In partnership with
This week the person using AI was on the other side of the case, and he was not a lawyer at all. An Oxford doctoral student took a small claim against SSE to the county court, produced more than a hundred pages of documents with the help of consumer chatbots, and won. Alongside it, the courts minister warned against a justice system where City firms run ahead and the courts stay put, and two sets of firm finance research landed in the same week with the same uncomfortable finding about pricing.
AI in Practice
The claimant with no solicitor and a hundred pages of documents
Legal Futures reported on 15 September that Lyle Hopkins, an Oxford doctoral student with no legal training, succeeded in a small claim against SSE at Oxford County Court before District Judge Walker. The dispute was an ordinary billing one: SSE had charged him at commercial rates across two accounts for a single property, including for a defunct meter, having been told he was a domestic residential consumer. The judge declared him not liable for the £1,091 claimed, awarded him £750 for distress and inconvenience together with interest and expenses, and made a declaration against further debt collection. A separate claim for £250 in respect of damage to his credit score was dismissed for want of any legal basis.
The detail which matters for practitioners is not the outcome but the preparation. Hopkins says he used a succession of consumer chatbots to produce over a hundred pages of material, including the bundle, a skeleton argument with authorities, and a cross-examination plan. His own summary of it ("AI was the difference between coping and not") is the most honest description this author has seen of what these tools currently do for a person in his position. It is effectively a mirror of the real strength of having a solicitor represent you in these types of dispute: organising the material and reducing the stress. He was careful about the limits too, warning that the tools "tend to want to tell you what you want to hear" and that a user has to check every reference.
Some caution is needed about how far this goes. There is no published judgment, so the account is based on the report and on the claimant's own description of his case. Of course, a small claim in the county court binds nobody, and the outcome turned on SSE's conduct (the judge is reported to have said the company had subjected Hopkins to "a rollercoaster ride", continuing even after an ombudsman had found for him) rather than on any clever point of law. This was, on any view, a claim which ought to have been settled long before a hearing.
In this author's opinion (as a solicitor who regularly consults on small value disputes), that is rather the point.
Many claims used to go away because the claimant could not face the paperwork or cost of a solicitor to do it for them. These claims are the ones where AI can change the picture. Volume defendant work is priced on the assumption that most complainants will give up. A litigant in person who arrives with a skeleton argument, a properly paginated bundle and a plan for cross-examination is not a stronger opponent on the law itself, but he is a far more expensive one to defeat, and a much harder opponent to wear down.
The Civil Justice Council's working group has reached a similar place from the other direction: its June update records broad agreement that no additional formal AI requirements are needed for professional drafting, while the questions it has not closed are about witness statements and about self-represented litigants. Those are the two places where AI is doing the most work and where the least is currently required of anyone.
Takeaways
Act: Look at how your team triages low-value defended claims, and ask whether the assumption that an unrepresented claimant will not follow through still holds. If your costs model depends on it, the model needs revisiting rather than the file.
Watch: The Civil Justice Council's draft report, expected later this year, and in particular whatever it proposes for litigants in person, because that is the open question the rest of the profession has been treating as settled.
Risk: Authorities cited against you by an unrepresented opponent which nobody has verified. The verification burden does not shift because the other side has no solicitor, and a fabricated citation going unchallenged is a problem for the court and for you.
