How Legal AI became more accurate than lawyers

In his book Tomorrow’s Lawyers, lawyer, academic, and tech expert, Professor Richard Susskind, says “The next two decades will bring more changes to legal institutions and lawyers than have the last two centuries

This prophecy came a step closer this week as legal AI proved itself more accurate than lawyers for the first time on a staple legal task –reviewing and approving contracts. The study was overseen by top US law schools and veteran corporate lawyers (including for instance Bruce Mann, a former senior partner at top US law firm, Morrison Foerster a Harvey Specter- like deal-maker who has handled more than 300 IPOs and over 200 mergers & acquisitions).

In the controlled experiment, 20 top US-trained lawyers took on a legal AI platform, LawGeex. Both the experienced corporate lawyers and the AI pored over five unseen Non-Disclosure Agreements to find a list of common 30 issues (vetted by contract experts from Duke University and the University of Southern California). Each participant (the AI included) was given 4 hours to issues spot clauses in the contracts. The result saw lawyers achieved an 84% average accuracy in spotting the issues in the correct places. The average time taken by the 20 lawyers to review the all the NDAs was 92 minutes. In contrast, the AI scored 94% accuracy, completing this feat in only 26 seconds.

How did this happen? 

LawGeex, part of a growing legal AI field, has been trained to detect issues on more than a dozen different legal contracts, ranging from software agreements to services agreements to purchase orders. This specific research focused solely on NDAs – the most common form of business contract. NDAs are typically used to create a legal obligation to secrecy, and compel those who agree to them to keep information confidential and secure.

The LawGeex AI was trained on tens of thousands of NDAs, using custom-built machine learning and deep learning technology. The machine was trained based on an exclusive corpus of documents that presented the LawGeex algorithm with a variety of examples, which allowed it to distinguish between different legal concepts. This level of technology for analyzing legal documents has only been possible with advances in computing over the last five years.

Computers convert the text into a numeric representation. The image below is a visualization of how computers read the text. Each dot represents one paragraph in the semantic space. The different colors shown represent different legal issues. Pink dots, for example, represent samples of non-compete issues, and purple ones represent governing law sections. Training an AI engine is similar to training a new lawyer – exposure to different examples is crucial in developing a deep understanding of the legal practice.

Explosion of Legal AI 

This research arrives as we have seen an explosion of legal AI. The promise of AI for lawyers lies in the automation of previously tedious manual processes that have not changed for decades. It is allowing lawyers to devote more time to valuable and strategic work. The infographic below is taken from a report on the changing face of legal technology for in-house lawyers. This reveals AI’s appeal in daily work including drafting and reviewing contracts, mining documents in discovery and due diligence, answering routine questions or sifting data to predict outcomes.

AlphaGo moment for Lawyers – but not Go-like emotions

It should be said that unlike previous AI vs Human contests, few are bemoaning humanity’s defeat at the hands of the machines when it comes to this particular aspect of legal contract review and approval. Of course, in the annuls of AI vs Human competitions, most famously AlphaGo famously beat Lee Sedol a 9-dan professional in a five-game match of the ancient game. In that historic tournament even the developers of AlphaGo program were pained to see the noble Lee Sedol beaten by their machine, and what this might mean for humanity. This however was not an ancient game, but a victory over every day and “in the way” NDAs that no lawyers enjoy reviewing. Like many of the common mundane legal tasks, AI and better technology are helping to solve legal and business problems. In the case of contracts, alone, the typical Fortune 1000 company maintains 20,000 to 40,000 active contracts at any given time. But The International Association for Contract & Commercial Management (IACCM) has found that 83% of businesses are dissatisfied with their organization’s contracting process. In addition, in the real-world NDAs (used in this study) slip to the bottom of the pile, taking companies a week or longer to approve – a process that frustrates other departments and slows down deals.

In the words of one participant in the test, Grant Gulovsen, an attorney with more than 15 years’ experience: ““Participating in this experiment really opened my eyes to how ridiculous it is for attorneys to spend their time (as well as their clients’ money) creating or reviewing documents like NDAs which are so fundamentally similar to one another”.

Though the law is historically slow moving (to quote Professor Susskind again the profession is only behind the clergy in its love of change), the profession is waking up to the challenges of using AI and technology, slashing hours of manual processes. In one case, that has grabbed headlines, JP Morgan uses AI to do in seconds what it took its lawyers 360,000 hours to do. The program does the mind-numbing job of interpreting commercial-loan agreements. It is fair to say this is not what lawyers went to law school to do, and AI is helping alleviate the excesses of this type of lawyering.

Lawyers waking up to a present and future for Legal AI

The American Bar Association (ABA) has already modified its rules to extend a lawyer’s the duty of competence to keep “abreast of changes in the law and its practice” to include knowing “the benefits and risks associated with relevant technology.” Many State Bars have followed, extending lawyer “competence” beyond knowledge of substantive law to a duty of technological competence.

With the steady march of technology to help with the basic parts of tasks, lawyers are able to do more strategic advising and less drudgework. What is clear is that lawyers failing to capitalize on the competitive advantage of such technology are unlikely to thrive into the next decade.

About the Author

Jonathan Marciano is Director of Communications at LawGeex. He regularly writes on legal technology and its intersection with business. He has formerly worked at successful startups and at Big Four consultancy, Ernst & Young.