This dataset was created through an anonymous survey of solicitors in England and Wales, conducted between 12 November 2019 and 13 January 2020. Respondents answered a series of questions regarding their use of AI technology, as well as their training for and attitudes to the use of technology in their work. After discarding partial responses, the dataset comprises a total of 353 valid responses.The proposed research will explore the potential and limitations of using artificial intelligence (AI) in support of legal services. AI's capabilities have made enormous recent leaps; many expect it to transform how the economy operates. In particular, activities relying on human knowledge to create value, insulated until now from mechanisation, are facing dramatic change. Amongst these are professional services, such as law. Like other professions, legal services contribute to the economy both through revenues of service providers and through benefits provided to clients. For large business clients, who can choose which legal regime will govern their affairs, UK legal services are an export good. For small businesses and citizens, working within the domestic legal system, UK legal services affect costs directly. Yet unlike other professions, the legal system has a dual role in society. Beyond the law's role in governing economic order, the legal system is more fundamentally a structure for social order. It sets out rules agreed on by society, and also the limits of politicians' ability to enact these rules. Consequently, the stakes for AI's implementation in UK legal services are high. If mishandled, it could threaten both economic success and governance more generally. Yet if executed effectively, it is an opportunity to improve legal services not only for export but also for citizens and domestic small businesses. Our research seeks to identify how constraints on the implementation of AI in legal services can be relaxed to unlock its potential for good. One major challenge is the need for 'complementary' adjustments. Adopting a disruptive new technology like AI requires changes in skills, training, and working practices, without which the productivity gains will be muted. We will investigate training and educational needs for lawyers' engagement with technology and programmers' engagement with law. With private sector partners, we will develop education and training packages that respond to these needs for delivery by both universities and private-sector firms. We will investigate emerging business models deploying AI in law, and identify best practice in governance and strategy. Finally, we will compare skills training and technology transfer in the UK with countries such as the US, Hong Kong and Singapore, and ask what UK policymakers can learn from these competitors. To the extent that these issues are also faced by other high-value professional services, these parts of our results will also have relevance for them. However, the dual role of the legal system poses unique challenges that justify a research package focusing primarily on this sector. There are constitutional limits to how far law's operation can be adjusted for economic reasons: we term this second constraint 'legitimacy'. We will map how automation in dispute resolution might trigger constitutional legal challenges, how these challenges relate to types of dispute resolution technology and types of claim, and use the resulting matrix to identify opportunities for maximum benefit from automation in dispute resolution. A third constraint is the limits of technological possibility. AI systems rely on machine learning, which reaches answers by identifying patterns in very large amounts of data. Its limitations are the size of the datasets needed, and its inability to provide an explanation for how the answer was reached. This poses particular difficulties for law, where many applications require or benefit from reasons being given. We will explore the possibility for frontier AI technologies to deliver legal reasoning. The research will involve a mix of disciplinary inputs, reflecting the multi-faceted nature of the problem: Law, Computer Science, Economics, Education, Management and Political Economy. Working closely with private-sector partners will ensure our research benefits from insights into, and testing against, real requirements.

The survey was run anonymously using the Qualtrics platform. Invitations to participate were distributed by email to 10,000 randomly-selected solicitors. In order to increase survey participation, subsequent survey invitations were sent to under-represented groups of respondents. Further details of survey methodology, participant information, and the survey questions are included in the data documentation.