Lawyers are among the latest white-collar professionals to learn that artificial intelligence (AI) simply does some things better than they can.
Specifically, a study released by LawGeex, a maker of AI due diligence software, found that AI was better at finding loopholes and other risks in proposed non-disclosure agreements than profession veterans.
Each lawyer participating in the study took an average of 92 minutes to go over the five contracts in the study, and was 85% accurate in his/her recommendations for changes. In contrast, the LawGeex AI software took 23 seconds to review all five contracts, and was 94% accurate in its recommendations for changes.
"LawGeex answers the question 'can I sign this?' within one hour, reducing legal bottlenecks and shortening contract turnaround time," says Noory Bechor, the company's CEO.
Granted, the LawGeex study, which pitted 20 seasoned attorneys against its AI software, might not be considered rigorous by some academics. However, Bechor says LawGeex considers the results of the shoot-out sufficiently convincing for its target audience: law firms and in-house counsel scrambling for ways to use technology to reduce costs and increase efficiencies.
"Our main priority is working with business in giving them back their time away from the drudgery of basic contract review," Bechor says. "We would be happy to speak with academics interested in conducting further research, but have no immediate plans to carry out further academic research."
Interestingly, LawGeex software is just one of a number of offerings in the rapidly burgeoning genre of AI products designed to make it easier for law firms to polish off due diligence needs.
Kira Systems, for example, makes contract analysis software that specializes in analyzing documents and highlighting specific content requiring further review by humans.
Leverton provides automated due diligence for real estate contracts, and can analyze documents in 20 languages.
eBrevia does similar contract analysis, while ThoughtRiver focuses on risk management in contracts, portfolio reviews, and investigations.
Still other AI due diligence packages include LegalRobot, Ross Intelligence, Casetext's Cara, Loom, Judicata, Everlaw, DISCO, Catalyst, Exterro's WhatSun, Brainspace Discovery, M&A Due Dilligence Robot, LexCheck, LegalSifter, Seal, and Luminance.
Experts in legal AI see the emergence of AI due diligence programs as significantly altering the way attorneys will deal with due diligence.
"No one can seriously question the coming marketplace importance of AI-driven due diligence software for reviewing contracts, and for performing other types of legal tasks," says Jason R. Baron, Of Counsel in the Information Governance and eDiscovery Group of Drinker Biddle & Reath LLP. "It's certainly happening here where I work at Drinker Biddle & Reath, and it's happening as well at other agile law firms that are attempting to use cutting-edge technology to provide new efficiencies in the delivery of legal services."
Kevin D. Ashley, a professor of law, and of Intelligent Systems, at the University of Pittsburgh, agrees. "AI-driven due diligence software will gain a significant foothold in legal practice in the next five years. Automatically annotating the different types of provisions and sentences and the various parameters will help attorneys be more efficient and productive in focusing their analyses."
Adds Guido Governatori, senior principal researcher and team leader of Data61, Australia's leading digital research network and part of the Commonwealth Scientific and Industrial Research Organization (CSIRO), "Definitely, there will be more and more applications like this in the coming years to support the legal profession. Depending on the jurisdictions, there is a growing pressure to offer legal services at a low cost."
In the long term, there's also a good chance some attorneys dealing with AI due diligence software will suffer the fate that befell many blue collar workers during the Industrial Revolution: fewer employment opportunities, according to Matthias Grabmair, a systems scientist in the Language Technologies Institute of Carnegie Mellon University. With due diligence AI, "large amounts of text will be processable within a short amount of time, likely reducing the number of experts needed for a given analysis task," Grabmair says.
In fact, as AI due diligence software enjoys increasing acceptance, more lawyers will take on supervisory roles, double-checking the work product of such systems, according to Grabmair.
Of course, human legal experts will still be needed to handle more complex cases that cannot be crunched with an easy run through the computer, Grabmair says.
"Such 'hard cases' can be reasonably classified either way, supported by competing analyses or arguments," says Edwina L. Rissland, Professor Emerita of Computer Science at the University of Massachusetts, Amherst.
Adds the University of Pittsburgh's Ashley, "When it comes to drawing semantic or inferential connections across multiple provisions in a contract or across multiple contracts, that is where human attorneys still have a clear advantage."
Before use of AI due diligence becomes widespread, some AI legal experts say there are still a few important loose ends that need tidying.
Rissland, for example, says the legal reasoning underpinning any AI due diligence software will need to be fully transparent and fully explainable. "While many folks — too many in my estimation — might be willing to accept the reason of 'because the computer says so' due to over-confidence in software, time pressure, laziness, lack of diligence, etc., thoughtful decision-makers and advisors like lawyers cannot afford to do so.
"Software should provide an understandable explanation of the reasoning behind the decision. This will become vitally important if robo-law applications are launched for the masses to handle everyday matters. A naked "Yes" or "No" is not enough."
Equally crucial, AI due diligence software developers need to get busy programming provisions that will safeguard the software against tricksters looking to game the systems, according to Carnegie Mellon's Grabmair.
"The more widely used and understood analytical technology becomes, the more one needs to plan for how it can be fooled," Grabmair says.
Joe Dysart is an Internet speaker and business consultant based in Manhattan, NY, USA
No entries found