In partnership with
Most weeks this newsletter looks at what lawyers are doing with AI. This week we are taking a different view: a report from the legal charity JUSTICE finds that a sixth of people with a legal dispute have taken it to a chatbot, and that some of them are using it to check whether their own solicitor has behaved reasonably. Alongside it, a LexisNexis survey suggests a portion of the profession now says it could not do the work without AI, and the Online Procedure Rules come into force on Monday.
AI in Practice
Your client is checking your conduct with a chatbot
Recent research has shown just how important client communication is in the age of AI tools.
JUSTICE and the Administrative Fairness Lab published What AI Chatbots Can Teach Us About Unmet Legal Needs on 1 September. The researchers surveyed 3,287 people across the UK, of whom 1,428 reported a legal dispute in the previous two years, and 233 of those (a sixth) had discussed the problem with a chatbot such as ChatGPT. Only 6% used a chatbot and nothing else, so for most people it sat alongside other sources of help rather than replacing them. Use skewed heavily by age: 26% of those aged 18 to 24, against 10% of those aged 55 to 64.
The findings that will interest (and perhaps concern) practising solicitors are not the headline numbers but the uses. People asked chatbots to explain legal jargon, to draft complaint letters, refund requests and formal warnings, and to provide reassurance when they were frightened about a hearing or worried about losing a job. They also, in the researchers' words, used them "to query and check the service they were receiving from their lawyers", asking for a second opinion on whether a solicitor's conduct was reasonable or might ground a complaint.
Two of the anonymised examples are worth quoting: one participant thought it "very unprofessional" that her conveyancing solicitor had verified a building's height using Google Maps, and another asked whether it was "appropriate" for a conveyancer to request mortgage funds "without my consent".
This author would make two points about that. The first (obvious to all solicitors) is that neither of those complaints are particularly well founded, and both are the sort of thing that a client would previously have grumbled about, been reassured, and it would be an end to the matter. A chatbot which is designed to be agreeable will tend to validate the grievance rather than explain the practice. The researchers make exactly that point. Evidence suggests that training chatbots to be friendly and validating makes them more likely to reinforce false beliefs, so a client seeking emotional reassurance may be the client receiving the least reliable answer.
In this author's experience of reviewing complaints as part of the day job, complaints that are clearly written by AI have a tendency to trend towards conspiracy and fraud allegations, rather than an administrative error or no consequence being simply that. Unfortunately, such clients tend to simply regurgitate what the AI has said, rather than express their concerns in their own words, which may well be far better founded than the elements AI has stressed.
The second point is that the file which survives this AI review is the file where the reasoning was explained at the time. If your client understands why you checked a building height the way you did, there is nothing for the chatbot to be indignant about. As always, communication with clients is king.
The access to justice framing deserves attention too, and it is less comfortable than the usual telling that chatbots will make law accessible to all. Chatbot users were younger, more likely to be male and more likely to be employed. Ellen Lefley, deputy legal director at JUSTICE, put the consequence plainly: AI is meeting some unmet legal need, "but not helping the most excluded or marginalised", and for them it may cause more confusion and inequality rather than less.
One caveat on the data: it was collected between the end of 2025 and January 2026, so it is roughly nine months old, and given how usage has moved in that period the real figures today are probably higher.
Takeaways
Act: Look at the last three client letters you sent and ask whether a lay reader could follow why each step was taken. The exposure here is not the decision, it is the unexplained decision.
Watch: Whether complaints arriving at firms start to carry the structure and vocabulary of a chatbot-drafted letter, which is a different intake problem from the complaints you are used to handling. This author has noted a marked increase within his own firm.
Risk: An agreeable chatbot confirming a client's misunderstanding of routine practice, and the firm learning about it when the complaint arrives rather than when the question was first asked. Minimise with client communication at all stages.
