Harvey, a generative AI startup focused on the legal profession, announced Wednesday that it has raised $21 million in funding.

Sequoia Capital led the Series A funding round, which follows up a $5 million funding round in November from OpenAI LP’s startup fund, the developer of the ChatGPT AI chatbot technology. The OpenAI Startup Fund joined this round again alongside SV Angel and Elad Gil.

Generative AI has become a powerful tool for professional services due to its capability of ingesting large amounts of training data on any subject and generating new content-based natural language queries. These chatbots are capable of engaging in humanlike conversation while being connected to vast amounts of knowledge, which means that they can summarize large documents, act as research assistants, write and edit legal briefs, and more.

Harvey harnesses the power of OpenAI’s GPT-4 large language model AI to provide these capabilities to a growing number of law firms in an invite-only service. Currently, there are more than 15,000 law firms on the waitlist. The law firm of Allen & Overy joined in February and the large accounting firm PricewaterhouseCoopers LLP launched a legal chatbot with Harvey in March.

At the time, Carol Stubbings, PwC’s global tax and legal services leader, called the technology a “game changer” for how the company would operate. It would allow its employees to focus on the aspects of their work that let them work more efficiently, by gaining insights into documents more quickly, doing research faster and skipping rote work so they can spend more time with clients.

“At Harvey, we believe that the future of AI will not be a chatbot that compliments your workflow; it will be the platform your workflow is built on,” Harvey co-founders Gabriel Pereyra and Winston Weinberg said in a blog post. “This is why we partner with entire firms. These close collaborations allow us to find solutions for lawyers that go beyond training data and treatises.”

Pereyra and Weinberg founded Harvey in 2022 based on their personal expertise in the fields of AI and the law. Weinberg worked formerly as an associate at the law firm O’Melveny & Myers and Pereya brought his expertise as a research scientist at Deepmind and as a machine learning engineer at Meta AI.

As part of building AI systems to empower legal firms, Harvey seeks to build them to be scalable, trustworthy and secure. The intent is to design them to be transparent and replicate the best practices of the firms that will use them. This is because legal work is based on reputation and establishing trust in the knowledge work that is done.

As generative AI systems continue to become integrated with more professional systems, concerns have arisen about their reliability and trustworthiness. Although they are extremely powerful, AIs such as GPT-4 can be prone to fits of “hallucinations,” or when an AI states false information with extreme confidence. For example, when Microsoft Corp.’s Bing AI search chatbot came online it was known for unreliability. To help combat that, Harvey said it will design its systems with legal best practices such as tiered reviews and transparency.

“Harvey is the first and best instantiation of a new breed of company: the AI super app. Harvey is in the business of giving people superpowers,” Sequoia partners Pat Grady and Charlie Curnin wrote about the announcement. “First, lawyers. Next, professional services. Eventually, all of knowledge work.”

Image: Harvey