Richard Susskind is an author, adviser and president of the Society for Computers and Law. His latest book, The Future of Law: Reflections and Predictions, revisits his forecasts three decades on. In this interview with Solicitors Journal, Richard discusses AI’s impact on law firm business models, the opportunities for smaller practices, and why improving access to justice may require fundamentally different legal services and institutions.
Your latest book, The Future of Law: Reflections and Predictions, revisits the predictions you made three decades ago about technology transforming legal services. Looking back, which developments have most surprised you, and where has the legal profession proved more resistant to change than you anticipated?
The first thing I should say is how different the world was in 1996. Only one in 25 people in the UK were online, and few lawyers had actually seen the World Wide Web. Few clients had received emails from their lawyers.
One thing I find humbling is what I call the “not yet invented technologies”. We didn’t have Google search, Wi-Fi, Wikipedia, social media, the iPhone or WhatsApp. We certainly didn’t have ChatGPT. It is impossible to anticipate the individual systems yet to be invented, and yet I think you can see the overall trajectory of change.
I try not to mark my own homework, but I think I judged the overall shape of the changes fairly accurately, but had little sense of the particular enabling technologies that would be devloped. Looking ten years ahead now, we cannot know what technologies have not yet been invented. But most people who work in AI think the most powerful, most impactful systems have yet to emerge.
With regard to the profession, I think I have always overstated the speed with which clients would demand fundamental change. Big corporate clients are under pressure to spend less on law firms and reduce internal headcount, and say they have more legal and compliance work than ever. Meanwhile, the big law firms are becoming more profitable. And yet general counsel have seemed reluctant to push for change.
As for law firms, I often joke that it’s hard to convince a room full of millionaires you’ve got your business model wrong. Why change if the model is working very well? The combination of current comfort and lack of client pressure has meant adoption hasn’t progressed as quickly as I expected.
AI is already changing how solicitors undertake research, drafting, due diligence and other routine tasks. But is the profession focusing too heavily on making existing processes more efficient, rather than reconsidering how legal services should be delivered in the first place?
I think there’s scope for both. One distinction I drew in the original book was between automation and innovation. Automation is computerising what we already do; innovation, in my terms, is using technology to allow us to do things that previously weren’t possible.
The story of legal technology since the late 1950s has essentially been one of automation, of grafting emerging and existing technologies onto current and past ways of working. Although AI has enormous innovative and transformative capabilities, I think over the next two or three years its dominant use will still be automation: looking for efficiency and productivity gains. People are spending on these systems and want a return on their investment.
That’s why I often say people are overstating AI’s impact in law in the short term, but understating it in the long term. We are now seeing clients recognising a way to reduce the cost of legal service while retaining the level of service they require.
The traditional law firm model depends heavily on billable hours, leverage and the development of junior lawyers through routine work. As AI increasingly takes on some of these tasks, how sustainable is that model, and what might a fundamentally different law firm look like?
There’s a deeper question: how advanced is AI going to become? Will we see modest progress or explosive progress?
What I call “model one” suggests lawyers and law firms will continue broadly as they always have, as a professional advisory service charging for human time, but with a very powerful new tool at their disposal. Important questions arise. For example, if time is saved and junior lawyers aren’t used, do the financial savings accrue to the client or the firm?
With Model 1, there’s no fundamental transformation. In contrast, “Model two” presupposes explosive advances and assumes we get to artificial general intelligence, or near AGI: systems that match human performance across the full range of our cognitive tasks. If AGI is achieved, most lawyers’ work could then be done by AI systems, or at least the systems would be capable of doing most work. We move from lawyers supervising AI output to relatively autonomous systems operating without human supervision at operating at human or above-human level. Questions about training young lawyers and whether hourly billing works dissolve because we’ve moved into an entirely different era where human labour is no longer at the heart of legal service.
The advance of technology, however, doesn’t necessarily lead to its adoption. For regulatory, cultural, commercial or ethical reasons, even if we have AGI, law firms may not embrace it. But others – such as start-ups – may well. I often say to law firms: the competition that will beat you won’t look like you. I think clients will show no loyalty to the traditional law firm way of delivering services if their required outcomes can be delivered at lower cost, more efficiently and conveniently.
The professions don’t occur in nature like rivers and trees. The modern legal profession emerged after the Industrial Revolution as a way of sharing scarce knowledge and expertise in a print-based industrial society. We shouldn’t be surprised to find entirely new ways of sharing that expertise in an AI-based digital society. Mere adaptation of our current practices and institutions may not be sufficient to support these new ways.
We don’t need to predict whether the future will be model one or model two, but surely we need to plan for both. None of us knows what the future holds.
Much of the discussion about AI centres on large commercial firms with the resources to invest in new technologies. Do you think smaller and high-street practices will ultimately benefit from AI, or could technological change widen the competitive gap between different parts of the profession?
I think AI hugely empowers smaller firms and enables them to undertake work they couldn’t have done on their own in the past. It offers the equivalent of a team of associates or researchers, enabling them to scale up and operate at speeds that weren’t previously possible. There’s a great equalising effect.
You may say that the more expensive systems are available only to very large firms. But I think many general-purpose AI systems will enable smaller firms to be far more productive and efficient, and to extend their reach.
