Human lawyers will survive the arrival of AI because clients need to trust the advice they receive, and a reliable answer from a machine is only the start of that trust. Sir Geoffrey Vos made the argument on 29 October 2026 in the opening keynote to the Law Society of Scotland’s annual conference, two days before he retired as Master of the Rolls. He gave it from the Royal Courts of Justice rather than in Edinburgh, because the Lady Chief Justice held his valedictory hearing that afternoon.
The audience was Scottish, but he spoke as the head of civil justice in England and Wales. He set aside his recent theme, the use of AI in judicial decisions, which this site covered in what AI judges would mean for your small claims. His question this time was whether legal skills, critical legal reasoning and human legal judgment can be preserved in future generations. The title, he said, was slightly tongue in cheek, and his answer to it a resounding no.
The premise and the six reasons
He started from what he called the obvious premise. Anyone with a legal problem, litigant or business, now turns first to ChatGPT, Gemini, Claude or another large language model, and most lawyers use the same tools, or dedicated legal ones such as Harvey and Legora, in their research. Clients who test your advice against a chatbot are already part of practice, as set out in clients using AI chatbots to check your advice.
From there he gave six reasons for rejecting talk of the end of lawyers. Everyone in the system will use AI extensively, and already does. A client accepts advice because of trust and confidence rather than technical competence alone. People take a long time to come to terms with an adverse position. The law is often a small part of a legal problem, since most disputes carry human issues that a reliable opinion on the law does not resolve by itself. People value human advice so highly that they will accept an idiosyncratic or even wrong decision if they think they have been heard. Modern life also generates more legal problems, and many of the intractable ones will still need a human lawyer.
Where he drew the line, and what he asked of lawyers
He tied that trust to experience. Clients will only trust a machine if their experience shows the trust is justified, and he returned to the example he used in his lectures on machine decisions. eBay decides 60 million cases a year, 90% of them without human intervention and apparently without much push back, but those, he stressed, are small claims. We should not assume that clients will extend that trust to more significant problems in their lives, such as major commercial cases or cases in which they stand to lose their homes, their jobs or their children.
He was as direct with the profession. Lawyers should stop pretending that AI may go away and stop saying endlessly that it hallucinates and is too dangerous to use. Every guideline issued by the professions and the judiciaries, he pointed out, says a lawyer can use a machine’s output only after checking it thoroughly and being satisfied that it is accurate and true. The lawyer remains wholly responsible whatever tool produced the draft. He asked lawyers to use AI to save their clients time and cost, to recognise its potential to widen access to justice, which he called a great prize, to be realistic about what machines can properly do, and to avoid protectionism.
What your firm should take from it
Treat the trust he describes as the service you sell. A client who has already asked a chatbot arrives with an answer and wants to know whether to believe it. Ask at the first meeting what they have read and deal with it point by point. Then spend the time AI saves you on the work a model does not do. The examples are ours rather than his: breaking the bad news he says people take so long to accept, and the family dispute behind the boundary claim, which is the human issue he says sits under most legal problems.
Show the saving in the price. Where AI cuts the hours a matter takes, a fixed fee that passes part of that saving to the client tells them what your use of AI gives them. It answers the comparison with a free chatbot better than any warning about hallucinations.
Train your people to evaluate rather than to accept. Sir Geoffrey said lawyers will still need training to evaluate the work of dedicated AI tools. Your juniors once learned that judgement by producing the first draft themselves, and the change in that work is the subject of AI and entry-level legal work in the new jobs plan. Build the checking into their supervision, with a named person signing off what the tool produced.
The judiciary publishes the full text of Are human lawyers a thing of the past? as an open web page with a PDF copy, and it runs to 37 short paragraphs.
If you want to find where AI saves time in your matters and how to show that saving in your fees, talk it through with us.
