Dixon: Legal judgement cannot be delegated
Agentic AI could become the “default doorway” into the justice system, “identifying legal problems before people recognise them, explaining rights, suggesting actions and pursuing processes on their behalf”, the Law Society has predicted.
Researchers said this could represent either the “largest expansion of legal capability in history” or justice could become “something people subscribe to, or something they are excluded from”.
The legal system could move from human decisions assisted by machines to machine actions supervised by humans.
“As agentic systems begin drafting claims, negotiating outcomes, communicating with institutions and making choices about next steps, the question of responsibility becomes increasingly complex.
“The challenge may no longer be deciding whether AI made a mistake, but understanding where human accountability begins and ends when decisions emerge from a chain of human and machine actions.”
Though agentic AI was not yet deployed in the justice system, “the conditions for its emergence are beginning to develop” in the coming years.
The Law Society reviewed developments in China, the US, Estonia and the UK, which they combined with interviewing 19 people “at the leading edge of AI in the legal sector” and 11 experts in the justice system for The future of agentic AI in the justice system: implications for courts, tribunals and alternative dispute resolution.
The report predicted that the “new gatekeepers of justice” may not sit in courts or law firms, but inside technology platforms.
“If commercial AI systems become the main route through which people understand their rights and navigate disputes, access to justice could become shaped by private incentives, business models and invisible design choices.
“The future battleground may not be whether people have legal rights, but whether they have meaningful control over the systems that interpret and activate those rights.
“The next decade may determine whether agentic AI becomes the greatest expansion of legal capability in history, or creates a new layer of invisible gatekeeping between people and justice.”
The report said the back office of justice could become its “hidden brain”. While, for generations, administration has been treated as the invisible machinery that supports justice, agentic AI could transform it into “an active intelligence: predicting demand, prioritising cases, allocating resources, identifying patterns and determining pathways through the system”.
It went on: “The most influential decisions about how justice operates may no longer happen in courtrooms, but inside automated systems that few people see, understand or can challenge.”
The Law Society also warned that AI agents could create a “self-driving justice system” as they connected across judicial case management, scheduling, communication, data analysis and resource allocation to provide “a court system that can respond in real time”.
Agentic AI introduced the possibility of judicial conclusions reached through a chain of human and machine reasoning. “The critical question may not be who signs the judgment, but who – or what –shaped the pathway that led there.”
If AI systems routinely identified authorities, framed legal questions, ranked arguments and suggested interpretations, they might begin to influence “not only the conclusions judges reach, but the intellectual landscape” within which conclusions are formed.
“The greatest transformation may not be machines making decisions, but machines quietly defining what counts as a reasonable decision in the first place”.
But the Law Society insisted that some parts of justice were “not automatable” – not because of capability, but because of legitimacy.
“Deciding contested facts, assessing credibility and exercising irreducible discretion are not just complex tasks – they are constitutive of justice itself. Removing the human decision-maker from these functions risks undermining the legitimacy of outcomes, even where accuracy is preserved.”
The report warned that legal regulation “may be trying to govern yesterday’s technology”, with many models assuming that humans remained the central decision-makers.
“But if future systems operate through complex chains of action that no individual fully understands, traditional oversight may become less a safeguard and more a comforting assumption.
“The challenge may be developing forms of governance that can supervise autonomous systems rather than simply review their outputs.”
Researchers warned that accountability “may arrive after harm rather than before it”.
They explained: “The first major test of agentic AI governance may not be a carefully designed framework; it may be a high-impact failure that exposes unresolved questions about who was responsible, who was harmed and who had the power to intervene.”
One of the few areas “where the technology aligns cleanly with system need” was in handling high-volume, low-discretion processes – document handling, triage, scheduling and post-decision execution – to reduce delay and free constrained human capacity.
The Law Society said it was calling for clear accountability for agentic AI in justice, transparency of AI decision-making, “meaningful human oversight”, and public debate before widespread use.
Vice-president Brett Dixon commented: “Legal judgement cannot be delegated. Not in courts and not in the legal sector. Professional expertise that shapes, checks, questions, improves and approves AI outputs is needed now more than ever.
“The most important question becomes not what AI can do, but what should remain human. There is still time to shape the boundaries of AI before the market does it for us.”
Leave a Comment