In a rare public forum appearance, Harvey co-founders\u00a0Winston Weinberg, its CEO, and\u00a0Gabriel Pereyra, its president, spent over two hours answering questions from the legal tech community in a Reddit AMA earlier today, addressing everything from their $8 billion valuation to how they compete with legal research giants and what the future holds for AI in [\u2026]<\/p>\n
In a rare public forum appearance, Harvey co-founders\u00a0Winston Weinberg<\/a>, its CEO, and\u00a0Gabriel Pereyra<\/a>, its president, spent over two hours answering questions from the legal tech community in a Reddit AMA<\/a> earlier today, addressing everything from their $8 billion valuation to how they compete with legal research giants and what the future holds for AI in law.<\/p>\n The AMA \u2014 short for \u201cask me anything\u201d \u2014 took place on the subreddit legaltech<\/a>, which is devoted to discussions of legal technology and innovation. It was the third vendor AMA the subreddit has hosted.<\/p>\n When pressed on his company\u2019s eye-popping $8 billion valuation, Weinberg was notably modest: \u201cWe need to earn that valuation everyday.\u201d<\/p>\n He broke down the investment analysis into two components: math and momentum. On the math side, there are approximately 10 million global legal professionals, and Harvey currently serves only single-digit percentage points of them.<\/p>\n On momentum, the company has seen an 81% increase in daily active users as a percentage of monthly active users since launching in 2023, with power users showing engagement levels comparable to Slack or email.<\/p>\n But Weinberg\u2019s longer-term vision reveals the real bet investors are making. Today\u2019s $30 billion legal tech market represents just 3% of the $1 trillion total legal services market.<\/p>\n \u201cThe simplest answer here is that the tech penetration into the legal market is going to change massively,\u201d Weinberg said, \u201cand if we build a great product we hopefully capture some of that very large upside.\u201d<\/p>\n Importantly, Weinberg emphasized this is about task automation, not job automation.<\/p>\n He also said he sees plenty of room other legal AI startups.<\/p>\n \u201cI don\u2019t think a single player is going to capture all of the pretty enormous amount of value that will be created in the next 10 years in this space,\u201d he said.<\/p>\n Another question addressed Harvey\u2019s competitive position now that it has partnered with LexisNexis. Historically, Harvey sat atop general AI models and firm data, while competitors such as Westlaw CoCounsel and Lexis+ AI were deeply integrated with proprietary legal research databases.<\/p>\n Weinberg called Lexis \u201can insanely trusted data source\u201d and said customers using both platforms see tremendous value from the integration. The partnership has enabled Harvey to build specialized workflows like drafting motions for summary judgment and motions to dismiss that combine Lexis data with Harvey\u2019s drafting capabilities.<\/p>\n \u201cI think if we can nail these two it\u2019ll be a really cool step change in data+drafting,\u201d Weinberg said. \u201cHopefully we\u2019ll have some more in the near future!\u201d<\/p>\n When asked specifically how Harvey\u2019s research capabilities compare to Lexis Prot\u00e9g\u00e9 on complex workflows, the response focused on Harvey driving the research strategy itself \u2013 determining which searches to run, analyzing initial results, and deciding what to look for next \u2013 rather than simply relying on Lexis to do the heavy lifting.<\/p>\n About 20% of Harvey\u2019s team are lawyers, and this is intentional, Pereyra said. \u201cOur goal is to build with and for the industry we serve.\u201d Lawyers and engineers work side-by-side on everything from workflow design to evaluating new model performance.<\/p>\n When asked the key to Harvey\u2019s client growth and adoption, Weinberg answered that the three main drivers were:<\/p>\n Interestingly, when asked if there was one product decision they would \u201cun-do,\u201d Pereyra said he would revisit their vault product design.<\/p>\n They initially tried to support two very different workflows in the same interface: asking simple questions over hundreds of documents and performing complex review over thousands of contracts. This led to significant scaling challenges that required substantial engineering work to resolve.