EvenUp.

01 // The SceneThe lawyer never opens the envelope

A demand letter, drafted at 2:14 a.m. by a model named Piai, lands on an insurance adjuster's desk in Phoenix. Nobody at the firm stayed up writing it.

Picture a personal injury law office in 2026. The partner is at her daughter's recital. The paralegal is asleep. A client - rear-ended on the 101 four months ago, three cracked ribs, $42,000 in medical bills - is waiting for someone to fight for him. And somewhere in a data center, EvenUp's claims intelligence platform is doing the fighting. It reads the medical records, builds the timeline, names every provider, prices the pain, and drafts the demand. By breakfast, the lawyer has a 47-page package that would have taken a week to assemble by hand. She signs it. The envelope goes out. Nobody, except the adjuster on the other end, ever thinks about the labor it replaced.

That is what EvenUp does. Not artificial intelligence in the abstract, but artificial intelligence applied to one specific, deeply unglamorous corner of the American legal system - the corner where most injury victims live or die financially.

02 // The ProblemAn industry held together by paperwork

Personal injury law is, by some estimates, a $50 billion annual industry in the United States. It is also, by almost every estimate, an absolute disaster of paperwork. A single case can involve thousands of pages of medical records spread across a dozen providers, ICD codes that need to be matched to treatments, bills that need to be reconciled, missing dates, missing diagnoses, missing context, and a demand letter that must somehow make all of it legible to an insurance adjuster who would prefer not to read it at all.

The work is not technically hard. It is just relentless, repetitive, and easy to do badly. Done badly, it costs the client money - sometimes life-changing amounts of money. Done well, it requires a kind of attention to detail that lawyers, frankly, would rather spend on the courtroom. The result is a justice gap. Clients with deep-pocketed firms get thorough demand packages. Clients without them do not. The settlement reflects the paperwork, not the injury.

It would be funny if it were not, very specifically, the reason some people walk away from accidents broken and broke.

03 // The BetThree founders, one inconvenient observation

EvenUp was founded in 2019 by Rami Karabibar, Raymond Mieszaniec, and Saam Mashhad. Karabibar had spent his prior years at Waymo, where the friction of injury claims at scale was visible up close. Mieszaniec had watched the same friction from the personal side, after a serious accident in his own family. Mashhad was a practicing attorney who had seen, from inside the law firm, how much settlement value evaporated when a demand package was assembled in a rush.

Their bet was small and slightly heretical. AI would not replace personal injury lawyers. It would absorb the part of the job that lawyers handed off to junior paralegals at 7 p.m. on a Thursday. That hand-off was where the quality dropped, where the client lost money, where the justice gap widened. Close that gap, and you do not just build a software company. You move the floor on outcomes for every plaintiff in the country.

This sounded, in 2019, like a perfectly reasonable thing to say at a Y Combinator demo day and a perfectly unfundable thing to actually do. The total addressable market was real but ugly. The customers were small law firms with cash-flow problems and a deep institutional skepticism of anyone who used the word "platform" in a sentence. Selling AI software to plaintiff lawyers, the joke went, was like selling vegetables to teenagers.

04 // The ProductPiai, and the long tail of the medical record

The thing EvenUp actually sells is called the Claims Intelligence Platform, and the engine under the hood is a proprietary model called Piai. Piai was not built by pointing a general-purpose LLM at a stack of injury records and hoping. It was trained on hundreds of thousands of personal injury cases and millions of medical records, then wrapped in a workflow that includes more than a hundred in-house lawyers, paralegals, and medical experts who review the output before it leaves the building.

This last detail matters. Most legal-AI startups have spent the last three years discovering, painfully, that lawyers do not want unsupervised AI in the loop. EvenUp built the supervision in from the start. The model drafts. The humans review. The lawyer signs. The settlement gets paid. Each step accumulates training data that makes the next case faster.

A short, unromantic list of what the platform does:

05 // The ProofThe numbers, then the receipts

It would be tedious to argue in the abstract that this is working. So here are the numbers, in the order an investor would ask for them. Over 2,000 personal injury firms are on the platform. Roughly 20% of the Top 100 firms in the country are customers. More than 200,000 cases have run through the system. The platform has helped claim over $10 billion in damages for injury victims. EvenUp has raised $385 million across four rounds in two years, all of them preempted by investors who came in before EvenUp went out to ask.

06 // The MissionClose the gap, on purpose

Every venture-backed company has a mission statement. Most of them, if you read them out loud, fall apart immediately. EvenUp's is unusually load-bearing. The stated mission is to close the justice gap - to make sure outcomes for injury victims depend on the facts of the case, not the resources of the plaintiff or the workflow of the firm. That sentence does real work, because the product actually advances it. A small firm using EvenUp's demand letter can put together a package that, before AI, only a well-staffed shop could afford to produce. That is the gap, in one paragraph, getting smaller.

There is a version of this story where EvenUp ends up being mostly a margin-expansion tool for large firms that were already doing fine. That version is plausible. The version EvenUp wants is the one where smaller firms, plaintiff lawyers in secondary markets, and clients without rich representation start to get outcomes that look more like the ones the big firms get. The honest answer is that it is too early to know which version wins. The data so far - the cases, the dollars, the customer mix - suggests both are happening at once.

07 // TomorrowWhat changes when paperwork is not the constraint

Strip out the AI talk and EvenUp is, fundamentally, a company that has decided to industrialize the boring middle of a high-stakes process. That is a less exciting pitch than "AI is coming for the lawyers," and a more interesting one. Because if the boring middle of personal injury law gets cheap, fast, and accurate, then the rate-limiting step of the entire industry moves. Lawyers spend less time on records and more time on clients. Cases close faster. Settlements get priced closer to the underlying facts. Insurance companies, who have used the friction of bad paperwork as a negotiating tool for decades, have to negotiate against well-prepared opposing counsel even when the plaintiff's lawyer is a solo practitioner in a small town.

That is not a tomorrow that arrives all at once. It is the kind of tomorrow that shows up in settlement data three or four years from now, quietly, in a chart almost no one outside the industry will look at. The number that matters is the gap between what an injury actually costs a person and what they actually get paid. EvenUp is betting the entire company that AI can shrink it.

08 // ClosingThe envelope, revisited

Six years ago, three founders looked at the most tedious part of personal injury law and decided it was the most important. They were not wrong. EvenUp now sits at roughly two billion dollars on the back of a thesis that is, by Silicon Valley standards, almost embarrassingly modest - that the way to fix justice is to fix the paperwork, one demand letter at a time. The story is not that AI replaced the lawyer. The story is that the part of the job nobody wanted to do anymore got done, well, by a machine, while the lawyer was at her daughter's recital. The envelope is sealed. Somewhere in Phoenix, an adjuster reads page one.