In a 2015 Altman Weil survey of law firm managing partners, 47% said that, in the next five to ten years, they could envision a law-focused artificial intelligence (AI) tool replacing paralegals, 35% could envision the replacement of first-year associates, and 19% could envision the replacement of second- and third-year associates. As a current law student, I can’t help but wonder whether I can beat the clock and survive a robot invasion.
Artificial Intelligence and its Applications in Law
Despite the traditional aversion to risk of new technology, lawyers have been paying more attention to recent technological advancements, partly due to increasing demand from clients to cut legal fees after the 2008 crisis. Moreover, the development of AI technologies, such as machine learning, advanced natural language processing (NLP), and data analytics, have enabled computers to perform tasks once thought to be only feasible if completed by people. Computers used to be solely capable of deductive reasoning, which means tasks could be automated only when step-by-step rules are provided. Older technology was replacing clerical and support staff, but AI could start to displace lawyers themselves.
AI has been hard at work according to Michael Mills, Chief Knowledge Officer of Davis Polk and Wardwell. It is currently applied in five main areas in the legal profession: legal research, E-discovery, outcome prediction, self-service compliance, and contract analysis. Traditional legal research databases like Lexis and Westlaw are incorporating NLP and machine learning to refine search results. E-discovery software uses predictive coding based on a “seed samples” of documents provided by lawyers to sift through the gigantic discovery data sets, which can significantly cut cost for clients. Outcome prediction tools use data mining and analytics to forecast outcomes of litigation. Self-service compliance provides answers to legal, compliance and policy questions based on the facts or context provided by clients for issues ranging from disputing a parking ticket to complying with Foreign Corrupt Practices Act. Contract analysis platforms can “read, interpret and summarize” key clauses, which enables lawyers to understand the rights, obligations and risks in a company’s contracts faster. More importantly, the AI software could manage and classify existing contracts, and potentially help streamline the processes through contracts’ lifecycle.
Problems with Adopting AI Technologies
Despite many benefits AI software brings to the legal profession, new problems are created. Structured tasks, such as document review, can be done faster and cheaper with the assistance of AI software. Many engineers and scholars believe that if the software is trained with suitable “seed samples” and lawyer guidance, the result could be more accurate as well. In Lola v. Skadden Arps, the Second Circuit concluded based on a high level generalization, that computers can perform document review as well as humans. 620 Fed. Appx 37 at 45 (2d. Cir. 2015). This conclusion is problematic because neither the scholars nor the court seriously inquired how the machine approaches the task differently from a human and how accurate the software is. Moreover, reducing a legal task to a clear set of computer-doable rules may “over-simplify the legal issue and overlook the complexity.” Potentially, using AI for document review could incur errors that are not expected or immediately apparent.
Another pressing issue is that of accountability and legal ethics. The quality of results of a particular software varies greatly from case to case, depending on the quality and quantity of prior training data. Most tools don’t have the mechanism to determine whether and how to notify a user that the computer’s “best” answer is not very good. Legal professionals with sufficient knowledge can bear the risk and assess the quality of results themselves and accept the responsibility for their results. The problem is more salient when the legal tool is consumer-facing. Would the AI tools be held to the same legal ethics rules? As AI systems are increasingly used by lawyers and end users, it will be interesting to see whether the American Bar Association will issue ethics opinions on AI.
Is AI Eating the Legal Work from Bottom Up?
The notion that junior lawyers’ work, such as document review, could be replaced by robots earlier than the more complex work partners do is generally true. However, seniority or experience is not the deciding factor. The real criteria is how structured and routine the work itself is. For example, the junior associates usually spend most of their time doing document review (litigation) and due diligence (corporate). Document review is almost purely structural, since it’s a process to identify relevance to existing topics and questions. Comparatively, due diligence for a deal contains a more unstructured component of actively looking for surprising or unexpected terms. More experienced litigation attorneys do more legal writing while the corporate counterparts work on document drafting. Legal writing, especially when crafting argument, requires much “conceptual creativity and flexibility” to advance one’s position. Document drafting is relatively consistent and structured, with more uses of standard templates and forms.
Client interaction and other communications are the most unstructured components of lawyering. While communicating with others, we navigate the other party’s language, gesture, and emotions even though we can’t quite articulate it all. Although tools like Cataphora can recognize sentiment in an email message, AI is not yet capable of understanding unscripted human interaction. Since partners spend much more time interacting with related parties than associates do, they are not as affected by the growth of AI technology.
Get in the Game Now?
It’s inevitable that technology advancements will force us lawyers to change how we work. A recent study by McKinsey & Co estimates that 23% of lawyer’s time could be automated, while a similar study by Frank Levy at MIT and Dana Remus at UNC School of law concludes only 13% could be automated. However, the decreased legal cost of AI adoption may broaden access to justice and expand the overall legal market. Most importantly, adopting AI technologies in the legal field enables lawyers to devote their time to more valuable work like legal strategy and analysis, which is the reason why we came to law school in the first place, right?