The long awaited tech disruption of the legal sector
Artificial intelligence, specifically machine learning, is not a new topic. IBM’s Deep Blue beat world chess champion Garry Kasparov over a series of chess matches in 1997. Since then, artificial intelligence has drastically improved in performance and is utilized across different fields, from matching riders with drivers on ridesharing apps to predicting criminal activities in urban areas.
Many people see legal work as the final frontier for machine learning because much of the work requires an understanding of complex human language. The challenges of natural language processing were highlighted when IBM’s Watson responded “Toronto” to a question on “U.S. Cities” on the game show “Jeopardy” in 2011. Such challenges are made more complex in law by seemingly long and convoluted drafting, sometimes mired with old English and Latin phrases.
However, things have rapidly changed over the last few years, driven primarily by increasing advance of technology. By way of illustration, a recent demonstration by LawGeex, an automated contract review service provider, showed that its machine-learning AI algorithm beat 20 experienced lawyers in both time and quality in the reviewing of non-disclosure agreements, a common task among junior lawyers.
In response to disruptions from legal tech and increasing demands from clients, Clifford Chance, an international law firm with one of the most pre-eminent legal practices in the world, is one of the first traditional “BigLaw” law firms to experiment with the provision of AI augmented legal service.
To “make sure the best ideas from legal and operations are coming forward and able to percolate,” Clifford Chance appointed Bas Boris Visser, the firm’s Amsterdam managing partner at the time, as Global Head of Innovation and Business Change in 2015. He was charged to “anticipate, respond and adapt to key factors shaping the legal industry.” Since then, Clifford Chance has started experimenting with and deploying a number of new machine learning software to support its lawyers’ day-to-day jobs. The most notable examples are Clifford Chance’s partnership with Kira Systems and the launching of Dr@ft.
Kira is a “machine learning software that identifies, extracts, and analyzes text” in documents. It came with some “knowledge” out of the box (with little set-up time) and it was further trained by specialist lawyers at Clifford Chance for over one year “to deliver more effectively against the specific requirements of Clifford Chance’s clients.”
Junior associates and paralegals spend most of their time sipping through thousands, even millions, of documents as part of the legal due diligence exercise in an M&A transaction or the discovery process in a litigation proceeding. Kira is most effective in such work, whereby high volume of documents are reviewed under a short period of time. An internal research found that Kira could save about 20% of the team’s time in an acquisition due diligence exercise involving more than 4,000 documents, even though the team chose to check Kira’s findings manually. Under the traditional hourly billing model, such savings on lawyers’ time (and therefore legal fee) can be significant to clients.
Dr@ft is an automated document assembly system developed internally by Clifford Chance. It enables clients to generate contracts quickly and automatically within Clifford Chance’s private cloud after answering an online questionnaire. Dr@ft adds value by saving up to 85% of lawyers’ time, promoting consistency across documentation, and reducing burden on clients’ legal and compliance teams.
To remain competitive in the increasingly crowded legal sector, Clifford Chance must continue to improve its AI tools (through supervised learning) and adopt new tools where appropriate. Despite the lead time to train Kira, it is not difficult for other law firms to replicate Clifford Chance’s success. The basic Kira software is not proprietary to Clifford Chance and is open to partnership with other law firms. Other similar machine learning software, e.g. LawGeex and eBrevia, exist in the market, too.
Looking into the future, as more law firms adopt Kira (or similar software) in their analytical work, the legal landscape will look commoditized and undifferentiated. Also, as AI tools become more intelligent and refined, questions will arise as to whether clients can directly acquire the necessary legal services from the software developer. As such, Clifford Chance should try to acquire Kira or at least strike an exclusive deal with Kira to make it proprietary to the firm. Clifford Chance should also integrate Kira, Dr@ft, and all future AI tools more closely with the rest of the firm and consider how to price future work where AI tools (with heavy upfront cost but little variable cost) play a heavy role in the overall service provision.
In addition to the pricing dilemma, it is also worth considering how the workforce of Clifford Chance will look in the future: What role will junior lawyers play? Will the firm employ more coders than lawyers?
[791 words]
Geek Squad- thanks for the interesting take on machine learning in the legal context. I think you are correct in that applying machine learning to the relatively simple, repeatable, and time-intensive task of reviewing the due diligence paperwork during the initial phase of a legal proceeding is a solid use case. I also think you bring up an interesting point in the future composition of the legal workforce as a result of the integration of machine learning. Perhaps it will be a zero-sum game where coders are selected instead of junior lawyers at the entry level; or, perhaps a select number of developers can be integrated into a law firm to reduce the grunt work that these junior lawyers need to muddle through, thereby freeing them up to be allocated to developmental tasks typically reserved for more senior lawyers.
It will be a big challenge for legal professionals as the article mentioned. Law firms in the future will have to consolidate, hire fewer junior lawyers but still groom them to become partners and employ more technology in order to stay competitive as clients will continue to be more demanding about the cost they pay for legal service.
