Watch out, attorneys. The bots are coming--and they're getting good at your job.

For some lawyers, contract review takes up a huge chunk of time and it can be extremely tedious. 

That's why the co-founders of LawGeex began their company in 2014. The Tel Aviv-based startup creates software that uses artificial intelligence to study contracts, flagging any language or stipulations that seem out of the ordinary.

LawGeex released its software to the public early last year. Recently, though, the company decided it wanted to truly put it to the test. In a study overseen by attorneys from Duke University and Stanford University, the startup had 20 experienced lawyers separately study five new non-disclosure agreements, while the A.I. did the same. 

The results: The lawyers came in at a solid 85 percent accurate. The software? Ninety-four percent. What's more, while the average lawyer took 92 minutes to complete the task, the company's A.I. did it in just 26 seconds--so in addition to being more accurate, it was 200 times faster than its human counterparts.

Noory Bechor, Lawgeex co-founder and CEO, spent six years as an attorney at Israel's largest law firm before deciding to pursue the new venture. "I had a growing frustration with how inefficient the legal world is," he says. "It's very repetitive and mundane, and there was no real technology that helps lawyers do their jobs better and more efficiently."

In his early professional days as a paralegal, Bechor had spent much of his time reviewing documents. "Once you've seen hundreds of examples of a specific contract type," he says, "the concepts keep repeating themselves. I said, if this is so repetitive, it can be automated."

Through an acquaintance, Bechor was introduced to Ilan Admon, a tech industry vet. Together they developed a proof of concept and raised more than $8 million from angel investors and VC firms including Recruit Co. and EverythingMe.

LawGeex's software compares incoming contracts with a set of standards that can be preset by the user. It then identifies clauses that might need altering before the contract can be approved. The product has already gained traction with in-house legal teams: LawGeex counts Sears, Deloitte, Skyview Capital, and Key Energy among its early clients, and it says that list also includes major banks and insurance companies.

Developing A.I. that can study contracts for uncommon language or clauses isn't quite as straightforward as it might sound--the software wouldn't be very useful if it flagged every uncommon piece of information. Someone signing a lease in Chicago, for example, probably doesn't need to be alerted to the fact that their landlord's address is in Michigan, but they would want to know about a stipulation that charged them five months' rent for breaking the lease.

To solve that, LawGeex used attorneys to help train its A.I. That helped give the software the ability to distinguish between information that's not just uncommon from a quantitative perspective, but from a qualitative perspective as well.

Which raises the question: Did those lawyers just play a part in rendering themselves redundant? Bechor says no. "We really try to focus on the value this solution creates in terms of efficiency, making business move faster, saving money and saving time," he says. Repeating the popular Silicon Valley talking point, he says he sees the software as giving workers time for more interesting, creative tasks. Although, he admits, "If I'm a person that only reviews these simple contracts for a living, I'm going to have to adapt."

Bechor sees the software as useful to businesses, which can often bypass legal review for the sake of speed or cost efficiency. "When they do that," he says, "they end up taking on increased legal risks."

Like with any technology, the software has its limits--so for most lawyers, their jobs are safe for now. "We're not claiming to be more accurate than lawyers all the time and in any type of work," Bechor says. "What we are showing is that on the mundane, repetitive, simple stuff, technology can actually do a better job that humans."

5 Things Artificial Intelligence Can Already Do Better Than Humans
Published on: Feb 26, 2018