<\/p>\n Sharing details about Harvey\u2019s early sales approach, Weinberg said the most effective tactic was demonstrating Harvey on specific matters lawyers were actively working on. For litigators, for example, he would pull their latest brief from PACER and have Harvey draft counter-arguments.<\/p>\n \u201cThis worked super well with litigators because they like lasered onto the screen reading every word,\u201d he recalled.<\/p>\n It was risky \u2013 if the AI hallucinated, lawyers would immediately call it out \u2013 but when it worked, they were highly engaged.<\/p>\n For M&A lawyers, he used the same approach with EDGAR filings, though Weinberg noted it worked \u201cnot as well as the briefs with litigators.\u201d<\/p>\n The origin story itself is compelling: Weinberg and Pereyra were roommates and best friends in San Diego, never expecting to start a company together. When Pereyra was brainstorming AI assistant ideas with friends from Google Brain, Weinberg suggested, \u201cWhy don\u2019t you build a legal assistant?\u201d<\/p>\n They cold-emailed Sam Altman and Jason Kwon (then OpenAI\u2019s general counsel) with the pitch and raised funding before landing any customers.<\/p>\n Participants in the AMA did not shy away from difficult topics. When asked about hallucinations, Pereyra acknowledged that fine-tuning is just one of many techniques they use, including RAG, inline citations, agentic methods and other applied research approaches.<\/p>\n However, he pushed back on the notion that AI needs to be perfect. \u201cOne common misconception in legal is that an AI system needs to be 0 hallucinations to be useful. Our experience has been that these systems are incredibly valuable and the rate of hallucinations will continue to decrease.\u201d<\/p>\n When confronted with claims from an alleged former Harvey employee about low user engagement, Weinberg was direct: \u201cIt\u2019s unclear to us that that individual actually worked at Harvey at all, but our DAU\/MAU and renewal rates would suggest that person has inaccurate and\/or outdated information.\u201d<\/p>\n Pereyra added that some early customers now see 80-90% monthly active users with strong weekly and daily engagement.<\/p>\n To the question of why law firm fees are not going down despite efficiency gains from AI, Pereyra said that making individual lawyers more productive does not cleanly map onto making entire matters cheaper, because legal pricing is not about the sum of hours worked but about delivering complete outcomes.<\/p>\n However, he noted that firms and clients are beginning to restructure matters in ways that can be more profitable for firms while providing clearer value for clients. \u201cThat\u2019s ultimately our goal.\u201d<\/p>\n However, he believes that, in the long run, both firms and clients stand to benefit.<\/p>\n \u201cHarvey is already enabling discussions where firms and clients restructure matters in ways that are more profitable for the law firm while also providing clearer value and predictability for the client,\u201d Pererya said. \u201cThat\u2019s ultimately our goal: to help both sides rethink how legal work is delivered and priced so everyone wins from increased productivity.\u201d<\/p>\n Over the course of the AMA, Weinberg and Pereyra revealed some of their plans for the coming year. Among them:<\/p>\n Perhaps the most pointed question came from a tech-savvy attorney who asked what Harvey can do that Claude Opus 4 cannot. Pereyra outlined what he sees as six key differentiators:<\/p>\n The AMA offered unique insights into a company that has matured significantly since its 2022 launch \u2013 but that still faces skepticism about its valuation, competitive moat and real-world impact.<\/p>\n However, it was refreshing to see the co-founders\u2019 willingness to engage with and be transparent around some tough questions, many from actual users.<\/p>\nThe Lexis Partnership and Competing on Research<\/strong><\/h3>\n
Product Development Philosophy<\/strong><\/h3>\n
\n
The Early Days: Building Trust<\/strong><\/h3>\n
Addressing the Tough Questions<\/strong><\/h3>\n
What\u2019s Ahead<\/strong><\/h3>\n
\n
The Differentiation Question<\/strong><\/h3>\n
\n
Bottom Line<\/strong><\/h3>\n