Reshaping the Banking Experience
Service Sourcing: The Deconstruction of Legal Technology Service...
To Digitize or not to Digitize!
eDiscovery as the CIO
Adopting Artificial Intelligence in the Provision of Legal Services
By Jaromir Savelka, Data Scientist, Reed Smith LLP and David R. Cohen, Partner, Reed Smith LLP
Impediments to the Adoption of AI and How to Overcome Them
There have been at least three primary impediments to the adoption of AI in the law:
• Legal Services Are Very Complicated: Much of the advice lawyers give not only requires careful analysis and judgement, but also is extremely tailored to the specific facts and circumstances in each situation. The creation of an AI system that could model and/or replace expert human judgement requires either the definition of exact rules experts follow when making decisions (expert systems), or a data set of past experiences and outcomes from which the system “figures out” rules— (machine learning). Some AI systems rely on a combination of both rules and machine learning. The development of precise rules can be expensive because it requires considerable time investments from highly-skilled and knowledgeable human experts—and sometimes those experts themselves cannot identify or properly weigh all of the factors that go into their subjective decision-making. Conversely, learning from data often requires sizeable data sets of recorded experience that may not be readily available. A big part of the answer is to focus AI development on legal services where there are clearly defined tasks and data for the AI to build off of—examples include technology assisted review of documents (“TAR” or “predictive coding”) for relevance to particular litigation or litigation issues, and technology assisted contract classification or drafting.
• Profit Incentives Have Been Mixed: In any profession it can be difficult to get workers excited about developing the technology that can theoretically replace them. Most outside lawyers still get paid by the hour, and developing technology to replace lawyer hours can make law firms less profitable rather than more profitable, at least in the short-run.
However, that may be changing as companies continue to demand more value from their outside lawyers and push for pricing models beyond the traditional hourly rate. Industry expert Kevin D. Ashley, Professor at the University of Pittsburgh’s Law School and Graduate School in Intelligent Systems, notes that that technology will enhance, not replace, human lawyers in the foreseeable future. See Ashley, K.D., Artificial Intelligence and Legal Analytics: New Tools for Law Practice in the Digital Age (Cambridge University Press, 2017). Likewise, innovative law firms are starting to realize that, if they can “build a better mousetrap” that delivers more client value, they will attract more client work, and ultimately earn greater profits.
• Technology Vendors Cannot Practice Law: In some other fields, technology developers have been able to fully understand certain services that humans previously provided and replace those services with AI. In law, however, there are professional restrictions in the U.S. that prohibit non-lawyers (or even lawyer employees of any company or firm that is not 100 percent lawyer-owned) from providing legal advice. Although some have argued that those restrictions are anti-competitive, and should be relaxed or eliminated, the truth is that legal services do tend to be complicated and it is difficult for non-lawyers to develop and apply the technology without a lot of help from practicing lawyers. A case in point is TAR. Despite being the biggest AI success in law to date, it may also be the biggest disappointment—even today, more than a decade after the first TAR programs were developed, most litigation document review is still manual. One reason is the lack of an easy and universally-applicable or accepted workflow for using TAR. Consequently, some companies that have tried to use it have found that, with the concomitant protocols, negotiations with opposing parties, and potential need for judicial oversight, trying to use TAR in place of human review can end up being more expensive than traditional key word filtering and human review. No company should ever try to apply TAR in place of human review without the close involvement of lawyers who are knowledgeable about TAR.
A Brighter Future for AI and the Law?
Despite the above impediments to large-scale adoption of AI in the legal industry, we foresee a brighter future. A number of law firms and in-house legal departments are now teaming with software developers and vendors to find new ways to use technology to deliver legal services better and less expensively. A handful of law firms have even launched technology subsidiaries devoted to developing technology tools for the provision of better and/or less expensive legal services. One example is Gravity Stack, Reed Smith's technology subsidiary, and their partnerships with Heretik and LegalSifter to expedite contract review and analysis employing AI. Other examples include programs that can help budget, optimize and report on litigation document review, analyze and provide initial advice in regard to data breaches, assist with due diligence project management, and even programs that can anonymize or pseudonymize documents to assist with GDPR compliance or other protection or redaction of confidential data. While not all of these tools utilize machine learning, each of them leverage technology to reduce time, reduce costs, and enhance quality of the legal services provided.
Based on the above developments, and continuing advances in the science and technology of AI, you can expect that the next decade will be far more exciting and transformative than the last decade has been with regard to the growing development and adoption of AI for legal applications.