AI and the Law

By nature, lawyers are very resistant to new technology and change. It took, for instance, a worldwide pandemic and the resultant quarantines, for Kenyan litigation to start utilizing video conferencing tools. But technology and innovation have finally caught up with the legal world


Over the last couple of decades, technology and automation have taken over large swaths of jobs and careers ranging from simple manual labour to advanced tasks. For a long time, this inexorable march of technology has limited its impact to blue-collar occupations that are simple to automate and outsource. 

For the last century or so, the legal world was safe from disruption by technology; despite changes such as telephones and computers, a lawyer in the modern-day argues before a judge in a manner not too different from his predecessors centuries ago. By nature, lawyers are very resistant to new technology and change. For example, it took a worldwide pandemic and an imposed quarantine for Kenyan litigation to start utilizing video conferencing tools. But technology and innovation have finally caught up with the legal world, or more accurately, artificial intelligence is here. 

To understand how new technology impacts and alters a particular field, it is essential to appreciate that either necessity or economics drives technological change. To use the Covid-19 pandemic as an example of necessity driving innovation, within less than two months lawyers and judicial officers in Kenya and globally were forced to learn and utilize videoconferencing to ensure that the judicial system did not grind to a halt. However, when necessity drives innovation, people tend to do the bare minimum needed to do their tasks successfully. 

On the other hand, when economics drive innovation, technology moves in and changes the field irrevocably. For a long time, technology was not good enough to allow economics to change the field of law. With menial tasks such as contract review, case research, transcription, and even filing, it was always easier to utilize interns, paralegals, and pupils simply because there were no tools available to do those tasks. However, once the tools emerge, technology slowly but steadily takes over. An example of this is the E-filing system introduced by the Kenyan Judiciary in 2020; despite the growing pains, law firms are slowly arriving at the realization that, unlike the previously tedious process that required physical visitation to the court case (which in turn created a need for interns and clerks to do those tasks), a single lawyer can file their matters from the comfort of their own office saving the money that would have otherwise been spent on clerks, transport, or even interns.

Artificial Intelligence is the peak of our current technology; the processes that took decades to change are now happening in a matter of months and years. Things that we previously considered impossible are now being achieved; experts went from “computers will never be better than humans at the board game GO” to having an AI program, AlphaGo being the reigning world champion in a span of two years. Similarly, AI in law is now accomplishing tasks that were previously considered out of abilities of computer programs. 

For the longest time contract review was a task for partners and associates who would go through an entire contract word by word to ensure that there were no loopholes, missing data and that they were compliant with the relevant laws and regulations. In the present day, there are thousands of legal AI software that with a single click can analyze a contract in seconds, identify risks and gaps in the contract and provide the lawyer with an outlined and a highlighted report. In addition, research indicates that the AI-assisted contract review has a greater rate of accuracy as compared to human only contract review. Research suggests that the AI algorithms have an accuracy rate of 95% in contract review far ahead of human lawyers who had a rate of 85%. 

Jurisdictions such as the USA have had AI tools being used in strategic decision making where they are provided access to a large corpus of cases, arguments, transcripts, rulings, and even personality profiles of judges, opposing counsel, and juries to expose patterns and biases. With this information, AI tools are able to guide the arguments and language to be used in the courtroom for the greatest impact and resonation with the judges and juries while addressing potential arguments by opposing counsel. AI tools are also being used to produce the first draft of legal briefs with the lawyer focusing on tweaking the document to develop a final copy.

As noted above, the primary driver of adoption of Artificial Intelligence in the legal world will be economics. When a law firm can deliver better results for its clients with reduced costs, then it is merely a matter of time for innovations to be adopted as industry standards. In fact, the main differentiator will be the time taken by a law firm to adopt technology vis a vis the general industry; the early adopters will have the opportunity to define themselves in the new world. In this world, most of the menial work in the legal world, such as endless filling, drafting, review, research, and even renewing will be taken over by technology. The large group of legal professionals who have carved a niche in these sectors will find themselves out of work. When a program readily drafts contracts, avails it for human review, and then automatically files it at the relevant court, a single lawyer can handle work that took an entire law firm. In a study published by the Richmond Journal of Law and Technology, a law firm was able to eliminate 100 workers and reduce costs by 95% by using an AI program to filter through thousands of hours of audio recordings as opposed to using people.

