Glossary

Harvey

Looking to learn more about Harvey, or hire top fractional experts in Harvey? Pangea is your resource for cutting-edge technology built to transform your business.
Hire top talent →
Start hiring with Pangea's industry-leading AI matching algorithm today
A Pangea Expert Glossary Entry
Written by John Tambunting
Updated Feb 20, 2026

What is Harvey?

Harvey is a generative AI platform purpose-built for legal work, developed by Counsel AI Corporation and founded in 2022. Unlike general-purpose AI tools adapted for law, Harvey runs on custom large language models trained on legal data and fine-tuned to each enterprise client's own work product. It handles research, drafting, contract analysis, document review, and matter knowledge management across every practice area and jurisdiction. The platform crossed $190M in annual recurring revenue by end of 2025 and serves over 1,000 clients including Allen & Overy, PwC, HSBC, Comcast, and the majority of AmLaw 100 firms. A $160M Series F at an $8B valuation closed in December 2025; by February 2026, Harvey was reportedly raising again at $11B.

Key Takeaways

  • The majority of AmLaw 100 firms now use Harvey, making it near-mandatory knowledge for large-firm legal tech roles.
  • Custom fine-tuned models trained on a firm's own work product produce outputs that match house style — not generic boilerplate.
  • Enterprise pricing starts around $1,000–$1,200 per seat per month with a 20-seat minimum, putting it out of reach for small firms.
  • Harvey scored highest overall in the first major legal AI benchmark study, including 94.8% accuracy on document Q&A tasks.
  • A mid-2025 LexisNexis integration gives Harvey access to the full US legal library — a database moat general AI tools cannot replicate.

What Harvey Does: Legal AI as Infrastructure

Harvey's strength is treating the law firm itself — not just the internet — as the knowledge base. Beyond research and drafting, the platform indexes a firm's past matters, memos, and precedent documents so that any lawyer can query institutional knowledge that previously lived only in senior partners' heads. Think of it as the difference between a paralegal who just graduated law school and one who has personally read every file the firm has ever produced.

Core capabilities include case law research grounded in the LexisNexis library (with citation validation added in mid-2025), contract analysis at scale for due diligence and review cycles, custom drafted documents calibrated to the firm's style, and matter knowledge unification that prevents associates from duplicating research already done on prior engagements. A 2025 benchmark found AI tools completed legal tasks six to eighty times faster than lawyers — Harvey scored highest overall across tested tasks.

Harvey vs CoCounsel vs Lexis+ AI

Thomson Reuters CoCounsel is Harvey's closest enterprise competitor. In the first major industry GenAI benchmark, Harvey scored highest overall while CoCounsel came in second with strong marks (77.2% on document summarization). CoCounsel's advantage is deep Westlaw integration for US case law research; Harvey's advantage is broader workflow coverage and faster model iteration. For a firm already on Westlaw, CoCounsel is the path-of-least-resistance choice. For firms that want fine-tuned models and broader document work, Harvey wins.

LexisNexis Lexis+ AI competes and partners simultaneously — LexisNexis withdrew from the 2025 benchmark but licenses its library to Harvey. As a standalone, Lexis+ AI is research-first with Shepard's validation built in; Harvey is broader, covering the full legal workflow rather than research alone.

Spellbook targets contract-focused mid-market firms at a lower price point and is the right choice for teams that don't have Harvey's minimum seat count or budget.

Pricing: Enterprise-Only, No Public Tiers

Harvey does not publish pricing. The platform runs on annual enterprise contracts with a reported minimum of 20 seats and per-seat costs estimated at $1,000–$1,200 per lawyer per month — meaning minimum annual commitments in the range of $240,000–$288,000 before implementation fees, training, and customization costs for fine-tuned models.

This cost structure is a deliberate market positioning choice. Harvey competes at the top of the market: global law firms and Fortune 500 legal departments where the efficiency gains on a single major transaction can justify the annual contract. Solo practitioners and firms under 50 lawyers have no realistic path to Harvey; for them, Spellbook, Clio's AI features, or general-purpose tools with legal prompting are the practical alternatives. No free tier, no self-serve signup.