Read: JUSTICE, and the reporting at Legal Futures and Artificial Lawyer
On your radar
15% of large firm lawyers say they are now dependent on AI: LexisNexis surveyed more than 500 UK lawyers for a report published on 2 September and found that 15% of those in large firms, and 13% in-house, describe themselves as dependent on AI to do their job. On frequency, 34% use AI tools for legal work every day and a further 32% several times a week, with drafting, legal research and document review the leading uses. Most respondents remained worried about hallucinations, and a majority thought AI use is leaving clients and the public with an incorrect understanding of legal issues, which sits neatly beside this week's lead. The report comes from a vendor with an obvious interest in the answer, so treat the dependency figure as a self-description rather than a measurement. Why it matters for UK lawyers: if a third of your fee earners are using these tools daily, the question of what supervision looks like has already been decided in practice, whatever the policy says. Ask your practice managers what proportion of the firm is in that daily group, because that number is the one your supervision arrangements have to answer. (LexisNexis, Artificial Lawyer)
The Online Procedure Rules come into force on Monday: The Online Procedure (Rules and Practice Directions) Rules 2026 take effect on 7 September 2026. Flagged here twice already, and repeated because it is now four days away and because of one point which has had less attention: the Civil Procedure (Amendment) Rules 2026 amended CPR Part 2 so that where proceedings are governed by Online Procedure Rules, the CPR does not apply except as the online rules, the relevant practice directions or regulations provide. At the outset the new code reaches only possession proceedings, and even there its extent is set by a separate practice direction. Why it matters for UK lawyers: this is the first procedural code written for digital proceedings from the start rather than adapted to them, and it disapplies the rules you know. If you do possession work, read the practice direction before Monday rather than after your first online claim. (legislation.gov.uk, Civil Litigation Brief)
A UK firm is competing on published fixed fees and same-day turnaround: Kyra Law, interviewed by Artificial Lawyer on 2 September, is a UK practice built around AI from the start rather than retrofitted, serving startups from pre-seed to Series B on commercial contracts, shareholders' agreements, term sheets, funding rounds and EMI and SEIS/EIS work. Fees are published and fixed, turnaround is quoted in hours or days, advice is signed off by solicitors holding practising certificates, and the business operates on an unreserved basis rather than as an authorised firm. Pricing is set from data on past matters, which is the part worth noting: the claim is that AI makes fixed fees forecastable, not merely cheaper. Why it matters for UK lawyers: the competitive threat to a mid-market commercial practice is not a chatbot, but a smaller outfit publishing a price for work you quote on request. If you act for early-stage companies, look at what a comparable piece of work now costs on a published fixed fee and decide what you are charging for. (Artificial Lawyer)
Reminder: the Legal Services AI Growth Lab closes to applications on 27 September: Covered here on 14 August, and repeated because the deadline is now three weeks away. The Ministry of Justice is inviting law firms, conveyancing businesses and lawtech developers to bring real AI proposals through a supervised process with input from the Legal Services Board, the SRA, the Council for Licensed Conveyancers and the ICO. Why it matters for UK lawyers: if a project in your firm has stalled on a confidentiality or privilege question nobody will answer, this is the route to put the question to a regulator directly. Check the eligibility criteria this month rather than next. (GOV.UK)
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.
Speak naturally. Send without fixing.
Wispr Flow turns your voice into clean, professional text you can send the moment you stop talking. Not rough transcription you have to clean up. Actual polished text — ready for email, Slack, or any app.
Speak the way you think. Go on tangents. Change your mind mid-sentence. Flow strips the filler, fixes the grammar, and gives you text that reads like you spent five minutes writing it.
89% of messages sent with zero edits. Millions of professionals use Flow daily, including teams at OpenAI, Vercel, and Clay. Works on Mac, Windows, and iPhone.
For Review
What AI Chatbots Can Teach Us About Unmet Legal Needs (JUSTICE and the Administrative Fairness Lab)
The report behind this week's lead, and the section to read first is the one setting out how participants actually used chatbots, because it includes anonymised extracts of the conversations. If you supervise a consumer-facing team, those extracts are the clearest picture available of what your clients are asking about you when you are not in the room.
Read or listen: JUSTICE
The AI-Dependent Lawyer (LexisNexis)
The survey behind the second item, useful mainly for the frequency data rather than the dependency headline. Worth having to hand if you need to show a management board that daily AI use is already the norm across the firms you compete with, though it is vendor research and the framing reflects that.
