney-curated playbooks, not just language model inference; built for in-house teams, not firms; works out-of-the-box on day one, not after a three-month implementation.
LegalOn's 25% market share among Japanese publicly-listed companies gives Lewis something most US legal AI founders don't have: proof of what happens when an enterprise legal team fully adopts AI-native contract review at scale. That data - and the engineering built around it - is the reason Goldman Sachs led the round. And it's the reason Lewis is, by most measures, exactly mid-stride.