Skip to content
PlaintiffStack
Blog

Industry5 min read

Workers' comp is the most under-served practice in legal AI

The best-funded plaintiff AI companies were built for personal injury. Applicant workers' comp has had far less attention, and its work suits page-cited AI.

By The PlaintiffStack team

Follow the money in plaintiff legal AI and you arrive at one practice: personal injury.

EvenUp raised a $150 million Series E in October 2025 at a valuation above $2 billion, bringing it to $385 million raised (EvenUp). Eve raised a $103 million Series B in September 2025 at a valuation over $1 billion (LawSites). Supio has raised $91 million (Supio). All three sell first to personal injury firms. EvenUp's home page names personal injury and nothing else (EvenUp). Supio's names personal injury and mass torts (Supio). Eve reaches further: its site now has pages for labor and employment, Social Security disability and workers' compensation.

Applicant comp isn't empty. Amber Legal sells an AI agent to California comp firms, plaintiff and defense, from $799 a month. Comp case management systems have handled forms and filing for years: MerusCase, for one, files DWC packets through JET File (MerusCase), and SmartAdvocate has a workers' comp module (SmartAdvocate). But measured by funding, product pages and marketing, comp gets a small share of the attention PI gets. That gap matters, because comp work suits the right kind of AI unusually well.

Why the money went to personal injury

Personal injury has a clean unit to sell against: a demand letter, sent to a carrier, with a dollar figure behind it. A vendor can price per demand, show a firm what it gets, and repeat that in every state with the same product.

Comp is harder to sell into, for three reasons.

  • The fees are smaller. An applicant attorney's fee in California is a percentage of the benefits awarded, and the WCAB has to approve it (DWC fee disclosure form). A product priced at hundreds of dollars a demand takes a large bite out of that.
  • The process is the state's own. California has panel QMEs and AMEs, utilization review and independent medical review, EAMS and ADJ numbers. A national product has to learn each state's board separately.
  • The files run for years. A comp case can stay open across years of treatment, several med-legal evaluations and more than one dispute about benefits. There's no single document at the end that the whole product can be built around.

None of that makes comp a bad market. It makes it a different one.

Why comp work suits page-cited AI

Most of the work in an applicant file is reading, extracting and checking against a rule. That is the work AI does well, as long as every fact it pulls out points back to its page.

Comp also punishes a wrong answer. In July 2025 the WCAB sanctioned a California firm $750, reduced from $2,500, over a petition that failed to cite the evidentiary record and appeared to cite authorities that did not exist. The judge speculated that AI may have been used; the firm said there was no finding about AI, that the attorney had left, and that it had added safeguards (Business Insurance). Whatever produced that petition, the lesson for comp firms is the same: a fact or citation nobody traced to its source is a liability.

What to ask a vendor

If you run a comp practice and you're looking at AI, these questions sort the products quickly.

  1. Was it built for comp, and for California? Hand it a QME report, a UR denial and a TD payment history, and see what it does with each.
  2. Does every fact cite its page? Not the document. The page, and the words on it.
  3. Can it read your scans? Comp files are full of faxed reports and scanned records. Ask what share of a real file the tool will actually read, and what it tells you about the rest.
  4. Does it know the deadlines, and show the rule? A deadline you can't check against its rule is one you'll have to compute again.
  5. Does it do the benefits math? TD periods, rates and late payments, laid out so a case manager can check them.
  6. Does it file in EAMS? Many AI tools don't. Know what you'll still do by hand.
  7. How does the price compare to a comp fee? Work it out per case, on your own inventory.
  8. Who decides? The tool should propose. A person on your team should accept.

Where PlaintiffStack stands

Workers' comp is built into PlaintiffStack, on California rules: phases from intake through benefits, treatment, QME/AME, the WCAB and settlement, and pages for benefits, body parts, authorizations, med-legal, the WCAB and liens. Case tools run a benefits audit (TD periods, rates and late-payment exposure), write the QME cover letter and records index, and draft WCAB petitions with the caption and ADJ number. Each morning, checks flag certified disability with no payment, IMR deadlines after a UR denial, and records due to the QME before the evaluation.

Every fact an agent proposes cites the document, page and words it came from, and a person accepts it before it lands in the case.

What it doesn't do yet, said plainly:

  • No OCR. Agents read searchable PDFs, Word and text files. Scanned records and faxes aren't read.
  • No PD rating calculator.
  • No EAMS filing. PlaintiffStack tracks the ADJ number and watches the case, but you file through your current filing tool.

The price is $250 per case with every app included. AI usage is billed separately: each tool shows an estimated cost before it runs, and how usage is billed is part of your quote. In a practice where fees are a percentage of benefits, per-case pricing has to earn its place, so run your own numbers on the pricing page.

See the workers' comp practice page, our guide to AI for workers' comp attorneys, or the guide to workers' comp case management software.