Harvey CEO: How a 31-year-old Runs an $11B Company
Winston, the 31-year-old CEO of Harvey (an AI legal tech company valued at $11B), discusses his management philosophy, decision-making frameworks, and lessons learned building the company. He covers prioritization systems, hiring for resilience, the future of AI in law, and how founders must balance building a machine versus fixing bottlenecks.
Summary
Winston describes his productivity system: a 200-400 page Google document that tracks motivational reminders, key metrics dashboards, quarterly goals, and a daily ranked task list. The core value of the document is that repeatedly re-ranking tasks forces meta-level thinking about priorities, which he identifies as when he performs best. He argues that prioritization must be completely redone every 3-6 months as a leader.
On decision-making, Winston outlines a framework centered on identifying one-way vs. two-way doors (defaulting to 99.9% being two-way), assessing whether a decision helps or hurts the current P0 priority, and being realistic about who will execute. He uses a 'paragraph test' for meetings — if you can't easily write a paragraph justifying a meeting, you shouldn't take it. He argues people say yes to things that create visible short-term progress (like hiring a CRO) rather than fixing the real underlying problem (like product quality), because external validation from investors feels good immediately.
Harvey's origin story involved Winston (a lawyer) and his co-founder Gabe (an AI researcher from Google Brain and Meta) experimenting with GPT-3 in early 2022. They tested the model on Reddit legal questions and had three lawyers evaluate the outputs — 86 out of 100 responses were deemed sendable without edits, which convinced them to build the company. They cold-emailed Sam Altman and OpenAI's general counsel, pitched on July 4th, and raised their initial funding from OpenAI without approaching other VCs.
The darkest moment came in early 2024 when they attempted to acquire a company 10x their headcount through a leveraged deal structure, signing the term sheet before securing full funding. They fell ~$200M short of the $700M needed, ultimately declined high-risk debt financing, and the deal collapsed. Winston processed the failure in about 24 hours and reframed it as forcing them to build the company properly through hiring, product development, and scaling processes.
On stress and resilience, Winston advocates for 'stress maxing' — deliberately experiencing stressful decisions early at lower stakes to build tolerance for larger ones. He notes that as CEO of an 800-person company, over-reacting to threats causes company-wide thrash because the CEO's stress amplifies through the organization. He identifies that the people who fail at Harvey never did so from making too many mistakes, but from decision paralysis, fear of hiring people better than themselves, and inability to scale.
On the future of law, Winston believes AI will fully commoditize work-product delivery (contract review, diligence, research) but that decision-making and advice will become more valuable over time. He predicts law firms may not shrink because AI enables them to take on far more projects, even with fewer people per project. He describes a regulatory landscape where lawyers must be barred to give legal advice (unauthorized practice of law is a felony in most states), making fully autonomous AI law firms currently impossible except in Arizona and Utah sandbox environments.
Key Insights
- Winston argues that re-ranking his daily task list multiple times per day — rather than setting it once — is the core driver of his best performance, because each re-ranking forces meta-level thinking about what actually matters most.
- Winston contends that founders say yes to the wrong things because external stakeholders (like VCs) reward visible short-term actions (e.g., hiring a CRO) with immediate praise, while the actual fix (improving the product) has a multi-quarter delay in visible results and requires tolerating outside pressure during that window.
- Winston claims that nobody at Harvey has ever been let go for making too many mistakes — every person who failed did so through decision paralysis, fear of hiring people better than themselves, or inability to scale, all rooted in an aversion to failure rather than actual poor performance.
- Winston argues that in professional services, AI will fully commoditize 'work product delivery' (contract review, diligence, research) but that advisory judgment — which relies on personal relationships, contextual experience, and reading people — will actually become more valuable over time because that data is very difficult to distill into models.
- Winston identifies that the biggest mistake he repeatedly makes as a founder is stepping away from product and compensating with more sales activity, describing it as something that helps for a couple of quarters but ultimately fails to scale, because only product scales.
Topics
Full transcript available for MurmurCast members
Sign Up to Access