Artificial Lawyer recently caught up with Jake Heller, the founder and CEO of Casetext, a US-based legal AI research system that uses natural language processing (NLP) to help lawyers gain insight into their cases.
Perhaps the first thing that hits you when you see Casetext is its simple interface. There is a window on the webpage that hosts CARA, Casetext’s automated legal research assistant, and you simply drag and drop your document into that. The NLP then gets to work.
This kind of interface says to lawyers with no tech skills: have no fear, you’re going to be able to use this. And given that some lawyers wonder if they now need to code, this is a good response. I.e. no, you don’t need to code to use legal AI systems.
But, what does Casetext, or more specifically the CARA function that you can drag and drop documents into, actually do? Jake Heller, founder, CEO and former Ropes & Gray lawyer, explains: ‘CARA finds relevant case law related to your matter. You drag your brief in and it will find other cases with the same facts or with the same legal topic.’
So far so good. But is this just a standard database search? No, says Heller. Casetext does much more because it also finds new cases that are not listed in the brief. Fundamentally, it uses semantic analysis to consider millions of other legal documents that may be related to what is in the brief.
This process can also reveal unmentioned cases that may be of use to one of the parties in the litigation, as the other party may have deliberately not cited certain cases at an early stage in order to gain some advantage.
Once you’ve run your brief through CARA, you can click on any of the results to view the full text of the opinion, and some of these have been annotated with articles on certain cases.
Some of these articles are provided through partnerships with top law firms; others are authored directly on Casetext through its publishing platform, LegalPad. These articles, as well as advanced data science tools such as the Heatmap, Key Passages, and Summaries from Subsequent cases, help a lawyer to understand the issues discussed in a case.
Artificial Lawyer asks if there was a problem getting hold of the legal data, which remains a major barrier to legal AI research companies around the world, i.e. you cannot provide a user with a great AI analysis if your dataset is too small. If the treasure trove of legal data, such as court decisions, is stuck behind a pay wall, or spread across dozens of hard to use, arcane, Government websites, then no matter how good your AI research technology is it’s going to be hard to provide meaningful insights.
This was a barrier Casetext also faced, but Heller approached the challenge head on. He explains that they had no choice but to pay for access to the huge collections of US case data they needed to provide the service users would need.
‘We had to see it as a capital cost,’ says Heller, who adds that there are data companies out there who will sell this information. It just isn’t cheap.
So much then for access to justice, especially given the fact that the courts are paid for by taxpayers. Clearly there is a gap between where ‘public’ legal data is stored and the ability for people to access it. No wonder then that providing access to justice is very much a driving force in Heller’s thinking.
But, how did Casetext come into existence? Heller explains that he started coding when he was nine years old. ‘I fell in love with code,’ he says.
He found his way to law school where he was President of the Stanford Law Review and eventually became an associate at a Big Law firm, Ropes & Gray. It was a world where there was usually a high price for legal information. There was no equivalent of a ‘Github for law’ as Heller puts it, noting the well-known open source platform for coders to download and try out the creations of fellow tech enthusiasts.
‘I could see that open source legal information was the future. The old model was like Encyclopedia Britannica, which charged for access,’ Heller adds. ‘As compared to Wikipedia, which was about data science and community sourcing.’
Heller says he kept this idea in his mind for several years, but no one seemed to be acting on this fact that the legal data market model was out of date. So, he decided to quit the day job and do something about it himself. And so Casetext was born in 2013.
He raised capital, including via Y-Combinator, and the company went through several iterations and improvements.
‘There is not a single line of my code from 2013 that is still there,’ Heller notes.
Clearly this was not an easy journey, but the venture has worked. Perhaps this is in part because the user interface is so convenient. As Heller adds: ‘There is no sign-up. You can just use it. The barrier to use is very low.’
And that raises in turn a new question: how do you make a profit?
Heller says that although access is free to many users, such as law students, the advanced services Casetext provides, like CARA research automation, are available through a paid subscription. A lot of large law firms are now paying to use the system, along with smaller firms that perhaps in the past would not have had the resources to conduct this kind of research manually.
While Heller is driven by the hope of creating a legal Wikipedia or Github, he is also realistic about returns on investment.
‘We have a counter that goes ‘ker-ching’ every time we have a new paying customer,’ he says. ‘We opened a bottle of champagne for the first one. It was a wonderful feeling.’
Casetext is clearly innovative, but is it a legal AI company? In response, Heller goes through the broader capabilities of the system.
As well as NLP it is using machine learning. ‘It gets smarter. For example, if you don’t click on any of the results CARA provides it notices. It keeps tweaking its variables,’ he notes. So, yes, this is AI at work.
He adds that the system is set up to emulate a real litigator in their research work and to try and predict the types of prior cases a judge might look at when they came to examine a new case.
He also points out that in the future the system could be more proactive and contribute more to a lawyer’s work. These are, after all, early days for legal AI.
Heller hopes it will reach the point where many litigators simply use Casetext as an automatic first point of call. ‘CARA it, before you file it,’ is the motto Heller wants the legal market to have on its lips.
And what about the other legal AI research tools, such as Premonition, Ravel, Lex Machina, ROSS and also the big legal publishers? Are we looking at a ‘one legal AI research tool to rule them all’ situation where one of this group will dominate all the others? The answer is: no.
‘Lawyers are omnivores. They use a lot of different tools,’ Heller explains and stresses that no single legal AI research system solves all the problems, rather lawyers will use whichever tools they need for each type of need.
And will Casetext come to the UK and other markets? Heller says that the UK is an eventual goal, as well as Canada. But, the US, which is the world’s largest legal market and Heller’s home, is going to remain the priority for now.