Read: Legal Futures, and the Civil Justice Council's current work page
On your radar
The courts minister has warned about a two-tier justice system: Sarah Sackman, Minister for Courts and Legal Services, told the City of London Law Society's "London Law Imperative" conference on 16 September that she did not want a position in which "large City law firms are at the cutting edge of AI use while the courts are stuck in the analogue age". She pointed to the AI Growth Lab as an operating sandbox, said the Ministry of Justice is already deploying transcription and case summarisation tools for judges, and trailed a Global Business and Law Summit in December at which a strategy will be launched. No implementation timetable for court systems was given. Why it matters for UK lawyers: judicial case summarisation is now a live deployment rather than a proposal, which means AI-assisted material may already be in the process on your matters without appearing anywhere on the file. If you conduct civil work, put the December summit in the diary and treat it as the point at which the government's position becomes concrete. (Legal Futures)
Firms are cutting billable hours with AI and raising billable hour targets anyway: BigHand's 2026 annual law firm finance report, drawn from more than 800 senior finance and legal leaders, found that nearly two thirds of firms report a decrease in billable hours while 99% plan to increase billable hour targets, and that only 34% have updated their pricing models to reflect AI efficiencies. Reporting on the same research notes that 31% say AI has improved profitability per matter and that 35% of partners are uncomfortable discussing AI use with clients. It is vendor research from a business which sells pricing software, so read the framing accordingly, though the gap between those two numbers is difficult to explain away. Why it matters for UK lawyers: where the pricing model has not moved, an efficiency gain becomes an unrecorded discount to the client or an unrecorded squeeze on the fee earner, and usually both. Ask your finance team which of your matter types have had their fee assumptions revisited since AI tools came in, and treat the ones which have not as the priority. (BigHand, Legal IT Insider, Artificial Lawyer)
79% of SME firms have now committed specific funds to AI: Research by the accountants Crowe with the Institute of Legal Finance & Management, reported on 16 September, surveyed 39 firms with revenues between £1.7m and £69m and found that 79% have set aside money specifically for AI tools, with around half naming profitability and margin rather than turnover as their main strategic objective. Revenue growth held at 11% to 12%, headline profits rose 12% in London and 6% regionally, and staff costs remain the dominant pressure, with 88% planning pay rises capped at 5%. Why it matters for UK lawyers: a budget line is not a plan, and a firm which has committed funds without deciding what it is buying them for tends to end up with several overlapping licences and no supervision arrangement covering any of them. If your firm has an AI budget this year, find out who owns the decision about what it is spent on. (Legal Futures)
Google's legal offering is still not a settled product: Following the launch of Gemini Enterprise for Legal on 25 August, covered here as the lead on 28 August, Google demonstrated "Cloud Legal AI Helpers" on 17 September: configurable assistants connected to sources of the firm's choosing, offered in Pro and Flash versions, for tasks such as redlining and brief writing, alongside NotebookLM connected to as many as 300 sources. The write-up is lukewarm rather than hostile, noting that the naming across Google's legal products remains unclear, that there is no proprietary legal content behind any of it, and that the suite is "not yet tied down into a fully unified set of products". Why it matters for UK lawyers: this is included as a follow-up because the August launch prompted procurement conversations in several firms, and the honest position three weeks on is that the thing being evaluated is still moving. If your firm is in that conversation, ask for the product roadmap in writing before committing to an infrastructure decision you will not easily reverse. (Artificial Lawyer)
Ad Break
In order to help cover the running costs of this newsletter, please check out the advert below. In line with my promises from the start, adverts will always be declared.
Product teams aren’t short on ideas. They’re missing a system.
Jira Product Discovery gives teams one place to capture customer feedback, prioritize ideas with consistent frameworks, and build living roadmaps everyone can align on. And when it’s time to build, those decisions connect directly to delivery in Jira, so everyone can see how the roadmap turns into real work.
For Review
Use of AI in preparing court documents (Civil Justice Council)
The working group's current work page, which sets out where the consultation has landed and what remains open. Read the June update on responses: the finding that respondents broadly saw no need for AI-specific requirements in professional drafting is the part most commentary has missed, and it tells you what the eventual rules are unlikely to say.
Read or listen: Courts and Tribunals Judiciary
2026 Annual Law Firm Finance Report (BigHand)
The research behind the second radar item. Go to the material on write-offs, lock-up and pricing models rather than the AI headlines, because the useful argument for a management meeting is the one about aged work in progress, and the AI findings land harder when they sit inside it.
Read or listen: BigHand
AI hallucination cases, UK (Natural and Artificial Law)
A running tracker of confirmed and suspected fabricated citation cases before the courts of England and Wales, which stood at 64 when last updated in May. It is not this week's news and should not be read as such, but it is the most convenient single reference if you need to show a sceptical colleague that this is a pattern rather than a handful of anecdotes, and the author's observation that use by litigants in person is rising sits directly alongside this week's lead.
Read or listen: Natural and Artificial Law
Practice Prompt
Try the below prompt to work out how to handle a defended claim where the opponent has no solicitor and their documents look AI-assisted, which is the position this week's lead makes more likely rather than less. It produces a verification plan and a proportionality assessment, not a view on who wins. Ensure you fill in context and constraints and other aspects marked with {}. Remember to adhere to the Golden Rules and do not upload confidential or privileged information to public tools.
