Eve works the whole case with you.

Here is a fact about the legal industry that is either obvious or slightly weird depending on how you look at it: there are two sides to a lawsuit, and they do not have the same amount of money. When an injured person sues a hospital, or an insurer, or a trucking company, the defendant usually shows up with a large firm, a deep bench of associates, and the patience to run out the clock. The plaintiff shows up with a smaller firm working on contingency, which means the firm only gets paid if it wins, which means the firm is quietly, structurally, always a little short on time.

Eve, a San Francisco company founded around 2020 by Jay Madheswaran, Matt Noe, and David Zeng, decided this asymmetry was the whole business. Most legal AI startups sell to whoever has the biggest software budget, which tends to be the big defense-side firms. Eve went the other way and built software exclusively for plaintiff firms - the personal-injury and employment-law shops - on the theory that AI is most valuable to the side that is chronically outgunned.

The pitch is not "make lawyers 10% faster." It is "let the small firm outmaneuver the well-resourced defense."

This turns out to be a good pitch, at least to investors. Eve raised a $14 million seed in 2023, a $47 million Series A in January 2025, and then - just eight months later - a $103 million Series B led by Spark Capital, with Andreessen Horowitz, Lightspeed, and Menlo Ventures piling back in. That round valued the company north of $1 billion, which is the point at which a startup officially becomes a unicorn and everyone has to stop being coy about it.

Eight months between a Series A and a Series B is fast. Fast usually means one of two things - real demand, or a hot market throwing money at a category. In Eve's case it is plausibly both. Legal AI is having a moment, and the numbers Eve puts up are, if you take them at face value, genuinely large: more than 450 firms as customers, 200,000-plus cases processed a year, and customers who have collectively recovered over $3.5 billion in settlements and judgments. It added 350 of those firms in the eight months between rounds.

You should take vendor-reported outcome numbers with the usual grain of salt - "our customers recovered $3.5 billion" is doing a lot of quiet work, since the customers, not the software, tried the cases. But the direction is real. Personal-injury litigation runs on an enormous, unglamorous pile of medical records - thousands of pages per case - and turning that pile into a usable chronology is exactly the kind of task nobody wants to do and AI is unusually good at. That is the wedge.

The founder story helps too. Madheswaran is a former Facebook engineer who did a stint as a venture capitalist at Lightspeed before starting Eve. Then Lightspeed backed the company. There is something pleasingly circular about a VC firm funding its own alumnus, and it is the kind of detail that makes a Series B feel less like a leap and more like a foregone conclusion.

The more interesting move, though, is linguistic. In June 2026 Eve renamed its platform "EveOS" and started calling it an operating system rather than an assistant. This is partly marketing and partly a real claim: Eve wants to run the firm's operations, not sit beside them. When a company reframes itself from "tool" to "OS," it is telling you where it thinks the value - and the lock-in - actually lives.