AI and The Legal Profession

Working in the legal profession for well over a decade gave me a pretty good insight into the mindset of others in the industry – or, more specifically, their attitude towards change and innovation. What follows is necessarily a generalisation to some extent but I believe the observation remains no less valid because of that.

I vividly remember giving a talk as a trainee back in 1999 on why the invention of Napster’s file-sharing growth was so important. Most of my colleagues at the time looked on with barely-disguised expressions of confusion and boredom. It wasn’t the first time that I felt that my attitude towards adopting and experimenting with change made me feel very different to others within the industry over the course of 13+ years.

I was always a fan of Richard Susskind’s work over that time. For a man who predicted in his book “The Future of Law” in 1996 that email would become the predominant form of communication between lawyers and their clients (provoking a response from the Law Society of England and Wales along the lines that Susskind shouldn’t be allowed to speak in public because he clearly didn’t understand the way that the industry functioned and or the rules surrounding client confidentiality), he’s continued to push the industry kicking and screaming into the modern era throughout.

I just watched a talk that he gave over the summer in which he gave a 50 year view on the impact of AI and the law.

Susskind starts by setting out the four stages of resistance that he inevitably sees from members of the legal profession when faced with technological progress:-

  1. This is worthless nonsense
  2. This is an interesting but perverse point of view
  3. This is true but quite unimportant
  4. I have always said so.

And it was this first stage that reminded me of that Napster talk all those years ago. It’s reminiscent of Gandhi’s quote that gets bandied around frequently in Bitcoin circles when someone brings up the usual adoption hurdles:

“First they ignore you, then they laugh at you, then they fight you, then you win”

Susskind’s history is interesting because he actually built a so-called Expert System with a leading expert lawyer back at the end of the 1980’s – basically transferring a human’s knowledge and expertise into a computer system for others to use. Not an easy call to take “a dense web of barely intelligible interrelated rules” and turn that into 5 1/4 inch floppy disks. But the end result was a system that would ask you a series of questions before giving you an answer.

Then on 6 August 1991, the web happened. But still the law firms didn’t cotton on to the fact that the world was changing. And of course, why would they? When your business is built on an hourly billing model, what possible use could you have for an Expert System that reduces a process that usually takes 10 hours down to 10 minutes?

But of course the signs are now undeniable and change is inevitable. As he points out, this key paper from 2011 points out that in terms of initial document reviews, intelligent search systems can now outperform junior lawyers and paralegals. And remember – that’s the worst that that technology is ever going to be in the future.

I’ve always been drawn to Susskind’s simple argument which goes along these lines: following Moore’s Law, the average desktop computer in 2020 will have more processing power than the human brain. And in 2050, the average desktop machine machine will have more processing power than the whole of humanity put together. So it might just be time for the legal profession to accept that change is coming. It just cannot be that the internet, computer science, natural language processing, speech recognition, big data, intelligent inference, machine learning, speech synthesis and so much more is transforming every single corner of society and yet somehow this effect will not extend into the legal profession – which after all is, of course, one of the most information- and document-intensive professions in the world!

His conclusion is that by the 2020’s, we’ll have legal IT systems that are not modelled on brains (i.e. we’ll move away from modelling AI based on human intelligence alone), fuelled by brute force computing, utilising speech recognition, with real-time machine language translation, natural language processing, an ability to discern otherwise hidden legal risks through the analysis of big data, perfect search and a mixture of deductive, inductive, analogical and lateral inference.

Face-to-face legal consultations will become the exception rather than the rule and “communities of legal experience” will develop – networks within which ordinary people who have consulted lawyers or solved problems themselves will share their experiences with others who want to access that knowledge.

And yet, the majority of lawyers are still in a state of denial. Most believe that the current state of the industry represents little more than a temporary blip in the standard state of affairs before things return to normal, with an economy similar to the one that existed before 2007. Yet whilst some more successful ‘firms’ are looking at the disaggregation of legal work (using paralegals, offshoring, on-shoring etc.), the real disruption will come over the next decade when technologies will be able to do the work that we originally thought could only be done by “intelligent human beings”.

Of course, it’s very easy to criticise from the outside. Which is why innovation often comes from elsewhere. And, with technology, hindsight is always 20/20. But disruption is a certainty. In a world in which Google’s stated aim has always been “to organise the world’s information and make it universally accessible and useful”, the information that provides the foundation of value upon which the profession is built is gradually being made free. I can’t wait to see where we get to once AI really starts to kick in.

It’s not hard to imagine a demand for having IBM’s Watson as an app on your mobile dishing out legal advice whilst it also saves your life, is it?


One Pingback/Trackback