What it actually is, who uses it, and what its real advantages and disadvantages are.
As a student of law, the topics I am currently studying include tort, property, and constitutional law. Harvey AI has yet to be covered, and I'm not sure if it will be. Because of this, I thought it would be good to spend as much time as possible understanding what Harvey AI is, as it seems to be coming up in every other LinkedIn and Instagram post I see in the world of legal AI.
In 2022, Winston Weinberg, a litigation lawyer, and Gabriel Pereyra, a former Google DeepMind research scientist, founded Harvey AI. Weinberg decided to use GPT-3 on a real landlord-tenant case and saw the potential for AI's use in law. The pair cold emailed Sam Altman, which ultimately secured them a seed round from the OpenAI Startup Fund.
Harvey takes AI and wraps it in legal-specific training, source-grounded answers, and enterprise security, built specifically for how large firms usually work. Harvey AI consists of five main tools:
| Tool | What It Does |
|---|---|
| Assistant | Streamlines analysis of contracts, regulatory filings, and discovery documents, and can draft and generate documents |
| Vault | Parks, analyses, and reviews vast amounts of documents at once, used for due diligence at scale |
| Workflow | Automates multi step, recurring legal tasks |
| History | Stores and organises past queries and work |
| Library | Stores firm specific templates, playbooks, and reusable material |
Unbelievably, 50% of the Am Law 100 use Harvey.1 This figure comes directly from Harvey's own company blog and is supported by Sacra, a private company data firm. With over 100,000 lawyers using the platform across 1,000 to 1,500+ organisations in 60+ countries, and this figure changing constantly due to the rapid increase in numbers, Harvey has become one of the most popular legal AI tools to use, if not the most popular.2
A&O Shearman was the first firm to roll it out, in 2023, giving over 3,500 lawyers access across 43 jurisdictions. It was reported that the firm saw over 60% weekly active usage within the first year.3
Harvey AI's valuation reached $11 billion, as of a March 2026 funding round.4
| Advantages | Disadvantages |
|---|---|
| Measured time savings. A&O Shearman reported 2 to 3 hours saved per lawyer per week, a 30% reduction in contract review time, and up to 7 hours saved on complex document analysis tasks. | Price barrier. Reportedly costing into six figures annually, based on third party estimates since Harvey does not publish official pricing, this makes Harvey AI only accessible to big firms and difficult to afford for solo practitioners. This suggests AI is not levelling the playing field, but rather doing the opposite. |
| Every answer is grounded to a specific source, rather than generated freely with no source behind it. | Hallucination risk. It hasn't disappeared, rather it has been reduced. As lawyers rely on Harvey AI more, the risk remains. |
| It is built around legal work, not legal work built around AI. The entire product is built around drafting, due diligence, research, and document review, in essence everything to do with law, rather than general chatbots which are built for day to day use. | Built for larger firms, not smaller ones. Even if a small firm can afford Harvey AI, one detailed review rated it 3 out of 10 for firms that are not enterprise sized.5 |
It is important to know a lot of the sources for these figures came from Harvey's own reporting and firms publicly using it, rather than a neutral third party. So take it with a slight pinch of salt.