Criminal Law Library Blog
Theo AI, a cutting-edge generative AI (genAI) startup based in the United States, has positioned itself as the first predictive AI platform designed specifically for litigation. Backed by an impressive $2.2 million in pre-seed funding, the company aims to achieve what no other legal tech firm has successfully accomplished: accurately predicting the outcomes of legal disputes well before their resolution.
The concept of AI-driven litigation prediction is not entirely new—Artificial Lawyer has been reporting on similar applications since 2016. However, Theo AI’s innovative use of genAI technology marks a significant departure from earlier approaches, potentially setting a new benchmark in the field of legal technology.
From Artificial Lawyer November 2024:
“As they explain: ‘Using a proprietary data model and prediction engine, Theo Ai helps legal professionals make educated decisions about the likely outcome of cases. The funding will be used to further enhance their prediction engine, expand practice categories, and accelerate customer growth.’
By analyzing similar cases and likely arguments, Theo Ai’s data model estimates the probability of winning a case, in addition to predicting the estimated award. Early users of Theo Ai found that the platform’s algorithms verified the results of their underwriting and due diligence teams. With Theo Ai, firms have access to a data-driven pipeline using real-time analytics and predictive modelling as new facts and evidence emerge.
Theo Ai thus enables customers to ‘identify and predict cases with the highest odds of success, uncover cases they might have missed, and access case summaries and key financial drivers all in a single offering’, they said…..
As they say: ‘With over 275,000 new lawsuits filed each day, choosing which cases to take is essential for the legal industry. The average mid-sized firm reviews roughly 650 cases per year, which can take anywhere between 7 to 30 days to manually review. With Theo Ai, that time is compressed into seconds – allowing legal teams to cover more ground and focus on winning cases.’
And as noted, if this really works the litigation funders will be lining up around the block to use it.
The startup is led by Alex Alben (UCLA Law Professor and Tech Executive), Patrick Ip (ex-Google and UCLA Law MLS) and Tiago Luchini (4x CTO/Founder). And the investment round was co-led by NextView and nvp capital with participation from Ripple Ventures, Beat Ventures, and SCVC Fund.
Plus, Co-Founder and Partner at NextView, Rob Go, added: ‘With backgrounds in both law and tech, Theo.Ai’s leadership team understands the complexities legal firms face and how to leverage advanced technology to address those challenges. Their experience allows them to build a platform that addresses the needs of the everyday economy and truly reflects the nuances of legal decision-making, giving customers a significant edge in strategy and case outcomes.’
OK, all well and good, and Artificial Lawyer applauds everyone’s enthusiasm here. However, there is one small challenge: half a dozen (or more) companies have tried over the years to use AI to make accurate predictions about case outcomes. It’s fair to say some have struggled. While those that have done alright have needed to leverage the expert views of experienced trial lawyers to quality check and advise on anything the tech predicted. I.e. they still had to rely heavily on human judgment nonetheless.
Why is this? Because court cases are really, really, really unpredictable. One witness can change it all. News going on around outside the court that month can change the view of the judge, no matter what their past positions. Parties can suddenly change tack and settle because of things where there is no public data that indicates a shift is coming. Maybe the lead trial lawyer had a bad day. Who knows?
And that’s the challenge.
It may work for smaller, more mundane cases, e.g. small insurance claims that are the same as 200 others also heard in the same area, but once you get into complex disputes between large corporates, who knows what will happen? Moreover, not all disputes end in a completed court trial – many commercial matters settle before court, or just as the case begins. That data is not necessarily public, or detailed even if some of it is made public.
And where companies such as the now defunct Gavelytics used past court data to predict elements of a case, e.g. whether a judge was likely to agree to a certain motion, although it was useful, it was only one small part of an overall case and could not – nor would try – to cover a whole trial.
Any road, if this site seems a bit skeptical it’s because we have been here many times before. The reality is if you could easily, and very economically, predict the outcome of court cases with high accuracy day-in and day-out then why would people ever go to court…? There would be no point, as you’d know before you began what the outcome would be, so the losing party would always quit while ahead.
AL wants to support this new startup, for sure, but also let’s be realistic about what’s being worked on here – predicting the future with AI is a tough one. Best of luck to them in any case, maybe they will crack the code on this one, unlike those that came before.”