A sceptic could argue that the examples being provided are simply cases for improvement in efficiencies of legal work rather than a fundamental change in how the law is approached and done. Fair enough, let’s discuss the radical changes that might occur as a result of AI being used in the legal world. In a couple of decades, we might have AI algorithms “hearing” cases and making rulings. At present it is not inconceivable to contemplate a world where an AI is used to prepare judgements in court cases with the magistrates or judges role being relegated to review; it is arguable that the only reason this is not happening presently is due to the ethical and professional considerations rather than lack of capability. In fact, the American Bar Association has already considered using AI as a tool to guide the judges in research and navigating materials provided in the court cases. In jurisdictions such as Arizona, Kentucky, Ohio, and Alaska courts are already using AI to establish the risk factors of the defendants when determining bail. While a human judge makes the final decision, they are following the instructions of an algorithm. Sooner or later, someone will suggest using AI to make the judgements – a genuinely impartial, incorruptible, and a legally omniscient (all laws, cases, and opinions available) judge, perhaps with human “oversight” to satisfy the need for equity.

Even in the possible future where we avoid making AI our judges and juries, they will still be able to predict the outcomes of most cases. At the end of the day, everything can (and will) be reduced to a set of data points, and AI algorithms are phenomenal at absorbing data and making predictions. And again, the future is already here, in the United States, researchers were able to predict the outcome of the Supreme Court cases with 75% accuracy, which doesn’t seem impressive until you realize that legal exports and scholars had a 51% accuracy. And AI keeps learning and getting better at an exponential rate. In light of this, LexisNexis developed a litigation analytics tool, Context, for the US jurisdiction to guide lawyers on the judges, the possible rulings, and how to shape their arguments based on similar and past cases. Bloomberg Law, Westlaw Edge, and Gavelytics are also offering similar tools. At some point in the not distant future, lawyers will be reduced to reiterating AI-generated arguments for judges who might then be making decisions guided by AI. 

Since we have explored a future where AI is in almost all areas of the legal world, it is essential to understand the inherent dangers of adopting this technology. The main challenge with AI and the decisions it arrives at is that they are opaque in a term referred to as a black box. For example, when an AI arrives at a conclusion guided by over a billion data points, it is virtually impossible to determine the exact factors or data points considered and or how they were weighed to arrive at the conclusion. This is in stark contrast with the legal world that insists on transparency in every decision, especially in the judicial arena. Through the Ratio Decidendi and Obiter Dicta, we not only demand that judges make their decisions publicly but that they also explain how and why they made those particular decisions. A judge relying on AI will never be sure how exactly the AI arrived at its conclusion. Whether he agrees with that conclusion is immaterial; a judgement affecting another human being was made by a system other than the duly appointed judge without an explanation.

The second danger is subtler; where the decisions made by the AI, whether in drafting a contract or deciding bail are guided by lessons from the past examples that have been simplified into data points. This means that all the bias that has been inherent in the system gets enshrined into the AI and used moving forwards. For example, an organization that had chosen to use AI in lieu of HR for hiring IT personnel was horrified to realize that the AI was actively discriminating against women since IT staff were historically made both in the company and in the industry in general. This is a stark example, but the AI could discover subtle associations that are harder to spot and implement them in its decision-making strategies. For example, men are statistically jailed for a longer period than women even when all the other factors are held constant, an AI learning from such a system would adopt the bias and keep implementing it moving forwards.

The final danger is more nebulous and yet the one I consider to be the most important, allowing AI to take over the legal world, especially the judicial part is akin to surrendering our destiny as humans. While the challenges I raised above can (and most likely will) be addressed with time and more research, the issue of having software and algorithms deciding the fates of people in such a direct fashion is too great a surrender of a critical part of the Hobbesian social contract; that only humans should stand in judgement of other humans. Perhaps a new jurisprudential school of thought is in order to champion against having AI gaining too much influence or even taking over the legal world.   

Sign Up