Legal services are traditionally provided by highly-skilled humans—that is, lawyers. During the past two decades, the costs of legal services have risen sharply but productivity has not. At the same time, technological progress with artificial intelligence (AI) has been dramatic. There is enormous optimism about the potential of AI as a general-purpose technology to deliver widespread productivity enhancements across the economy. What will be the impact of AI on the way in which legal services work is delivered, and can it provide a way to augment productivity? Much of the debate to date has focused on the capabilities of technical systems, asking ‘can machines replace lawyers?’ Yet while AI will surely render some roles redundant, humans will still be crucial for legal services for the foreseeable future. In our recent paper, we argue that understanding how AI will augment lawyering requires a focus on two inter-related aspects of these human dimensions: how the nature of legal services work will change, and how the firms that co-ordinate this work will be organized. We develop and substantiate our analysis with what is to our knowledge the most comprehensive empirical study yet conducted into the implementation of AI in legal services.
Where AI is deployed for legal tasks, it substitutes for humans. However, technical and economic constraints on AI’s deployment mean that more specialized and unique tasks will for the foreseeable future exclusively be performed by human lawyers, whose work will be augmented by the technology. Less obviously, AI will also create demand for new types of human role. We present the first empirically-grounded account of what these will look like in legal services. We draw on qualitative data from over fifty interviews with relevant professionals to detail the way in which AI is actually implemented. This requires a ‘pipeline’ in which the tasks to be performed must first be specified and the relevant data gathered and checked, before the system then performs its analytics; subsequently the output must be regularly reviewed by subject-matter experts. Most of these steps require human input from a range of different disciplines—including legal—working together in multi-disciplinary teams (MDTs). Delivering legal services through such technology pipelines therefore requires the assembly and management of MDTs in which some members have legal expertise. These ‘lawyers’ in turn serve to augment the AI system’s efficacy, as part of an MDT whose overall capability includes a range of other types of human capital, such as data science, project management, and design thinking. The work of these legal experts in MDTs is very different from traditional ‘lawyering.’
We also consider how these changes in legal work will impact the organizational structure of legal services firms. We present case studies of the implementation of AI in three different types of organization: law firms organised as partnerships, in-house teams in corporations, and alternative legal services providers (ALSPs) organised as companies. We find that law firm partnerships face unique challenges with the implementation of AI technology, not shared by the other organizational types. The core problem is with recruiting, motivating and managing the non-legal human capital needed to make the technology work. Non-legal human capital is hard for a law firm to recruit, as there is no obvious way for such persons to progress to partnership. And a management structure composed solely of lawyers is poorly suited to coordinating an MDT. Our cases studies of in-house teams and ALSPs suggest these problems are less intense in businesses organized as corporations.
We supplement these qualitative findings with quantitative analysis of a survey of practicing lawyers in England and Wales. We show that in a multivariate framework, respondent lawyers who work closely with non-lawyer professionals are significantly more likely to use AI applications than those who work exclusively with other lawyers, consistent with the importance of MDTs. We also find that respondent lawyers who work in law firm partnerships are significantly less likely to work with non-lawyers, and to deploy AI applications, than respondent lawyers who work in corporations.
Our findings have important implications for lawyers, law firms and the legal profession.
John Armour is the Hogan Lovells Professor of Law and Finance at the University of Oxford.
Mari Sako is a Professor of Management Studies at Saïd Business School, University of Oxford.
Richard Parnham is a Postdoctoral Research Fellow at Saïd Business School, University of Oxford.