Legal work is incredibly labor- and time-intensive, requiring piecing together cases from vast amounts of evidence. That's driving some firms to pilot AI to streamline certain steps; according to a 2023 survey by the American Bar Association, 35% of law firms now use AI tools in their practice.
OpenAI-backed Harvey is among the big winners so far in the burgeoning AI legaltech space, alongside startups such as Leya and Klarity. But there's room for one more, says Jerry Zhou and Kyle Lam, the co-founders of an AI platform for personal injury law called Supio, which emerged from stealth Tuesday with a $25 million investment led by Sapphire Ventures.
Supio uses generative AI to automate bulk data collection and aggregation for legal teams. In addition to summarizing info, the platform can organize and identify files -- and snippets within files -- that might be useful in outlining, drafting and presenting a case, Zhou said.
"After attending numerous conferences and meeting with hundreds of lawyers across the U.S., Lam and I decided to focus on personal injury and mass tort plaintiff law," said Zhou, who's also Supio's CEO. "These are practice areas that require compiling thousands of documents from multiple sources, and analyzing and finding information from the data within them."
Zhou and Lam are childhood friends whose career paths have often intersected. The pair worked at Microsoft, specifically in the Office 365 org, and together again at tax compliance software firm Avalara.
The idea for Supio came about after Zhou and Lam left Avalara to pursue building a business that could, in Zhou's words, "help understand complex data and identify critical connections within certain data."
"We pursued the legal industry because we knew it wasn't just document-heavy -- it was also due for technology innovation," Zhou said. "[These are] practice areas that require compiling thousands of documents from multiple sources and analyzing and finding information from the data within them."
Personal injury and mass tort cases, or civil suits filed on behalf of victims harmed by negligence, like the sale of defective products, typically unfold across paperwork including medical records, police reports, insurance claims, financial statements, consumer complaints and so on. What Supio does, Zhou explained, is generate demand letters -- letters outlining the legal disputes to be resolved -- as well as supporting documentation, while letting users search the evidence through a chatbot-type interface.
It sounds a lot like EvenUp, a startup that taps AI to generate legal documents to assess injury cases. Companies like Lawyaw and Atrium also apply AI to the task of drafting initial complaints.
But Zhou claims that Supio is more complex in its technical approach.
"Law is extremely complex and nuanced, and most creators of work productivity tools lack a true understanding of the legal documents lawyers ultimately have to produce, which inhibits the development of accurate [AI] models," Zhou said. "Supio has hundreds of models running at a given time with different functions to try to understand and classify documents. We then measure this against the work products that are expected and improve these results gradually."
AI like Supio's is powerful stuff in theory -- but also fraught. Given the sensitive nature of most legal disputes, lawyers and law firms might be reluctant to grant (or be prohibited from granting) a tool like Supio access to any case docs.
Late last year, the State Bar of California released guidance instructing legal professionals to refrain from putting clients' information into AI tools that "lack reasonable or adequate security." (Zhou says that Supio stores client data in its country of origin and has security protocols that adhere to privacy regulations including HIPAA and GDPR.)
Another concern with AI legaltech is AI's perjurious proclivities. Last year, a group of lawyers with the firm Levidow, Levidow & Oberman, P.C. tapped ChatGPT, OpenAI's AI-powered chatbot, to prepare a personal injury complaint against an airline. The result was disastrous: ChatGPT invented citations, misidentified judges and referred to airlines that don't exist. The federal judge overseeing the case ultimately imposed a $5,000 fine on the lawyers and their employer.
Courts are frantically preparing for a rise in inaccurate filings attributable the uptake in AI legal research tools.
In November 2023, the U.S. Court of Appeals for the Fifth Circuit proposed a rule (since scrapped) requiring that any professional filing legal paperwork drafted with the help of AI certify that a human reviewed the documents for accuracy and approved them. Earlier that same year, a district judge in Texas issued an order banning the use of generative AI to write court filings without a human fact-check.
The risks with AI are such that, in a recent survey of over 300 general counsel and senior legal officers at large corporations, 25% said that they believe their outside counsel shouldn't use AI. A separate poll by Thomson Reuters found that one in five law firms have issued warnings around the use of AI.
Zhou makes the remarkable claim that Supio's AI performs "better than human levels of accuracy" and "without hallucinations," i.e. it never fibs.
"Supio is providing flexible software featuring AI that can organize unstructured data and produce reliable results because we know timing and accuracy are critical," he said.
It's not clear what's meant by "human level"; Zhou didn't share any test or benchmark results. But I would note that just because AI can achieve feats like passing the bar exam doesn't mean it has the skills attorneys gain through experience and education. (The National Conference of Bar Examiners argues as much.) As for the "without hallucinations" bit of Zhou's claim, it's not backed up by data, either -- at least none that Zhou volunteered.
Yet some firms believe Supio has promise.
According to Zhou, Supio is currently working with around 30 personal injury and mass tort law firms and expects that number to reach 100 firms by the end of the year. The startup's annual recurring revenue has eclipsed $1 million, meanwhile, with most of the money coming from subscription fees that Supio charges based on case volume.
It could be a case of keeping up with the Joneses.
In a survey of legal execs published by LexisNexis, nearly all (90%) said that they anticipate their investment in generative AI will increase over the next five years; the same poll found that 43% of firms now have a dedicated budget for generative AI. Gartner predicts that the allure of generative AI will drive the legaltech market to $50 billion in value by 2027, almost double what it was worth in 2022 ($25.6 million).
Against this dramatic backdrop, Seattle-based Supio keeps chugging along. The company has 27 employees and expects to double headcount in the 12 months.
Having raised a total of $33 million, Zhou says that Supio aims to expand its customer base in the near term -- and eventually to "scale to address other law specialities."
Bonfire Ventures and Foothill Ventures also participated in Supio's latest tranche (a Series A). Zhou says it was oversubscribed but wouldn't give the valuation.
