HARVEY AI

Winston Weinberg spent years as a lawyer at Goldman Sachs. He knew better than anyone that legal work was drowning in repetition — contract review, due diligence, drafting, research — tasks that cost clients hundreds of dollars an hour and made lawyers miserable.

Gabriel Pereyra came from the other direction: he was a researcher at DeepMind, one of the most respected AI labs on the planet, and had spent years thinking about how language models could be applied to specialized, high-stakes domains.

The two met, figured out they were solving the same problem from opposite ends, and co-founded Harvey in 2022. The timing was almost comically perfect.

GPT-4 was about to drop. The legal industry was enormous — roughly $1 trillion globally — and almost completely untouched by real AI automation.

And crucially, law firms were starting to feel pressure from clients who didn't want to pay associate rates for document review that software could do in seconds.

They applied to the OpenAI Startup Fund almost immediately. OpenAI didn't just invest — they gave Harvey early, preferential access to their most powerful models.

That relationship became the rocket fuel. By the end of 2022, Harvey had signed Allen & Overy, a Magic Circle firm with offices across 40+ countries and a client list that reads like a FTSE 100 directory.

The legal world noticed.

Harvey sells AI software to law firms, corporate legal departments, and professional services firms. The pitch is simple: legal work involves an enormous amount of text — contracts, briefs, memos, research, due diligence packs — and large language models are very good at reading, summarizing, drafting, and analyzing text.

Harvey's platform sits on top of those models and is fine-tuned specifically for legal work, with guardrails, citation accuracy, and integrations that matter in a professional context.

Firms pay a subscription, typically at the enterprise level. The pricing scales with seat count and usage.

Harvey doesn't replace lawyers — they're careful about positioning it that way, and not just for PR reasons. The model is genuinely designed to make senior lawyers more productive, not to fire junior ones (even if the downstream effect may eventually look the same).

The revenue model is clean: recurring SaaS fees from law firms that need productivity tools, and enterprise contracts with large corporate legal teams. As they expand into adjacent professional services — think consulting and accounting — the addressable market grows considerably beyond just law.

Harvey's core product is its AI legal platform — a chat and workflow interface that lets lawyers query documents, draft contracts, conduct legal research, and summarize complex filings. It's built on top of frontier language models from OpenAI, fine-tuned on legal corpora and trained to prioritize citation accuracy and professional tone.

The platform includes document review capabilities that can ingest massive due diligence packs and flag risks, inconsistencies, or missing clauses in minutes rather than days. For contract drafting, it can generate first drafts based on precedents and firm-specific playbooks.

For legal research, it surfaces relevant case law with citations — though lawyers are still expected to verify before filing.

They've also built matter management integrations that slot Harvey into existing law firm workflows rather than requiring lawyers to switch platforms. The goal is to be the AI layer underneath how legal work already gets done, not a separate app that competes for attention.

Harvey's key insight was to go upmarket first. Most AI startups go mass-market: build something cheap, acquire users in volume, figure out revenue later.

Harvey did the opposite. They signed Allen & Overy — a firm that does not take meetings with unproven startups — before they had a finished product.

That one logo opened every door that followed.

The OpenAI relationship was the other unfair advantage. Having early access to GPT-4 while competitors were waiting in line meant Harvey's product worked better than anything else in the category from day one.

And OpenAI investing meant a credibility stamp that mattered enormously when going into sales meetings at conservative institutions like law firms.

From there, the strategy was expansion through elite institutions. PwC signed on — 4,000 lawyers across their legal services division.

Sullivan & Cromwell. A&O Shearman after Allen & Overy's merger.

Every big-name client made the next one easier. Law firms follow each other.

If your main competitor is using Harvey, you either adopt it or explain to your managing partners why you're not.

Legal AI has a trust problem that no amount of funding fully solves. Lawyers are professionally and personally liable for the work they sign off on.

When an AI hallucinates a case citation — which early legal AI tools did, publicly and embarrassingly — it doesn't just look bad, it can cost a client their case and a lawyer their career. The stakes are much higher than a chatbot giving you wrong restaurant recommendations.

Harvey has worked hard on citation accuracy and factual grounding, but the fundamental tension remains: LLMs are probabilistic, and the legal system is not. Every hallucination, every subtle error in a contract clause, is a liability event waiting to happen.

Convincing cautious, risk-averse institutions to trust AI output with real legal documents is a slow, painstaking sale.

The competitive landscape is also intensifying fast. Microsoft is embedding Copilot into the tools lawyers already use — Word, Teams, Outlook.

Thomson Reuters, which owns Westlaw (the research tool every lawyer uses), is building AI natively into their platform. LexisNexis similarly.

Harvey is fighting generalist tech giants with enormous distribution advantages on one side, and domain-specific incumbents with decades of trust on the other. Being the best product isn't always enough.

Harvey's investor list reads like a greatest-hits of the AI investment boom. The OpenAI Startup Fund came in during Harvey's seed round — a $5 million raise in late 2022.

That relationship was strategic as much as financial: OpenAI was effectively backing one of its own key deployment partners in a high-value professional sector.

Sequoia Capital led the Series A in 2023, a $21 million round that confirmed Harvey as one of the more serious AI infrastructure plays in the legal category. Sequoia has a long history of backing enterprise software early — Salesforce, ServiceNow, Stripe — and legal AI fit the thesis cleanly.

The Series B in late 2023 was where things got serious: $80 million at a $715 million valuation, led by Kleiner Perkins with participation from Google, Salesforce Ventures, Owl Rock, and a collection of other institutional names. Then in 2024, a Series C brought in another $100 million at a valuation above $1.5 billion, with Coatue Management leading.

The participation of Google and Salesforce Ventures is particularly notable — both have strategic reasons to see legal AI succeed on their platforms.