At A&O Shearman, the first global law firm to deploy Harvey at enterprise scale back in December 2022, lawyers report cutting contract review time by roughly 30% and saving two to three hours a week on routine tasks [Medium]. That single reference customer, a magic-circle-pedigree firm, is the wedge that has carried Harvey from a roommate project in Los Angeles to one of the most heavily funded application-layer AI companies in the market.
Harvey, founded in 2022 by Winston Weinberg and Gabriel Pereyra, sells a domain-specific AI assistant to law firms and corporate legal teams. The product handles contract analysis, due diligence, compliance work, and litigation support, routing requests across multiple large language models and synthesizing the outputs [Harvey Help]. The ideal customer profile is sharp and well-defined: Am Law 100 and Am Law 200 firms, plus the in-house legal departments of large multinationals that already buy seven-figure software contracts and have a partner or general counsel willing to sponsor a multi-year rollout. By 2024, Contrary Research counted 28% of the Am Law 100 among Harvey's clients, with reach into 235 customers across 42 countries.
The bet
The wedge is straightforward: take the highest-margin knowledge work inside a law firm, the kind that historically gets billed to clients in six-minute increments, and let an associate or paralegal complete it faster with AI assistance. Harvey runs on Microsoft Azure infrastructure, which matters for a buyer base that lives and dies by SOC 2, ISO 27001, and client confidentiality clauses [Harvey.ai]. The company has also begun publishing thinking on how billing models inside firms are shifting toward fixed, subscription, and value-based pricing, a tell about where Harvey wants its own contracts to sit over time [Harvey.ai].
The procurement cycle for this ICP is not fast. Selling into an Am Law 100 firm typically means a pilot with a single practice group, a security review run by the firm's CISO, partner buy-in across the management committee, and a multi-quarter rollout. The budget owner is usually a combination of the COO, the CIO, and the head of innovation or knowledge management, with the managing partner signing off on anything that touches client data. Renewals are the real test: in legal, a tool either gets embedded into matter workflows within the first year or it quietly disappears from the stack at renewal time.
Why it could be big
Harvey's funding history reads like a conviction trade by the AI investor class. Sequoia led a $300 million Series D in February 2025 at a $3 billion valuation, with the company telling Fortune that ARR had passed $50 million and was on track for $100 million within roughly eight months [Fortune, Feb 2025]. By August 2025, AllAboutAI reported Harvey had hit that $100 million ARR mark on schedule. Andreessen Horowitz then led a $160 million Series F in December 2025 at an $8 billion valuation [TechCrunch, Dec 2025]. By February 2026, Forbes reported Harvey was in talks to raise another $200 million at an $11 billion valuation, a round that SiliconANGLE later reported closed in March 2026 with GIC and Sequoia leading [Forbes, Feb 2026; SiliconANGLE, Mar 2026]. Total capital raised has reached approximately $1 billion [TheAIInsider, Mar 2026].
Feb 2025 Series D valuation | 3 | $B
Dec 2025 Series F valuation | 8 | $B
Feb 2026 reported round valuation | 11 | $B
ARR Feb 2025 | 0.05 | $B
ARR Aug 2025 | 0.1 | $B
The cap table is a who's-who: OpenAI Startup Fund (notable given Harvey's reliance on frontier models), Sequoia, Kleiner Perkins, GIC, Andreessen Horowitz, Coatue, Conviction Partners, Elad Gil, GV, DST Global, SV Angel, and REV. That mix of crossover funds and sovereign capital (GIC) suggests investors are underwriting Harvey as a durable software business, not a flip.
The team and traction
Weinberg, the CEO and co-founder, was a securities and antitrust litigator at O'Melveny & Myers before leaving after one year to start the company [Wikipedia; Forbes]. Pereyra, president and co-founder, spent roughly a decade in AI research at Google Brain, DeepMind, and Meta [Digidai, Nov 2025]. The two were roommates in Los Angeles when they began building. Alan Ghelberg joined as CFO in February 2024, a hire that typically signals a company preparing for late-stage rounds and the financial discipline that institutional buyers expect [websets.exa.ai]. Open roles include a Staff Product Manager and an Engineering Manager for Product Engineering, both pointing at deeper investment in the enterprise platform layer.
The honest counterfactual
What bears will point to: Harvey is an application-layer company building on top of frontier models it does not own, and the OpenAI Startup Fund presence on the cap table is a reminder of how dependent the product can be on outside model providers. If a Microsoft or an OpenAI decides legal verticalization is a first-party opportunity, or if a competitor like Spellbook (focused on contract drafting in Word) or Ironclad (a contract lifecycle management incumbent that is steadily adding AI) closes the gap, Harvey's pricing power gets pressured. The bull answer is in the deployment data: A&O Shearman's measurable productivity gains, 28% Am Law 100 penetration by 2024, and a multi-model architecture that lets Harvey swap underlying LLMs without re-papering the customer relationship [Contrary Research; Harvey Help]. Switching costs in legal software, once a firm has trained associates and integrated a tool into matter workflows, are notoriously high.
The realistic competitive set is narrower than the broader "legal AI" category suggests. Spellbook competes most directly on contract drafting inside Microsoft Word for small and mid-market firms, a different ICP. Ironclad competes from the CLM side, owning the contract repository and workflow but historically weaker on generative drafting and legal research. Thomson Reuters (with CoCounsel, acquired from Casetext) and LexisNexis (with Lexis+ AI) are the incumbents Harvey actually has to displace inside Am Law firms, and both bring decades of content licensing and existing enterprise contracts to the fight.
What to watch
The next twelve months will turn on three things: whether the reported $11 billion round closes on the terms SiliconANGLE described, whether Harvey can show net revenue retention above 120% on its earliest law firm cohorts (the metric that would justify the multiple), and whether the company expands beyond law into adjacent professional services like accounting and consulting, which Harvey's own job listings hint at when they reference "new customer segments and verticals" [Harvey.ai]. A first major customer logo outside law would be the strongest signal yet that the platform thesis is real.
ICP: Am Law 100 and Am Law 200 firms, plus Fortune 1000 in-house legal departments. Realistic competitive set: Thomson Reuters CoCounsel, LexisNexis Lexis+ AI, Ironclad on the CLM flank, and Spellbook in the SMB segment.