AI for law firms: where it helps, where it fails, and how to start
A plain guide for small UK law firms: where AI helps, the fake citation risk, SRA and Law Society duties, and a safe start.
The fear in every small firm is the same. Someone files a document that cites a case which does not exist. The partner finds out from the judge.
That has happened, and it is why AI for law firms needs care. It is also not the whole story. Used well, AI saves time on the dull parts of the job. This guide is for the owner or partner of a small UK firm. It covers where AI helps, the known failure, the duties that matter, and a safe way to start. This article gives general information only and is not legal advice for your firm.
Where AI for law firms helps today
General tools such as Claude or ChatGPT are good at drafting and sorting text. They are not legal databases. Keep the work to jobs where a lawyer reads the result.
- First drafts. Letters, attendance notes and routine clauses, written from your instructions for a fee earner to edit.
- Document review and summaries. A short summary of a long bundle, with page references you then check against the documents.
- Intake and enquiry replies. A polite first reply to a new enquiry, and the questions you need answered.
- Matter updates to clients. A plain English progress note drafted from the file notes, ready for approval.
- Knowledge search across your own precedents. Finding the right clause or template in your own library.
The SRA has said AI could help firms deliver administrative tasks more efficiently and at lower cost, freeing time for more complex work. Its report also notes that the bigger risk for some firms may be not adopting AI at all.
A good test for any job is this. Could a trainee do it, and could you check it in a few minutes? If yes, it is a fair candidate. If the check would take as long as the task, leave it.
Enquiry replies need a special note. A reply that sounds like advice can create expectations, so keep them to facts about your process and let a lawyer handle the rest.
The known failure: made up case citations
On 6 June 2025 the Divisional Court gave judgment in Ayinde v London Borough of Haringey and Al-Haroun v Qatar National Bank. Both cases involved material put before the court that was false.
The court said freely available generative AI tools are not capable of reliable legal research. They may cite sources that do not exist, and the answers can look entirely plausible.
It set out three points every firm should note:
- Lawyers who use AI for research must check it against authoritative sources before relying on it.
- That duty also applies to lawyers who rely on someone else's AI assisted work.
- Managing partners and heads of chambers must take practical and effective measures so that their lawyers understand these duties. In future hearings the court may ask whether they did.
The possible consequences range from a wasted costs order to a referral to the regulator. Contempt of court is possible in the worst cases, where there is knowledge that the material is false. In both of these cases a solicitor was referred to the SRA.
The lesson is simple. Never let AI be the source of a case, a statute or a quotation. Find it in a proper database, read it, and then cite it.
Duties that matter: confidentiality, competence, supervision
At the time of writing (November 2026), the Law Society's Generative AI: the essentials is the main guide for solicitors. It links AI use to the SRA Code of Conduct.
- Confidentiality. Do not feed confidential information into tools you do not control. Free online services carry the most risk.
- Competence. You need to understand what the tool can and cannot do, and test the vendor's claims.
- Supervision. Your duties apply to AI assisted work, whether it is yours or a colleague's.
- Honesty. Misleading the court or a client breaches the Code, even without deliberate dishonesty.
The guide also advises vendor due diligence, staff training, an internal policy, and keeping a record of inputs, outputs and errors. It suggests telling clients whether and how you use AI, and checking your insurance.
The SRA's Risk Outlook report on AI (November 2023) makes the same point from the regulator's side. Your firm stays responsible for the output, and you cannot hand that to an IT team or a supplier. It also advises against relying on an AI system to judge its own accuracy.
Client data and UK GDPR
Client files hold personal data, often sensitive. UK GDPR still applies when a tool handles it. The ICO's guidance on AI and data protection sets out what it expects on accountability, transparency, lawfulness, accuracy and fairness.
The ICO page was last updated on 15 March 2023 and is under review because of the Data (Use and Access) Act. Check the ICO site for the current position.
The Law Society notes that vendors may see what you enter, and that data may be processed outside the UK. So before any client material goes in, find out:
- Where the data is stored and who can see it.
- Whether the provider can reuse it.
- Whether you can delete it.
Write the answers down. You may need them for a client, an insurer or the regulator.
A safe way to start
- Write a short internal rule on what may go into AI tools and who checks the output. Our AI policy template for UK businesses can help.
- Choose one job with no court deadline and no new law in it, such as a first reply to an enquiry.
- Test with made up or anonymised material first. The Law Society suggests fictional data for testing.
- Make a fee earner read and approve every output. Nobody relies on a citation unless they have opened the source.
- Keep a short record of what you tried and what went wrong.
- Review after a month. Add the next job only if the first one earned its place.
When AI is the wrong tool
Do not use a general chatbot for legal research or for anything you will submit to a court without full checking. Do not use it where the answer must be exactly right and you cannot verify it quickly.
Judgement calls, advice to a client and anything involving privilege are for a lawyer. AI can prepare the ground. It should not decide.
Where Agentic Scale fits
For a law firm we start with the firm's own material. Precedents, house style and intake rules are written into an AI Brain, a set of plain files held in the firm's Drive and read by the firm's Claude account.
AI employees then draft, search and summarise from that material only, and a fee earner approves every output. Nothing reaches a client or a court unreviewed.
The usual first step is the AI Audit at £197, with the fee credited in full if a build starts within 30 days. It sorts your tasks into those that suit AI and those that stay with a lawyer. More detail is on our page about AI Agents for law firms.
Sources
Can lawyers use AI?
Is it safe to put client documents into Claude or ChatGPT?
What happened in the Ayinde case?
Who is responsible if AI gets it wrong?
An AI receptionist can answer calls and book appointments, but it needs rules, testing and a person watching. Here is what UK firms should check.
5 min read · 12 November 2026Where AI helps a small UK accountancy practice, where it must not work alone, what ICAEW and ACCA say, and a safe way to start.
5 min read · 12 November 2026Agentic AI plans and carries out jobs, not just answers questions. Here is what that means, where it helps and where to be careful.
5 min read · 12 November 2026Not sure where to start?Book an AI Audit.
We look at how you work today and tell you honestly what AI should and shouldn't do for you. £197, credited in full if you build within 30 days.