How Harvey Builds AI Around Domain Expertise

Good morning. Early this year, it became clear to investors that companies with a focus on the legal profession were highly vulnerable to an incursion by frontier AI developers. Harvey, an AI platform for legal work, has continued to show resilience since then, and it’s worth understanding how that’s so.

The four-year-old startup has fused deep domain expertise with its constantly evolving AI platform. I’m sure it’s difficult to maintain a constant mastery on both fronts. But companies that can make it work are in better position to build a good future.

Harvey yesterday unveiled a major update, Harvey II, that incorporates memory into the platform. While leading generalist AI platforms have been incorporating memory in a big way since 2024, this is newer territory for AI in specialized professional markets.

“Harvey II is all about moving from AI that responds to individual prompts towards Harvey that increasingly understands how you work in the context of the work that you’re doing,” Harvey Chief Product Officer Anique Drumright told me. “No lawyer starts from zero context or zero understanding of the client or zero preferences. So how do we build a product that more reflects how lawyers actually work?” (Read the full column here.)

Harvey has been building memory since January, but engineering is only one dimension of the effort. “We built Harvey II closely alongside our customers,” co-founder and CEO Winston Weinberg said on LinkedIn. “I’m grateful to every firm and legal team that helped shape the product.”

Weinberg didn’t set out to be a technology entrepreneur. He studied liberal arts at Kenyon College [full disclosure: I did too, once upon a time] and law at the University of Southern California Gould School of Law. Weinberg began his career as an associate at O’Melveny & Myers, specializing in securities and antitrust litigation.

Weinberg told me earlier this year that his experience as a junior lawyer was invaluable when he shifted course and launched Harvey. A senior role in the legal profession, he said, might have blinded him to the everyday friction of legal workflows. Being a junior associate allowed him to understand where the industry’s data lives, what practitioners actually care about and where inefficiencies exist.

Such proximity to the market is the foundation for companies that distinguish themselves from increasingly powerful general AI. The experience of frustration distinguishes humans from machines, at least for now. It isn’t fun, but in the right hands, it’s an invaluable part of the creative process.

Epic token consumption. Between January and June of this year, monthly token consumption on the Harvey platform grew 14-fold, to 14.5 trillion tokens a month from 1 trillion, co-founder and President Gabriel Pereyra told me. That level of demand goes a long way toward explaining why so much capital continues to pour into the construction of AI infrastructure, which is necessary to produce those tokens.

One more round. In March, Harvey raised $200 million at a valuation of $11 billion, in a round led by Sequoia Capital and GIC. The company, which has annual recurring revenue of about $350 million, is now looking to raise $500 million at a valuation of $15.5 billion, according to a person familiar with the matter.

Does your company shape AI around domain expertise? Send your feedback to me at steven.rosenbush@wsj.com (if you’re reading this in your inbox, you can just hit reply).

The WSJ Tech Council brings together CIOs, CTOs and CISOs advancing innovation and shaping the future. Join this trusted community where tech executives connect with peers to explore emerging trends and gain the perspective they need to stay ahead of disruption.

Request an Invitation

Follow Isabelle Bousquette on LinkedIn, Instagram, X, and TikTok for more behind the scenes on her tech and AI coverage, and lately, her contributions to the WSJ Leadership Institute’s new Executive Resilience series, where she’s profiling America’s top execs about their fitness and wellness habits.

Follow Belle Lin on LinkedIn and X for her latest reporting on enterprise technology and AI.

Steven Rosenbush is chief of the enterprise technology bureau at the WSJ Leadership Institute. He also has a column. You can follow him on LinkedIn.

Tom Loftus is the editor of The Morning Download. He suggests following Isabelle, Belle and Steve on their various social channels. But if you insist, here’s his LinkedIn.