You are assisting a solicitor in England and Wales acting for a defendant in a low value civil claim where the claimant is a litigant in person and their statements of case or skeleton argument appear to have been prepared with the assistance of an AI tool. Your task is to produce a handling plan covering verification, merits and proportionality. You are not advising on the outcome and you are not drafting the defence.
Context to apply:
- The claim, in outline, redacted of names and identifying detail: {value, cause of action, stage, track, hearing date if listed}
- What the claimant has served, and what about it suggests AI assistance: {e.g. authorities cited in an unusual format, American terminology, a level of structure inconsistent with the correspondence, citations you cannot find}
- The authorities and sources the claimant relies on, exactly as cited: {list, or say "none cited"}
- The correspondence history and any pre-action conduct on both sides: {summarise, including anything the client did which a judge would criticise}
- What your client actually did, as distinct from what it says it did: {be candid, or say "not yet established"}
- Any ombudsman, regulator or complaint finding already made, and by whom in whose favour: {describe, or say "none"}
- The commercial position: {cost of defending to trial, cost of settling now, any reputational or precedent concern, any insurer or funder involvement}
Produce the plan under these headings:
1. Verification of what has been cited
For each authority, statutory provision or source the claimant relies on, set out how to establish whether it exists, whether it says what it is said to say, and whether it is good law. Distinguish clearly between a citation you have checked, one you cannot locate, and one you have not yet checked. Do not assume a citation is fabricated because you have not found it, and do not assume it is sound because it looks plausible.
2. What to do about anything which does not check out
Set out the options where a citation appears not to exist or to be materially misdescribed, covering what to put to the claimant and when, what to put before the court and in what form, and the tone appropriate to an opponent who is unrepresented and may not understand what has happened. Flag the duty to the court which applies to you whatever the claimant has done.
3. The claim underneath the documents
Strip out the presentation and state the claim at its highest as a court would understand it. Identify which elements are genuinely in issue, which are admitted or unarguable, and where the claimant's own material helps your client. Presentation quality and merit are independent of each other, so say plainly which one you are assessing at each point.
4. Where your client is exposed
List the points on which the claimant is likely to be right, or on which your client's conduct would attract judicial criticism, including any regulatory or ombudsman finding already made. Rank these by how much they would cost at a hearing, and say which of them are capable of being cured now.
5. Proportionality
Compare the cost of defending to a hearing against the cost of resolving the claim now, taking into account that an unrepresented claimant using AI tools is less likely to withdraw through inability to prepare than one who is not. Say what additional information would change the answer.
6. Handling the hearing
If the matter is listed, set out what to expect from a prepared litigant in person, what the court is likely to do about any AI-related issue, and how to raise one without it looking like an attack on a person acting for themselves.
7. What this plan cannot tell you
Flag every point resting on an assumption about the claim, the claimant, your client or the court's likely approach, and say what you would need to establish to replace it with a fact.
Constraints:
- {Add matter-specific constraints, for example an insurer with conduct of the defence, a portfolio of similar claims where settlement sets a pattern, a fixed costs regime, or a client instruction to defend on principle.}
- Apply the law and civil procedure of England and Wales, and the SRA Standards and Regulations. Where you are unsure whether something is a rule, a practice direction or ordinary practice, say so rather than asserting it.
- Do not invent case law, citations, procedural rules, practice directions or court guidance. If you cannot verify an authority, say that you cannot verify it.
- Do not advise on the merits as though the outcome were predictable, and do not recommend settling or defending as a conclusion. Set out the considerations and what would move them.
- Write in plain English, and keep section 1 practical enough that a paralegal could work through it.
- This is a case handling aid, not legal advice. The solicitor with conduct remains responsible for the claim and for everything filed in it.How did we do?
Hit reply and tell me what you would like covered in future issues or any feedback. We read every email!
Thanks for reading,
Serhan, UK Legal AI Brief
Disclaimer
Guidance and news only. Not legal advice. Always use AI tools safely.
Recommended Newsletters
Below are a few newsletters that I recommend, for various reasons. Check them out!
