Sir Geoffrey Vos has now argued twice in two months that AI judges will arrive in small disputes, and that money rather than principle will carry them there. The Master of the Rolls gave the John Lehane Memorial Lecture at the Supreme Court of New South Wales on 18 August 2026. He returned to the subject in his keynote at the Legal Geek conference on 14 October 2026, in his last three weeks in the post. His prediction is that people will come to accept, on economic grounds, machine-enabled resolution of small disputes at first instance. He holds the most senior civil justice post in England and Wales, so the prediction bears on our courts whoever he is addressing.
He drew a boundary around it. A machine-made decision, in his account, needs the free and fully informed consent of the parties, because the right to an independent and impartial human judicial decision sits deep in the constitutions he surveyed. Machine-assisted decisions are a different matter, and he expects pressure on every justice system to let machines at least advise judges. Private dispute resolution shows how far the economic logic already runs, since eBay decides sixty million disputes on its platform each year within twenty-four hours, ninety per cent of them without any human involvement. The pressure on the bottom of the county court list will arrive as a cost argument rather than a philosophical one.
The pressure on the bottom of the county court list will arrive as a cost argument rather than a philosophical one.
Assisted and made sit closer than they sound
Vos said plainly that the line between the two is hard to draw, and his reason was practical. A system that produces a reasoned recommendation in seconds stays on the assisted side of that line only while a judge has the hours to test what it has told them. He pointed to the American Arbitration Association, which has put an AI arbitrator to work on two-party document-only construction disputes with a human arbitrator reviewing the reasoning afterwards to protect trust and due process. Whether that review is genuine or a formality decides everything, and it is the question your supervisors already face when a fee earner brings them an AI draft to approve.
The settlement figure arrives before the judge
In October he added the step that matters most to a practice. Once machines offer reliable answers, he said, both sides in a small case will ask one what the court is likely to decide, and both will feel economic pressure to settle at the level the machine predicts rather than wait for a judge. That pressure works whether or not any court uses AI, because it acts on the parties before anyone reaches a hearing.
He also described the system he saw at the Shenzhen Intermediate People’s Court in September, where AI distils the evidence, lists the issues it thinks need deciding and drafts reasons for the judge. One judge there told him she now does six cases in the time one used to take. He said he was troubled by a machine listing the important facts when the judge has not read them all. The Standing International Forum of Commercial Courts, drawing on 71 jurisdictions, will likely develop model rules for the use of AI in deciding cases, he added.
What it changes for your low-value work
Nothing changes in your hearings this year. What changes is an assumption sitting underneath the way small practices price the bottom of the caseload. Where first-instance resolution of a modest dispute becomes faster and cheaper, the volume work you run at a fixed fee stops paying in the way it pays now. Open your list, find the small claims, the modest debt recovery and the low-value disputes you take at a fixed price, and ask what your fee buys the client that a quick machine decision would not.
The settlement point reaches you sooner than any machine judge. Expect clients, and the other side, to arrive with a chatbot’s view of what the claim is worth, since Vos told Legal Geek that almost everyone with a legal problem now asks one before going near a lawyer. When your advice departs from that figure, put the reasons in writing, because a number the client has already seen on a screen is hard to move without them.
The consent point belongs in the same conversation. If a client can agree to a machine deciding their case, somebody has to explain what it saves, what it costs and which right it gives up. That explanation is advice, and it lands on you rather than on the court.
Sir Geoffrey sets out the argument in full in his August lecture, The implications of the development of artificial intelligence on the judiciary and the legal profession. He carries it forward in his Legal Geek keynote of 14 October 2026, and the judiciary publishes both, open to read.
If you want to know which parts of your caseload a cheaper way of deciding the same dispute reaches first, that is the kind of question we work through with firms: talk it through with us.
