Harvey AI builds artificial intelligence tools specifically for the legal industry. It helps lawyers with complex legal research, document review, contract analysis, due diligence, and litigation support. The platform is designed to handle multi-document reasoning across thousands of files, provide accurate legal citations, and automate workflows for high-stakes legal matters. As of 2025, Harvey serves 700+ clients across 63 countries, including the majority of the top 10 U.S. law firms.

Harvey AI was founded in July 2022 by Winston Weinberg and Gabriel Pereyra. Weinberg was a securities and antitrust litigator at O'Melveny & Myers but only practiced law for one year before starting the company. Pereyra is an AI researcher with experience at Google DeepMind, Google Brain, and Meta. The unlikely pairing of a junior lawyer who understood the pain points and an AI expert who could build cutting-edge systems became Harvey's secret weapon.

Harvey deliberately went after the most prestigious, demanding law firms (like Allen & Overy, Latham & Watkins, and Paul Weiss) from day one. The strategy was based on how trust works in professional services: it cascades down from elite institutions, not up from smaller players. By earning the trust of top-tier firms first, Harvey created a credibility signal that made everyone else downstream (smaller firms and corporate clients) more willing to trust them. This contrarian approach became one of their biggest competitive advantages.

Harvey isn't just a "GPT wrapper" with a legal interface. They built domain-specific infrastructure that foundation models alone can't provide: line-by-line citation accuracy (critical in legal work), multi-document reasoning across 100,000+ documents simultaneously, workflow automation for complex processes like international mergers across multiple jurisdictions, and data permissioning systems that handle ethical walls between clients and conflict-of-interest checks. The moat isn't the AI model itself—it's the messy domain expertise and infrastructure that took years of collaboration with actual lawyers to build.

Harvey's valuation went from $5 million in November 2022 to $8 billion by December 2025. The growth was driven by a combination of factors: starting with prestigious clients who validated the product, creating a network effect where law firms began selling Harvey to their corporate clients, and hiring lawyers as product builders (nearly 50% of the team) to ensure the product met professional standards. They also benefited from timing—entering the market just as generative AI was exploding. By August 2025, Harvey had surpassed $100 million in annual recurring revenue.

The article doesn't specify exact pricing, but it mentions Harvey started with seat-based pricing (per-user licensing) and is now moving toward outcome-based pricing as workflows become more sophisticated. The platform is primarily used by large law firms and corporate legal departments. Harvey currently serves 700+ clients globally, though the focus remains on organizations that handle complex, high-stakes legal work where accuracy and domain expertise are critical.

According to Winston Weinberg, no. Harvey's vision is that AI will make the legal profession more attractive, not eliminate jobs. The idea is that AI handles tedious, repetitive work (document review, first-pass research, routine drafting), allowing lawyers to focus on high-value strategic advisory—the intellectually engaging work that attracted them to law in the first place. Early evidence supports this: surveys of law firms using AI haven't shown dramatic reductions in headcount or billable hours. Instead, they're shifting what lawyers work on and how efficiently they deliver results.

The biggest lessons: go after your hardest customers first if you're in a trust-based industry, build domain complexity that foundation models can't replicate (that's your real moat), hyper-personalize everything (especially demos—show your product working on their actual work), hire domain experts as product builders (not just advisors), design for network effects and multiplayer value, and don't wait until you're "qualified" to start. Weinberg had only one year of legal experience when he founded Harvey. He didn't wait for permission or decades of expertise—he saw a gap, found the right co-founder, and moved.