Inside Harvey AI: $11B, $300M ARR, 960 Employees, 12 Offices, 13 Trillion Tokens a Month
Sourcery
1 DAYS AGO
Inside Harvey AI: $11B, $300M ARR, 960 Employees, 12 Offices, 13 Trillion Tokens a Month
Inside Harvey AI: $11B, $300M ARR, 960 Employees, 12 Offices, 13 Trillion Tokens a Month
Sourcery
1 DAYS AGO
Winston Weinberg, CEO of Harvey, discusses the company's rapid growth and strategic evolution in the AI legal space. He provides an inside look at Harvey's San Francisco headquarters and shares insights on scaling from $100M to $300M ARR, the shift to cloud agents, and the company's global expansion.
Harvey's growth is driven by a shift to cloud agents, which doubled usage quarter-over-quarter, and a focus on post-training models with synthetic data for legal applications. Token usage surged from 1 trillion to 12-13 trillion monthly, while DAU/MAU rose from 36% to 52%. Weinberg emphasizes constant reinvention, with the company prioritizing hiring and development every six months. He notes that Harvey's primary competition comes from foundation labs like OpenAI and Anthropic, not other legal tech companies, and claims a win rate over 70%. The company sells legal intelligence, positioning it as an intelligence allocation problem, and creates specialized models for legal tasks that perform as well as frontier models at 100 times lower cost. Weinberg predicts all companies will eventually sell tokens and intelligence, competing with large models.
00:00
00:00
Cloud agents doubled usage quarter-over-quarter
00:47
00:47
A tiny gavel demanded faster product latency.
03:17
03:17
Harvey has grown to nearly 1,000 employees and 12 offices.
04:11
04:11
Expansion reacts to large customers and considers legal market size.
05:18
05:18
Only 25 of 200+ lawyers work in traditional roles
06:34
06:34
Loud lunch culture with nearly 1,000 employees
07:42
07:42
Organic collaboration and cross-functional friendships define the culture.
09:05
09:05
This is my favorite room in the office.
10:09
10:09
Eight Airbnbs were easier than finding a real office
12:58
12:58
The speed of building has accelerated
14:20
14:20
ARR grew from $100M to nearly $300M
15:35
15:35
Harvey hasn't used most of its raised capital.
20:09
20:09
If it feels annoying to write, it's not important.
21:10
21:10
Initially it was very difficult to convince lawyers to join Harvey
22:14
22:14
Underperformers are let go, unlike law firms.
23:26
23:26
Buying talented teams, not legacy technology.
27:01
27:01
Faster decision-making at 51% certainty
28:46
28:46
Every company will eventually sell intelligence
30:09
30:09
Vertical products and specialized models optimize cost between frontier and commoditized models.
31:22
31:22
Foundation labs are the main rivals, not legal tech companies
32:07
32:07
ROI must be shown for every token spent
