Tavrn.

Here is a fact about personal injury law that nobody puts on a billboard: most of the work is filing. Before an attorney argues that a client deserves $175,000, someone - usually a paralegal, usually for weeks - has to fax hospitals, chase records, read hundreds of pages of clinical notes, and assemble them into something a jury or an insurer will actually respond to. It is slow, it is manual, and it is the single largest hidden cost in a contingency-fee practice. You do not get paid until the case settles, and the case does not settle until the paperwork is done.

Tavrn's insight is that this is a software problem wearing a legal costume. The company, founded in 2022 and based in San Francisco, builds AI agents that do the filing. Its platform retrieves medical records from providers across all fifty states, turns those records into structured chronologies, drafts demand letters straight from the case facts, handles client intake, and surfaces documents for discovery. The pitch, printed plainly on its homepage, is not visionary or grand. It is: "Move more cases to settlement faster."

The people behind it are an odd pair, in the way that startup founders often are. Pedro Paulino, the CEO, is a Harvard College dropout who credits competitive chess for whatever pattern-recognition he brings to the job. Vitor Vavolizza, the CTO, taught himself to code at eleven and left Pomona College to build the thing. Before they scaled anything, the two spent close to a year sitting next to actual paralegals and attorneys, watching where the hours disappeared. This is not the usual order of operations for an AI company, and it shows in the product: Tavrn is aimed with unusual precision at one vertical rather than at the whole abstract category of "legal work."

What it actually does

Start with retrieval, the least glamorous and possibly most valuable piece. A firm submits a single request; Tavrn's agents fan out to thousands of providers, track the status of each request in real time, and deposit organized records into the case file. The company says it processes roughly a thousand of these requests a day and delivers 95% of records within 48 hours - about twelve days faster than the manual grind. If you have ever waited on a hospital records department, you understand why a law firm would pay for that alone.

From there the work compounds. A medical chronology - the structured, dated narrative of a client's injuries and treatment - normally takes a paralegal hours to build. Tavrn says it produces one in under 24 hours, and reports cutting review time by as much as 90%. The demand letter, the document where a case becomes a dollar figure, gets drafted by AI directly from the records and the facts, then handed back to a human to review and send. There is an eDiscovery layer for surfacing digital documents, and an intake tool for the earliest, messiest stage of a case.

Why anyone believes it

Selling software to lawyers is famously hard, and for defensible reasons - the downside of a hallucinated citation or a leaked medical file is not an embarrassing tweet, it is a malpractice exposure. Tavrn's response was not to out-hype the competition. It was to lead with the least exciting material imaginable: SOC 2 Type II, HIPAA, and ISO 27001 certifications; 256-bit encryption at rest; TLS 1.2 in transit; and a promise not to train models on customer data. For a buyer who fears compliance more than they crave novelty, the boring checklist is the sales pitch.

The economics do the rest. Tavrn tells prospects they can expect something like a sixfold return on the platform's cost within a single quarter, along with faster settlements and lower per-case expense. Those are the company's own figures, and worth reading as claims rather than audited results - but they are the kind of claims a firm can check against its own timesheets within a few months. That is a short feedback loop for an industry that usually measures adoption in years.

The money

Investors have noticed. Tavrn took $1.4 million in a pre-seed round from A* Capital in early 2023, then closed a $15 million Series A led by Left Lane Capital in July 2025, with A*, Hummingbird Ventures, and Box Group also participating. That brought total funding to roughly $21.6 million. Paulino has framed the ambition in the plainest possible terms: to make Tavrn "the must-have platform for modern law practices." The company sits in a crowded field - EvenUp, Supio, and a handful of others are chasing the same medical-records-to-demand-letter pipeline - which suggests the market is real and the race is on.

What makes Tavrn worth watching is not that it uses AI; nearly everything claims that now. It is the choice of target. The company went after the part of law that no one romanticizes - the retrieval, the chronology, the endless following-up - and treated it as infrastructure to be automated rather than a chore to be tolerated. If the boring middle of litigation really can run as a background process, a lot of paralegal hours are about to look very different. And the firms that adopt first will settle their cases while the rest are still on hold with the records department.