Agentic AI and justice: why human judgement must remain central

Artificial intelligence that can carry out tasks on someone’s behalf could change how people access legal services and how the justice system operates. However, the Law Society of England and Wales has warned that the technology is advancing ahead of the safeguards needed to oversee it.

Its foresight study, The Future of agentic AI in the Justice System, examines the opportunities and challenges presented by increasingly autonomous systems. The central concern is whether clear boundaries around responsibility, oversight and human decision-making will be established before these tools become widely used.

Efficiency cannot resolve every problem

The attraction of automation is understandable in a justice system facing substantial delays, with some hearings already scheduled for 2030. AI could help with administrative workloads and make it easier for people to navigate legal processes.

Yet the Law Society cautions against treating technology as a solution to longstanding structural problems. Inadequate funding, unreliable data and processes that still depend on paper will not disappear simply because an AI system is introduced. Automation could instead carry those weaknesses into new workflows.

There is also a more immediate pressure: people are already using AI to help them understand proceedings, and incorrect or misleading information can leave courts with additional work to untangle.

The risks extend beyond inaccurate answers

When a system can act as well as generate information, a mistake may affect several stages of a process before anyone notices. Responsibility can also become difficult to trace where different people, organisations and technologies contribute to the outcome.

Uneven access is another concern. Those able to afford sophisticated tools and the expertise to use them may gain advantages over people without those resources, potentially widening existing inequalities in access to justice.

The study also raises questions about decisions being treated as administrative tasks when they have previously involved judicial judgement. If that shift happens through the design of automated processes, it may become less visible and more difficult to challenge.

“Legal judgement cannot be delegated.”

Brett Dixon, vice president of the Law Society of England and Wales

Dixon emphasised that professional expertise remains essential to reviewing, challenging and approving AI-assisted work. He also made clear that potential administrative savings do not remove the need to invest in court buildings and staff.

What safeguards is the Law Society seeking?

Used with appropriate controls, agentic AI could allow judges and practitioners to spend more time on matters requiring their expertise. However, the Society considers existing regulatory arrangements insufficiently adapted to this technology.

Why this matters to will-writers

Although the study focuses on the justice system, it raises relevant questions for will-writing and estate planning. As a practical reflection on its findings, practitioners considering AI should look beyond the time a tool might save and examine how its work will be supervised.

Useful questions include what the system is permitted to do, where professional approval is required and how errors would be identified before they affect a client. Clear records of checks and decisions can help explain how technology contributed to the work.

The wider message is that the boundaries need deliberate attention. Deciding which tasks can be supported by AI, and which decisions require human expertise, should form part of adopting the technology.

Sources: the Law Society of England and Wales, research report and accompanying announcement. The section on will-writing offers practical reflections on the findings.