Read or listen: LexisNexis
The Online Procedure (Rules and Practice Directions) Rules 2026 come into force on the 7th September: here is a summary (Civil Litigation Brief)
Gordon Exall's summary of the rules and what they do, which is the quickest way into a statutory instrument most litigators have not yet opened. Read it alongside the SI itself if you conduct possession work.
Read or listen: Civil Litigation Brief
Practice Prompt
This week's prompt is a favourite of the author's for testing a client letter or update. It stress tests against the questions your client is likely to put to a chatbot after reading it, which is exactly the gap this week's JUSTICE research showed. It works on a redacted draft, and what comes back is a list of the points a lay reader will not follow and the sentences which invite a complaint, not a rewrite of your letter. 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 who is about to send a letter or update to a client. Your task is to predict how a lay client will read it, and what they are likely to ask a general-purpose chatbot after reading it. You are not rewriting the letter and you are not advising on the underlying law.
Context to apply:
- The draft communication, redacted of names, addresses, figures and anything identifying: {paste}
- Matter type and stage: {e.g. residential conveyancing, post-exchange; employment tribunal, three weeks before hearing}
- What the client is most worried about, in their words if you have them: {e.g. "will I lose the house", "how much is this going to cost", or "unknown"}
- The client's likely familiarity with legal process: {first-time client, repeat commercial client, litigant with prior proceedings}
- Anything in the letter which reports a decision the client did not ask for and may not expect: {list, or say "none"}
- What the client has already been told about fees and timescales: {summarise, or say "nothing in writing"}
- Your firm's complaints procedure and whether it has been provided: {yes and when, or no}
Produce the review under these headings:
1. Plain reading
Restate what the letter says as a non-lawyer would understand it, in no more than 200 words. Do not correct or improve it. Where a sentence is ambiguous to a lay reader, give both readings.
2. Where the reasoning is missing
Identify each point where the letter states a step, a decision or a cost without explaining why it was necessary. For each, quote the sentence and say what a client would need to be told for it to make sense to them.
3. Questions the client will ask a chatbot
Draft the questions, in the client's own likely wording rather than legal language, that a reader of this letter would plausibly put to ChatGPT or a similar tool. Include the ones which start "is it normal that", "is my solicitor allowed to" and "should I complain about". Be realistic rather than charitable.
4. Where an agreeable chatbot would get it wrong
For each question in section 3, set out the answer a friendly, validating chatbot is likely to give, and say whether that answer would be accurate, incomplete or simply wrong about ordinary practice in England and Wales. Mark clearly where you are uncertain about the general practice rather than guessing.
5. Complaint risk
List the sentences most likely to generate a complaint, ranked, with the reason for each. Distinguish between a complaint about the work and a complaint about the communication, because the second is the one this exercise can actually reduce.
6. What to add, not what to change
Propose additions the solicitor could make to close the gaps found in sections 2 and 5, expressed as the substance to cover rather than as drafted text. Keep each to one or two sentences so the letter does not double in length.
7. What this review cannot tell you
Flag every point where the output rests on an assumption about the matter, the client or ordinary practice, and say what the solicitor would need to establish to replace it with a fact.
Constraints:
- {Add matter-specific constraints, for example a client who has already complained, a fixed-fee retainer, a funder or insurer in the background, or a deadline which limits how much can be added.}
- Apply the law and ordinary professional practice of England and Wales, and the SRA Standards and Regulations. Where you are unsure whether something is ordinary practice, say so rather than asserting it.
- Do not invent regulatory requirements, SRA expectations, case law, Legal Ombudsman outcomes or firm procedures.
- Do not advise on the merits of the underlying matter, and do not suggest changes to the legal advice in the letter. Your subject is how it will be read.
- Do not rewrite the letter or produce a replacement draft.
- Write in plain English, and keep the client-voice questions in section 3 genuinely colloquial.
- This is a communication planning aid, not legal advice. The solicitor remains responsible for the letter and for the advice 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!
