There is a 3.5 percent chance that lawyers’ jobs will be automated. That statistic seems appealing to those of us in the profession—especially relative to the chances for other “skilled” professions like financial advisers (58 percent of automation) and accountants (94 percent). However, this figure does not stand for the proposition that lawyer’ jobs will remain unchanged as the tendrils of artificial intelligence (AI) wind their way into jobs once thought too complex to be done by machines.
Legal work is diverse. There are several categories of work, from litigation to transactional, from bankruptcy to investigations. Each brand of job entails routine work like document and contract review, but also cognitively complex tasks such as oral argument and negotiation. And it is the former type of work that will be first in line to be handled by a computer.
Indeed, efforts in this space are already advanced. There’s LawGeex, a software company that handles contract review and approval that would otherwise be done by in-house counsel. LawGeex advertises substantial time and cost savings of 80 percent and 90 percent respectively. J.P. Morgan recently announced that its proprietary software COIN (short for “contract intelligence”) is now reviewing commercial loan agreements, saving 360,000 hours of legal work per year all while reducing mistakes. And then there’s ROSS, which uses machine learning to perform legal research. The software uses keyword analysis to sort among a database of documents and case law to deliver attorneys the answer to questions they’ve asked.
Notice that these are exactly the type of tasks you’d expect to be automated: time-intensive, often repetitive work. In the large firm context, lawyers had already begun to receive pushback from clients for the high rates they’d bill while doing tasks like legal research. “A lot of clients have refused to pay for the time spent doing research,” says Andrew Arruda, a cofounder of ROSS. But as more work is being done by AI attorneys like ROSS, doesn’t that mean less work to go around and fewer attorneys? Is the legal industry the next manufacturing sector?
Maybe.
LawGeex offers an initially compelling analogy that automation in the legal sphere is like autopilot for airlines: while pilots no longer need to do much of the grunt work in flying airplanes, we certainly still want just as many pilots in the cockpit.
But don’t be so sure that the legal industry is congruent in this sense. The trend in the legal industry is to find alternative ways to complete lower level work so that attorneys can focus on higher level tasks like the ones mentioned above. Many large firms had already begun to take lower-level work from high-billing associates and assign it to contract attorneys who cost far less to employ. The contract attorney arrangement, which offers lawyers more flexible schedules in exchange for lower salaries, seems to be working well. But the diligence and document review work being done by these lawyers will likely be eaten up by automation as the technology becomes more sophisticated.
The movement toward the contract attorney arrangement likely explains much of the continued divergence in salaries in the legal industry. Due to software like COIN, however, it no longer takes an army of attorneys to review hundred-page long contracts; and due to software like ROSS, it no longer takes hours or days to perform legal research. These tools are necessarily freeing up time and manpower for other tasks—but the amount of legal work is finite. As more ground-level work is eaten up by AI, lawyers can expect to see the number of reasonably paid, flexible-work-arrangement positions like document review attorneys shrink, and the bimodal salary distribution widen.
In addition, AI software is pushing the boundaries of what was once thought possible. The team behind ROSS is testing a feature that allows ROSS to actually write legal memos, a task traditionally assigned to law firm associates. And J.P. Morgan anticipates using COIN to help interpret regulations and analyze corporate communications. These developments suggest that not even jobs that require complex thought, creativity, and skills like reasoning and persuasion are safe from the slow and steady advance of automation. Much like autopilot in the airline industry, these software platforms will certainly still require review and operation by attorneys—but the question is how many lawyers will be needed.
It’s not all doom and gloom. Automation has the potential to free up lawyers’ time without reducing the need for lawyers at all in certain areas. Federal courts, for example, are notoriously backlogged. Some commentators agree that the automation has the potential to shorten dispute resolution times, and even pass along some of the cost savings to individuals who can’t afford high-quality legal services.