Harvey
Domain-specific foundation-model agents for elite law firms and enterprises — legal research, drafting, due diligence, litigation support and compliance. Embedded legal-engineering teams deploy and customise per firm.
The Business
Harvey builds AI agents purpose-built for elite law firms and corporate legal teams. Each deployment is customised by embedded legal-engineering teams who work alongside the customer firm, and the company has invested heavily in proprietary legal knowledge bases as a vertical advantage. Headline use cases span legal research, document drafting, contract due diligence, litigation support and regulatory compliance.
Customers and Distribution
Harvey discloses 100,000+ lawyers across 1,300 organisations as of March 2026. Customers skew towards AmLaw 100 and tier-1 international firms (A&O Shearman, DLA Piper, McCann FitzGerald), plus a growing in-house enterprise general-counsel segment (NBCUniversal, HSBC) and the Big Four (PwC). Distribution sits inside Microsoft Word and Azure, with channel partnerships through DocuSign, LexisNexis and DeepJudge. Go-to-market relies on embedded legal-engineering teams that work onsite or near-site with each firm; lower-touch self-serve plays a secondary role.
Model Strategy
Harvey started as an OpenAI-exclusive partnership and has since diversified to routing across Anthropic Claude (Sonnet/Opus 4), OpenAI (GPT-5 / o3 / 4.1 / 4o) and Google Gemini 2.5 Pro, with multi-cloud deployment on Azure, AWS Bedrock and Google Vertex. Custom-trained legal models built on top of these foundation layers are positioned as the moat — vertical fine-tuning, privileged co-development with marquee customer firms, and the company’s own benchmarks (BigLaw Bench, LAB).
At A Glance
The Numbers
Annualised revenue
Paid seats
Headcount (FTE)
Funding to date
Leadership Team
Harvey has also recruited extensively from US Big Law — named recent hires include partners and senior associates from White & Case, Latham & Watkins, Skadden, Gunderson Dettmer, Katten Muchin Rosenman, and Paul Weiss. Embedded legal-engineering teams sit alongside customer firms; this is the company’s distinctive operating model. CFO, CRO and CTO roles have not been publicly disclosed as separate appointments.
IM Framework Scoring
IM’s structured assessment of Harvey’s competitive position. The summary below is the headline; expand “Show the full analyst-grade analysis” near the bottom for the per-dimension reasoning and evidence. Methodology →
Funding History
| Date | Round | Raised | Post-money | Lead investor(s) |
|---|---|---|---|---|
| Mar 2026 | Series E second extension | $200M | $11.0B | GIC & Sequoia (co-led) |
| Dec 2025 | Series E extension | $160M | $8.0B | Andreessen Horowitz |
| Jun 2025 | Series E | $300M | $5.0B | Kleiner Perkins, Coatue (co-led) |
| Dec 2024 | Series D | $300M | $3.0B | Sequoia |
| Late 2024 | Series C | $100M | $1.5B | GV (Google Ventures) |
| 2024 | Series B | $80M | ~$700M | Kleiner Perkins, Sequoia |
| 2023 | Series A | ~$21M | ~$150M | Sequoia (OpenAI Startup Fund early) |
| 2022 | Seed | ~$5M | — | OpenAI Startup Fund, SV Angel |
Cumulative ~$1.0B. The Jun 2025 round was Series E at $300M / $5.0B post-money co-led by Kleiner Perkins and Coatue (not Sequoia) per Harvey’s Series E announcement blog post. The Dec 2025 round was a $160M Series E extension led by Andreessen Horowitz (not Sequoia, not $300M, not $8B headline raise) per the Harvey blog announcement. The Mar 2026 second extension at $200M / $11.0B post-money was co-led by Sequoia and GIC. Round-by-round figures from Harvey’s own blog posts, CNBC, Bloomberg, TechCrunch and PitchBook-equivalent named-press coverage.
Competitive Landscape
Harvey’s competitive set sits in three concentric rings: incumbent legal-research platforms with new AI layers (Thomson Reuters / LexisNexis), pure-play legal AI startups (Spellbook, EvenUp, Robin AI), and foundation-model providers offering general-purpose reasoning — notably OpenAI’s Deep Research, which Harvey’s own founders have publicly called an indirect competitor.
| Competitor | Positioning | Distribution edge | Threat profile |
|---|---|---|---|
| CoCounsel (Thomson Reuters) |
Research-led; backed by Thomson Reuters’ Westlaw + Practical Law content. Absorbed Casetext (acquired Aug 2023 for $650M); standalone Casetext retired April 2025. | Embedded in Westlaw; in the workflow of every AmLaw 200 firm. | High — existing customer relationships + proprietary content moat. |
| Lexis+ AI (LexisNexis / RELX) |
Citation-verified research grounded in LexisNexis content with real-time Shepard’s validation. | Same channel logic as CoCounsel — lives inside existing Lexis subscriptions. | High — same channel-control argument. |
| Spellbook | Word-integrated contract drafting / review for commercial lawyers; in-house and law firm teams. | Microsoft Word native; 4,000+ team installations. | Medium — narrower (contracts) but stickier in that lane. |
| OpenAI Deep Research | General-purpose research agent — not legal-specialised, but a credible research tool that many attorneys reportedly use unofficially. | Available to anyone with a $200/mo ChatGPT subscription; no IT procurement required. | High and asymmetric — OpenAI is Harvey’s foundation-model supplier and a potential competitor. |
| Eve Legal / EvenUp / Robin AI / NexLaw | Specialised plays in narrower lanes (litigation prep, demand letters, contract negotiation). | Smaller and more vertical; less direct overlap with Harvey’s AmLaw 100 / enterprise GC focus. | Low — medium — flanking risk if any one wins a vertical. |
Pricing benchmark: CoCounsel Core starts at $225 / user / month; Harvey is reported at $1,000+ / user / month, with one widely-cited figure of ~$40,000 / year for a 10-seat deployment. Spellbook is materially cheaper. Harvey is positioned at the premium end and competes on capability + embedded-team delivery rather than price.
