AI LEADERS
Latest research: July 2026
This latest Briefing AI Leaders research, distilling views from 50 senior strategic leaders at UK law firms, suggests that agentic genAI has moved well beyond the theoretical stage and is rapidly becoming a practical focus. Unlike traditional genAI that relies on direct prompting, agentic AI is defined here as technology capable of independently planning and executing multiple steps towards a user-defined objective, using systems as necessary within set parameters, before a human reviews/approves the results.
Please reach out to [email protected] if you have any questions or suggestions relating to the project.
Join Briefing AI Leaders
Get on the mailing list for our exclusive AI research, hear about dedicated roundtables for senior leaders like you making decisions about AI investment, adoption and planning in your peer firms, and be part of the AI Leaders community.
Executive summary
Most firms on the path of agentic AI deployment, but uncomfortable with full autonomy and approval needed, say leaders
More than a quarter of AI leaders report their firms are now piloting some agentic solutions, while 13% say they have implemented these within production workflows. More firms are active in the space — with agentic AI somewhere on their roadmaps — than are merely exploring the concept, signalling growing confidence that autonomous AI capabilities can deliver value. At the same time, adoption remains uneven. Some leaders indicate widespread use or pockets of mature deployment, while others see activity restricted to specific functions or a requirement for formal approval before people become in involved in any agent-related initiative.
Current use cases reveal that firms are prioritising operational efficiency rather than highly autonomous client service applications. The strongest interest centres on document-intensive legal work such as review and drafting, together with knowledge management and business development activity. Risk and matter management are also emerging as promising areas for application. Examples of deployed or developing solutions include large-scale document review, automated extraction and comparison of contract data, credentials management in BD, website enquiry routing, and policy signposting for internal users.
Two-fifths of those at firms where agents are already being explored have also already built their own bespoke agentic solutions, and over another third are in the process of doing so. Only one in 10 reports no exploration activity here at all. Most firms are combining API integrations with emerging interoperability standards such as the Model Context Protocol (MCP), enabling agents to interact with multiple systems and data sources without extensive custom development.
Governance, however, is the top concern with such transformative work. When assessing agentic AI opportunities, leaders place substantially more emphasis on output quality than on efficiency gains. The legal sector’s cautious stance reflects the reality that agents may access more systems and carry out more complex actions than conventional AI assistants. Consequently, organisations are likely to be reviewing governance frameworks, strengthening oversight mechanisms, and considering new controls over both information access and permitted actions. A third of firms have already made governance changes specifically to accommodate these technologies, their leaders say, while a majority are actively reviewing their approaches.
Very few are prepared to allow agents to act without explicit human approval, but two-fifths of leaders accept the idea of limited autonomy within approved processes. However, almost as many say they would only be comfortable with AI gathering the information or drafting recommendations for what could be done (38%). Human supervision continues to be regarded as essential for effective risk management.
The research also highlights a broader trend towards AI maturity. Almost half of respondents (47%) are confident that their firms are competitive with peers at driving genAI adoption. Weekly use of approved tools is now widespread across many, with significant numbers estimating that a majority of employees use such tools regularly. Alongside adoption, more leaders than at the start of the year say their firms have a number of 'AI readiness' activities underway, including role-specific workforce training, performance/outcome-tracking, workflow redesign and documentation, and sharing of use cases, with particularly significant increases in the number that are apparently rethinking pricing models and skills planning for a future workforce where AI is a common tool in the toolkit. Hybrid ‘build and buy’ technology strategies appear to be increasingly common as firms seek to balance bespoke capability with the advantages of commercial platforms.
Overall, the findings suggest that the legal sector has entered a new phase of AI evolution. The debate is no longer whether agentic AI has relevance for law firms, but how to deploy it safely, govern it effectively, and scale the benefits while preserving the human oversight that legal practice clearly still demands.