Smaller firms can turn around their strategy and direction in a weekend. Major firms often suffer from “supertanker syndrome”: they find it very hard to change direction, very hard to stop. Introducing AI is a major technological, cultural and training endeavour. A handful of solicitors can decide as a group to embrace it very rapidly.
That’s especially important because the technologies themselves are changing quickly. New versions and types of systems can be embraced by smaller firms more rapidly. This departs from some of my previous writings, where I was rather more pessimistic about smaller firms. We’ve all been surprised by the latest AI systems’ power and capability.
AI promises to make legal services more accessible and affordable, yet there are concerns about accuracy, accountability and the quality of advice. How far could AI realistically help address the unmet legal needs of individuals and small businesses, and what safeguards would be essential?
I think when we get into the 2030s, AI’s massive impact in law will be in empowering non-lawyers to do legal work for themselves. This is how we crack the access to justice problem, making legal service more affordable and available, at non-lawyers’ fingertips. The extent and pace of this are up for discussion.
We have to be careful not to overstate these systems’ inaccuracy. They’re improving rapidly, and many formal studies that highlight their errors relate to systems that are now a couple of years old.
Humans make mistakes as well. We shouldn’t compare AI with ideal humans, but with humans who make mistakes and with situations where there’s no access to human lawyers at all.
We need to distinguish between unmet legal need and poorly met legal need. For unmet need, the threshold isn’t especially high. I think there’s already a benefit from general-purpose systems that give people a rough sense of their rights and entitlements in ways that weren’t previously possible or affordable.
If you assume, as I do, that these systems are becoming increasingly capable, then they’re going to become increasingly reliable. But we do need ways of evaluating them. I’ve been doing a lot of work in AI in medicine and healthcare, where there are detailed regulations govern new devices coming to market and whose performance is appraised once they’re in use.
We don’t have formally agreed benchmarks in law against which to assess emerging AI systems’ reliability and compare their performance with humans. A great joint project would be to produce agreed standards and methods to assess the suitability and reliability of legal AI systems being brought to market for consumers and citizens.
As AI systems become more capable of producing legal advice and performing tasks traditionally undertaken by solicitors, how should professional regulation evolve? Are existing approaches to competence, confidentiality, liability and professional responsibility sufficiently equipped for this transition?
Again, it depends whether you’re envisaging model one or model two. In model one, people still go to lawyers for proper legal advice, and these lawyers use their judgment, knowledge and experience, with AI helping in the background. Fundamentally, the service hasn’t changed. Much of the regulation and ethics governing legal practice would remain as they are.
However, if AI systems autonomously dispense legal guidance to people who would otherwise have no help, we need a whole new set of precautions to protect users. This is a new category of legal support. How we regulate, contain or control these systems hasn’t been considered sufficiently.
Most debate about controlling AI in law has concerned its use by lawyers. The far greater impact, I insist, will be as a general-purpose everyday tool anyone can use. What do we do as a profession to ensure that AI-based guidance for non-lawyers is reliable and systems have been sufficiently road-tested?
We’re not yet having sufficient debate about these entirely new ways for people to understand, assert and enforce their entitlements. Discussing AI for direct use by lawyers is necessary, but it really isn’t sufficient.
You have argued that technology could transform not only legal practice but also the administration of justice. What changes would you most like to see in the courts and dispute resolution over the next decade, and what are the biggest obstacles to achieving them?
If you’re a model one thinker, you’re thinking it would be good if judges, court administrators and litigators used AI to speed up the traditional judicial process and make it more affordable. This makes sense.
But we’re already seeing public dispute services – tribunals, for example - struggling with self-represented litigants who have AI at their fingertips. People can now generate credible and plausible legal arguments, but these systems can also generate huge amounts of irrelevant material. We need standards and protocols to help here. We don’t have these yet.
I’m often asked how our court system will cope. The answer is that it won’t cope with this new world. We’ll need new forms of public dispute resolution mechanism, themselves underpinned by AI, to cope with a world in which most citizens have ready access to these systems.
We shouldn’t be surprised that we need new public institutions. We have to think deeply about what dispute mechanisms would be appropriate in the AI age.
In 2014–15, I chaired a Civil Justice Council group looking at online dispute resolution. We recommended the idea of an online court, and the idea was strongly backed by (now) Lord Briggs. We envisaged an entirely different form of dispute resolution delivered by the state. In the event, that wasn’t followed through. We should be dusting off some of that thinking and updating it in light of recent technological advances.
Looking ahead to 2035, what do you think will distinguish the solicitors and law firms that thrive from those that struggle to adapt? And what should practitioners be doing today to prepare for a profession that may look radically different from the one they entered?
If you believe in model one, it will be those firms that most imaginatively, creatively and pervasively embrace technology in their daily practice.
If you believe in model two, firms will have to pivot from being one-to-one consultative advisory providers to some kind of AI-based system development providers. I know that’s a huge leap, but the legal AI systems that empower clients won’t spontaneously arise. They’ll need to be developed.
The new role for tomorrow’s legal businesses, for a while at least, will be to build the systems that replace our old ways of working. This is where I think the great opportunity and excitement, and perhaps even the obligation, lies.
These businesses will require knowledge engineers, design thinkers, data scientists, risk managers and system developers: people designing, building and maintaining the systems that replace our old ways of working. I call them tomorrow’s lawyers.
Copyright & permissions
-cropped-uoskilsj.jpg&w=3840&q=60)
