A litigant in person using AI beat counsel at Oxford County Court, and the result was reported on 15 September. Lyle Hopkins, an Oxford doctoral student with no legal training and no funds to instruct a solicitor, brought a small claim against the energy supplier SSE over a defunct meter it had billed him for at commercial rates. He prepared the whole case with AI, meaning a paginated bundle of more than a hundred pages, a skeleton argument with authorities, a supplemental bundle answering late-served evidence inside a day, and a plan for cross-examination. District Judge Walker found comprehensively in his favour.
The judge declared that Hopkins owed nothing on the £1,091 SSE was demanding and awarded him £750 for distress and inconvenience, with interest and expenses, indicating that he would have allowed more had more been claimed. He recorded that the defendant seemed interested only in billing at commercial rates a consumer it knew to be domestic, at a property it knew to be a single dwelling, and he added a recital that continued debt collection would amount to harassment. One head of claim failed, £250 said to reflect damage to a credit score, because it disclosed no recognisable claim and no proven loss.
What changed on the other side of the table
The arithmetic of a defended small claim used to favour the represented party for reasons with little to do with the merits. An unrepresented opponent struggled to paginate a bundle, rarely found the right authority, and seldom answered late evidence in time. Preparation was the advantage and preparation cost money. Hopkins closed most of that gap on a student's budget, and he did it against a barrister.
Read what he did with the documents rather than what he did with the software. He went through the filed defence and found errors in it, including an assertion that SSE had not received his letter before action when the company's own emails showed that it had received it and answered it. That is ordinary litigation work done well, and the machine gave him the hours and the vocabulary to do it. The effect on settlement matters as much. SSE offered him more before the hearing than the court went on to award, and he refused because the offer was conditional on confidentiality and non-disparagement, and he wanted the outcome in the public domain. An opponent whose preparation costs him almost nothing is harder to buy off with terms that used to work, because the weight of running cost no longer sits on him alone.
What your firm does with this
Begin by treating documents from an unrepresented opponent as competent until you find otherwise. The instinct to skim a litigant in person's bundle has become an expensive one. Read the skeleton, check every authority in it, and check it properly, since a fabricated citation is now as likely as a poorly chosen one and the court will expect you to say which it is rather than argue past it. Hopkins gave consumers the same warning, telling them to verify every reference the machine quotes and to remember that these systems lean towards telling you what you want to hear.
Then look at what the file costs you. Where your opponent's preparation has become cheap and yours has not, a defended low-value claim stops paying long before judgment. That points at your own use of the same tools for bundling, chronologies and first drafts, and at offers made earlier and on cleaner terms. Hopkins put the point about your profession better than most vendors manage. Lawyers are best placed to use AI because they already understand the law, and the ones who adopt it properly will offer a good deal more for a good deal less, and take the market while they do it.
The account of the case, including the judge's wording and Mr Hopkins' own explanation of how he prepared it, is open to any reader at Legal Futures.
If your firm defends low-value claims and you want an honest view of what that work now costs to run, start with a conversation.