Limitations and Production Gotchas

Harvey's self-reported hallucination rate on BigLaw Bench tasks is 0.2% — roughly one fabricated claim in five hundred. That sounds small until you're working on a multi-issue brief with hundreds of claims, where several errors can slip through. Every Harvey output requires attorney review; the efficiency gain is speed, not oversight elimination.

A practical friction point that reviews rarely mention: when documents are uploaded, the platform's input context drops from 100,000 to 4,000 characters. Attorneys working on complex matters have to manually break sophisticated queries into segments to stay within limits — partially defeating the automation benefit. Harvey also struggles with poorly formatted, handwritten, or low-quality scanned documents, and its non-English legal coverage does not match its depth for US and UK law. The ROI case for mid-size firms remains hard to make until associate time is genuinely the binding constraint.

Harvey in the Legal Tech Hiring Market

Harvey's trajectory mirrors what Bloomberg Terminal did for finance: it is transitioning from a competitive differentiator into non-optional infrastructure. The February 2026 fundraise at an $11B valuation — up from $8B just two months earlier — is investors pricing Harvey as a platform bet, not a SaaS feature.

For legal technology professionals on Pangea, this means Harvey expertise is increasingly table-stakes for large-firm legal ops and knowledge management roles rather than a standalone specialization. Companies hire for legal technology directors, knowledge management leads, and legal engineers who can deploy Harvey firmwide, develop prompt standards, and train attorney cohorts. Fractional demand tends toward legal ops consultants who include Harvey implementation and prompt strategy as part of a broader AI-augmented legal workflow practice. Listing Harvey proficiency signals comfort with enterprise legal AI deployment, not just casual tool use.

The Bottom Line

Harvey has moved past the pilot phase and into the infrastructure layer of large-firm legal practice. With the majority of AmLaw 100 firms running it, custom fine-tuned models that match each firm's style, and a LexisNexis integration that general AI tools cannot replicate, Harvey is the dominant platform in enterprise legal AI. For companies hiring through Pangea, Harvey expertise signals a legal technology professional capable of driving AI adoption at scale — not just using a tool, but deploying it across a firm and making it stick.

Harvey Frequently Asked Questions

Is Harvey AI suitable for small law firms or solo practitioners?

No. Harvey is priced and structured for large law firms and corporate legal departments, with reported minimums around 20 seats and costs of $1,000–$1,200 per lawyer per month. Solo practitioners and small firms are better served by Spellbook, Clio's AI features, or general-purpose AI tools with legal prompting.

How does Harvey differ from just using ChatGPT or Claude for legal work?

The core differences are custom fine-tuning and proprietary data access. Harvey trains models on a firm's own work product, producing outputs that match house style and institutional precedent. It also integrates the LexisNexis US legal library for citation-grounded research — a database general AI tools cannot access. General AI models can draft legal text but can't replicate those two layers.

Does Harvey hallucinate or make up case citations?

Harvey's self-reported hallucination rate on BigLaw Bench tasks is approximately 0.2% — about one fabricated claim per 500. The LexisNexis integration added citation validation that reduces research hallucinations. However, every Harvey output still requires attorney review; the tool accelerates work, it does not replace professional judgment or oversight.

What skills should a legal technology professional have alongside Harvey?

Harvey expertise pairs naturally with broader legal operations knowledge: document management systems (iManage, NetDocuments), Microsoft 365 integration, prompt engineering for legal workflows, and change management for attorney adoption. The rarest and most valuable combination is a legal professional who can both evaluate AI output quality and manage firmwide rollout.

How quickly can a legal tech hire get productive with Harvey?

Attorneys with strong legal research instincts can reach basic productivity in a few days — the interface is built for lawyers, not engineers. Mastering prompt engineering for complex document review takes two to four weeks of regular use. Enterprise contracts include onboarding training; a fractional hire with prior AI-assisted legal experience can ramp within a week on standard workflows.
No items found.
No items found.