I wonder if we will soon see cross-training of lawyers and coders – for example, will junior associates differentiate themselves from other junior associates (thereby making themselves more valuable to potential employers) by becoming familiar with this technology? Separately, you note that the team ‘chose to check Kira’s findings manually.’ Given the high stakes, how will law firms become more comfortable engaging with this type of technological disruption, so that they can realize the cost savings of delegating this work to Kira-type platforms?
Nice piece Geek Squad. I agree that the lower level work currently done by junior lawyers is definitely susceptible to the AI technologies mentioned here. Regarding your point on law firms becoming commoditized and undifferentiated, I expect that firms will turn to emphasizing the expertise of their senior lawyers. However, that brings up the question of whether we will soon see a dearth of senior lawyers given that the junior ones are being displaced. I’m not sure how much expertise developed by senior lawyers is dependent on the the rote work they did in their early days, but if that is indeed a path they traditionally have to go through, then I would question what ways we can bring them up to speed now that the rote work will soon be replaced by machines.
Thanks for the interesting read Geek Squad. I think you lay out a clear reasoning on why AI will become important for the legal sector as well as a compelling story on how Clifford Chance has responded to this evolution. I have two main questions when reading this paper.
(1) How far will this go? What can be automated and what not?
I read an interesting article in the New York Times where the main take-away was that AI will lead to efficiency gains, but will not change the fact that clients still want to be guided by a trusted advisor: “For the time being, experience like mine is something people are willing to pay for.” [1]
(2) How will this impact the legal workforce? How many lawyers will we still need? And how will they be trained?
In many ways closely related to the previous question, the efficiency gains that are achieved will primarily hit junior lawyers – as you pointed out as well – because they are the ones doing the biggest chunk of the research and contract drafting. Interestingly, AI will not only replace part of these junior legal jobs, but might also make it more difficult for junior lawyers to be trained properly [2].
Happy to continue the discussion on both of these points.
[1] Steve Lohr, “A.I. Is Doing Legal Work. But It Won’t Replace Lawyers, Yet.”, The New York Times, March 19, 2017, https://www.nytimes.com/2017/03/19/technology/lawyers-artificial-intelligence.html, accessed November 2018
[2] Erin Winick, “Lawyer-Bots Are Shaking Up Jobs”, MIT Technology Review, December 12, 2017, https://www.technologyreview.com/s/609556/lawyer-bots-are-shaking-up-jobs/, accessed November 2018
Geek Squad, very interesting piece on the application of AI in the legal space. As I read your piece, I drew a similar analogy to the work of junior radiologists and pathologists in medicine. As AI is becoming more advanced in these spaces, there is a growing consensus that machines will be better equipped to read high volumes of imaging studies with less variability. That being said, the job of these physicians will not be replaced, but rather facilitated by machines. I see something similar happening at Clifford Chance where AI has facilitated the role of Junior Lawyers such that they will be spending less time reading legal documents. This will shift their work from the mundane to the complex; that is, spending more time interpreting the AI’s output. There are ethical and legal implications of shifting major responsibilities like M&A’s to machines. I believe a human will always be required in those transactions and AI is actually allowing us to spend more time on these more complex tasks. With that said, I don’t believe Junior Lawyers are at risk of being replaced, but rather, these positions will require a higher level of knowledge and training to be able to deal with higher-level analytical tasks.
Thank you for sharing, Geek Squad. I certainly see the benefits for a law firm that is looking to incorporate this technology in their organization. However, it does bring up the interesting point if lawyers are essentially cannibalizing themselves. At the junior level, moving to AI certainly raises the bar for hiring; the baseline for joining a firm is now what junior rainmakers were typically learning in the first few years. I see changes within the curriculums at law schools to reflect this. At the same time, there is no replacement for the reps junior lawyers receive in these early years so it becomes a slippery slope for the profession over the next 30 – 50 years. Overall, I don’t see coders as a threat to lawyers, at this time, AI is great for boilerplate agreements and grunt work. However, as a client of a complex transaction or case, I want to know that my lawyers know the ins and outs and can attest to anything that is shared with the other side.
Thank you for your essay – I find this touches many different fronts of AI including replacement of jobs in general and integration of AI into critical civil services such as medicine, law, and banking. My main concern is how the use of AI could propagate throughout the legal system. Would we allow AI to determine fair sentencing of crimes to remove bias of judges, for example? There are many issues surrounding the ethics of AI that still have not been solved or debated by academia let alone the general public. Just because we can do something does not mean that we should. For now, removing the burdensome tasks of paralegals seems great, but there seems to be little checks and balances on the speed with which this technology is spreading, and I would hope that people remain vigilant and prudent when it comes to AI applications, and always consider the long term ramifications of these changes to society.