In recent years, tools leveraging artificial intelligence have become increasingly pervasive in the e-discovery world, with terms such as continuous active learning and natural language processing entering the vernacular of lawyers, IT professionals, and clients seeking ways to reduce costs and improve quality. But while technology-assisted review (a.k.a. predictive coding) may be the most well-publicized application of AI in the legal community,machine learning and related branches of AI can be harnessed in a variety of other contexts to the benefit of our clients. In this article, we will consider how law firms can extract the most value from AI to help resolve the tension between delivering the high quality legal advice clients demand, while simultaneously keeping fees within clients’ tightening legal budgets.
Before proceeding any further, some level-setting is in order: just what is machine learning? The term—originally coined in 1959 by an IBM researcher—refers to the branch of AI permitting systems to learn and improve by analyzing data and gleaning patterns and insights without requiring explicit, if-then style programming. When Deep Blue was able to beat Garry Kasparov at chess, when Watson was able to beat Ken Jennings at Jeopardy, and when AlphaGo was able to beat world champion Lee Sedol at Go, they all were able to achieve those feats in part due to machine learning.
Although less newsworthy to the general public, machine learning first sent shockwaves through the legal profession just over a decade ago when predictive coding technologies started to hit the market. Promising to upend discovery (typically the most labor-intensive, inefficient, and expensive phase of civil litigation) by single-handedly separating the relevant wheat from the irrelevant chaff in vast data sets, these tools garnered outsized attention before encountering resistance from attorneys comfortable with the status quo and clients loathe to be the guinea pig in the literal court of public opinion. Fast forward to the present, and we now have many judicial decisions endorsing the prudent use of technology-assisted review (TAR) to augment, but not supplant, the lawyer’s role in discovery. We also have sophisticated corporations insisting that their outside counsel use TAR to keep discovery costs in check. And so, inevitably, we have innovative lawyers and legal technologists pondering ways to replicate the AI-powered disruption of discovery in other areas.
One such field ripe for AI disruption is on the transactional side of the house, where corporate lawyers routinely have to perform highly regimented review, drafting, and editing tasks that are tailor-made for machine learning and natural language processing engines. Before any merger or acquisition is consummated, the acquiring entity understandably needs to assess the risks and legal implications of the transaction, and it does so by engaging attorneys to perform due diligence. As part of that exercise, junior attorneys customarily review a critical mass of contracts, leases, and other agreements to which the target entity is a party, summarizing their contents and flagging any anomalies or irregularities in a systematic fashion.
Jason Lichter, Director of Discovery Services, Pepper Hamilton LLP
This tedious process has, over the last few years, been the focus of a litany of AI-powered startups promising faster, cheaper, and better due diligence workflows. At the same time, other tech companies are attempting to mechanize the drafting and negotiation of non-disclosure agreements and other largely-boilerplate contracts. These technologies may still be in their nascent stages, but—as in the discovery world—it seems inevitable that clients will come to expect their use as they mature and adoption swells.
Raising client satisfaction through purposeful use of AI is certainly one way to increase law firm profitability, but another is by reducing the mundane, non-billable time spent by attorneys entering—and by administrative personnel correcting—billing narratives
Litigation document review, deal diligence, and contract drafting are all services performed almost exclusively by lawyers, but there are many opportunities for AI to improve the functioning of a law firm that impacts the billable hour more indirectly, yet no less critically. One example is knowledge management, which incorporates people, processes, and technologies designed to help attorneys find and re-use existing resources and work product to avoid having to reinvent the wheel each time the same legal question arises. Just as e-discovery has evolved from sifting through boxes to running keyword searches to using analytics and now machine learning, that same progression is underway in KM, with leading document management and enterprise search platforms being fortified with AI algorithms to surface the most salient templates for the task at hand. Even ostensibly old-fashioned or poorly-crafted searches can be enriched via metadata tags applied automatically by classifiers working in the background to categorize emails and documents as they are drafted. As with most things AI, the client ultimately benefits two ways: reduced cost and improved quality.
Raising client satisfaction through the purposeful use of AI is certainly one way to increase law firm profitability, but another is by reducing the mundane, non-billable time spent by attorneys entering—and by administrative personnel correcting—billing narratives. From suggesting time entries based on learned attorney behaviors to flagging erroneous billing narratives before a client’s e-billing system does, AI tools can pay for themselves (and then some) by increasing law firm realizations with each monthly invoice.
Needless to say, a full recitation of all potential applications of machine learning is beyond the scope of this article. We would be remiss, however, not to mention the extent to which AI can help in the compliance arena, including as it relates to the General Data Protection Regulation, or GDPR, which imposes a patchwork of obligations on companies offering goods or services to European Union data subjects. Among them are onerous requirements concerning data breach response and severe financial penalties for violations. Borrowing from medicine, the best treatment lies either in the prevention or early detection, and in this day and age, the no cybersecurity defense is complete without AI-enabled intrusion detection software. The GDPR also restricts the processing and cross-border transfer of personal data, and the same classification engines discussed in the context of knowledge management can make it far easier to locate and segregate GDPR-protected data for specialized treatment. GDPR aficionados may recognize the irony here given that the GDPR has become a major thorn in the side of AI providers trying to reconcile their machine learning “black box” algorithms with the GDPR’s principles of accountability and transparency, but that is the subject of another article.
Rumors of the demise of the billable hour and the automation of the legal profession have thus far been greatly exaggerated. Yet these and other uses of AI will only grow in prevalence, particularly as law firms and their clients shift resources to cloud providers who imbue machine learning in their core offerings. Accordingly, it serves us all well to get ahead of the curve while AI remains a differentiator as opposed to a table-stakes capability.