Potential Risks
The case for Harvey at $11B rests on customer-data-driven legal models, an embedded delivery model, and early dominance of the AmLaw 100. The case against splits into five risks of differing magnitude.
Foundation-model dependency
Harvey started OpenAI-exclusive and has since diversified routing across Anthropic (Claude Sonnet / Opus 4) and Google (Gemini 2.5 Pro); Mistral is not in the live lineup per Harvey’s help centre. That partially de-risks the supplier-concentration problem but does not fully resolve it — Harvey’s custom-trained legal models still rely on the foundation layer for capability scaling. The asymmetric risk is OpenAI itself: Harvey’s founders have publicly acknowledged OpenAI Deep Research as an indirect competitor.
Incumbent distribution
Thomson Reuters (Westlaw / CoCounsel / Casetext) and LexisNexis (Lexis+ AI) are inside every Big Law firm’s procurement budget and workflow. Harvey wins on capability today; the incumbents win on distribution and proprietary content if the capability gap narrows. The Casetext acquisition price ($650M for an arguably comparable revenue profile in 2023) sets a cautionary comparable on what legal-AI exits can look like.
Hallucination and reliability ceiling
A 2024 Stanford study put hallucination rates for leading legal AI tools (Harvey included) above 1-in-6 queries. Harvey reports that its 2024 Assistant version reduced hallucinations by 60%; independent re-verification is sparse. Big Law’s risk appetite for citing fabricated authority is essentially zero, so the reliability ceiling is a hard constraint on the practice areas Harvey can credibly serve.
Valuation-to-revenue gap
$11B valuation on ~$190M ARR is a ~58x multiple. The growth rate (~90% YoY based on the IM tracker series) justifies a premium, but the gap implies the market is paying for a credible $5B+ ARR future. Compression risk is real if growth slows, if Thomson Reuters / LexisNexis materially close the capability gap, or if a foundation-model provider builds legal-specialised reasoning directly into its base product.
Embedded-team operating leverage
Harvey’s embedded legal-engineering model is a distinctive moat but a cost structure that scales sub-linearly. Each new AmLaw 100 firm needs onsite or near-site delivery. Margin expansion depends either on shifting to self-serve once the firm is trained, or on dramatically reducing per-firm delivery cost via tooling. The headcount trajectory (200 to ~650 in 16 months) suggests Harvey is still scaling delivery capacity rather than optimising it.
Recent IM Coverage
- Legal Sector Context Briefing May 2026.
Show recent press coverage of Harvey
- Dec 2025 — Andreessen Horowitz leads $160M investment in Harvey — Series E extension at $8B post-money.
- May 2026 — Docusign × Harvey partnership: Harvey’s legal reasoning platform integrated with Docusign Intelligent Agreement Management.
- May 2026 — Command Center launch + DeepJudge partnership.
- May 2026 — HSBC strategic AI partnership: Harvey embedded in HSBC Global Legal.
- Apr 2026 — CEO profile: Winston Weinberg on building Harvey.
- Mar 2026 — $200M Series E extension at $11B valuation.
- Feb 2026 — Gabriel Macht (Harvey Specter from Suits) brand partnership: launch of Harvey’s Instagram presence.
- Jan 2026 — LawNext podcast: Pereyra & Weinberg on building AI infrastructure for law.
- Nov 2025 — “Inside Harvey: How a first-year legal associate built one of Silicon Valley’s hottest startups”.
Curated feed of named-source coverage from approved publications — not a press-release aggregator. We exclude PR-wire reposts of the same release and unsourced “industry round-up” pieces.
Show the source register for the figures on this page
IM operates a primary-source-where-possible discipline. The figures above come from:
- Revenue: The Information (Aug 2025 $100M ARR) and Harvey’s own Mar 2026 round blog implying continued trajectory. IM triangulation to $190M for Jan 2026.
- Paid seats: Harvey’s $11B round announcement — “100,000+ lawyers across 1,300 organisations.”
- Headcount: Harvey LinkedIn company page, ~600–650 range post-Series E closes.
- Funding to date: Cumulative through Harvey’s Mar 2026 announcement; corroborated by CNBC and Bloomberg coverage.
Methodology & Disclaimer
For metric definitions, source-tier hierarchy, and decline-to-publish rules, see the tracker methodology. Confidence dots (• green / • amber / • red) follow the same convention as the AI Tracker.
Spotted a figure you believe is wrong? Send corrections to info@informationmatters.net.
Information Matters Framework scores are the considered opinion of the IM team — human and AI — applied to publicly-available evidence under a disclosed methodology. They are not statements of fact about the companies scored and they are not investment advice.
