Fracas Digital • Aug 6, 2026 • 9 min read
AI for Solicitors: The UK Tools Guide
AI for solicitors means three distinct categories of tool: legal research (Lexis+ AI, vLex Vincent, CoCounsel), drafting and contract review (Spellbook, Genie AI, Harvey, Luminance), and practice management (Clio Duo, LEAP Generator, Smokeball's built-in AI). None of them remove the solicitor's duty to supervise the output. The SRA has not written AI-specific rules, but its existing principles, competence, client best interests, and confidentiality, apply exactly as before, and a named human has to check the work before a client sees it.
That's the short version. Most of what's published on this topic comes from the tool vendors themselves, which means the roundups tend to rank their own product first and skate past the part that actually matters to a COLP: what happens when the output is wrong.
We build custom AI automation for SRA-regulated firms, so this guide is written from the seat of someone who has had to answer that question to a partner's satisfaction, not just write copy about it.
Is it legal for UK solicitors to use AI?
Yes, with conditions the SRA has been consistent about since its 2025 risk outlook. There is no dedicated AI rulebook. Instead, four existing obligations do the work:
- Competence (Principle 4). You need to understand roughly how the tool works and where it fails, not just that it produces plausible text.
- Client best interests (Principle 7). Speed and cost savings don't excuse a worse outcome for the client.
- Confidentiality (Code paragraph 6.3). Client data cannot go into a consumer-grade AI tool that trains on inputs. This is the single most common mistake we see: a fee earner pastes a client's instructions into a free chatbot to save ten minutes.
- Accountability. The SRA's position, echoed by the Law Society's generative AI guidance, is that using AI does not dilute a solicitor's duties. If the output is wrong and it reaches a client, the solicitor who let it through is responsible, not the software.
The clearest real-world marker of where this is heading: the SRA authorised Garfield.Law in May 2025, the first AI-driven law firm it has licensed. That was not a green light for unsupervised AI. It was a statement that the SRA will authorise AI-heavy models provided the accountability structure holds.
The best AI tools for solicitors, by task
Nobody needs every tool on this list. Match the tool to the task that's actually costing your fee earners time.
Legal research
Lexis+ AI and vLex Vincent are the two names that keep coming up in UK benchmarks because both are trained on UK case law and legislation specifically, rather than a general model bolted onto a search box. CoCounsel is capable but was built with a US practice in mind first, so UK citation accuracy needs closer checking.
None of these tools have a concept of truth. They retrieve and summarise; they don't know when they're wrong. Treat every citation as a lead to verify in the primary source, not as a finished answer you can drop into a skeleton argument.
Drafting and contract review
Spellbook and Genie AI sit inside Word, running first-pass drafts and redlines against a precedent bank. Harvey and Luminance are built for higher-volume review, priced more for firms doing enterprise-scale due diligence than a five-partner high street practice.
The pattern we see in firms we've worked with: drafting tools are genuinely good at producing a first draft in your house style once they're calibrated on your precedents. They are not good at knowing which clause a particular client relationship actually needs. That judgement call stays with the fee earner.
Practice management and intake
Clio Duo and LEAP Generator are the AI layers bolted onto the two most widely used UK practice management systems, and Smokeball ships its own AI features natively. All three can triage new enquiries and draft a passable first engagement letter.
Where these tools tend to run out of road is anything that crosses systems: an enquiry that comes in by phone, gets logged in one place, and needs a conflict check against records held somewhere else entirely. That's the gap between a subscription tool and a custom-built agent, and it's usually the first thing a firm discovers once it's actually tried to automate intake end to end. We've written a longer breakdown of that specific workflow in our guide to AI client intake for law firms.
Client-facing and admin
An AI receptionist that answers calls and web enquiries outside office hours and books them straight into the right diary closes a gap most firms don't realise they have. The enquiry that comes in at 6pm on a Friday would otherwise sit until Monday. Microsoft Copilot, which most firms already pay for through a Microsoft 365 licence, is worth configuring properly before buying anything else. Most firms we talk to have the licence and have never turned the legal-specific settings on.
When a subscription tool is not enough
A subscription tool solves a single, well-scoped problem: drafting inside Word, research inside a case-law database, matter notes inside your PMS. It stops being enough the moment the problem spans systems, which for most firms means intake, compliance, or anything that has to pull from more than one source before a human sees it.
At that point the choice isn't really "AI tool versus no AI tool." It's "off-the-shelf tool versus something built around how your firm actually works." Most vendors start from their product and ask you to adapt. A firm-specific build starts from the practice, which is a different conversation and, in our experience, a different result. If that's the gap you've hit, our AI automation for law firms work is built for exactly that handover point.
What breaks if you get this wrong
Three failure modes account for almost every AI incident we've seen or heard about in a UK firm:
- Hallucinated citations. A model invents a case that sounds plausible and cites it with confidence. This has already produced sanctioned lawyers in multiple jurisdictions. The fix is procedural, not technical: nothing goes to a client or the court without a named human checking every citation against the primary source.
- Confidentiality leakage. Client instructions pasted into a consumer AI tool that trains on inputs. Once it's in, it's not coming back out. The fix is infrastructure: use tools with a data processing agreement and zero retention, or self-hosted models where privilege genuinely demands it.
- Unsupervised output reaching a client. The AI drafted something reasonable-looking and nobody checked it before it went out. This is the one that actually gets firms in front of the SRA, because it's the clearest evidence the accountability structure wasn't there in the first place.
A practical checklist before you deploy any AI tool
- Confirm the tool has a data processing agreement and does not train on your inputs.
- Name the human who reviews output before it reaches a client, in writing, not as an assumption.
- Run a DPIA if the tool will touch real client data at any scale. The ICO's threshold catches most intake and drafting tools once they're live.
- Test the tool on closed matters first, not live ones, and keep a log of what it got wrong.
- Brief your COLP before rollout, not after an incident.
FAQs
Is AI legal for solicitors to use under SRA rules? Yes. The SRA has not published an AI-specific rulebook, but its existing principles apply in full: solicitors must maintain competence, act in the client's best interests, and protect confidentiality. Using AI does not move liability off the solicitor. A named human must supervise and check the output before it reaches a client or the court.
What is the best AI tool for legal research in the UK? Lexis+ AI and vLex Vincent are the two most commonly cited tools purpose-built for UK legal research, both trained on authoritative UK case law and legislation rather than general web text. Whichever you choose, treat citations as a starting point to verify, not a finished answer.
Can AI draft legal documents for a solicitor? It can produce a competent first draft. Tools such as Spellbook and Genie AI work inside Word to draft and review contracts. None of them should produce a document that goes to a client unreviewed.
Do solicitors need a DPIA before deploying an AI tool? Almost always, if the tool processes client personal data at any scale. Firms that skip this step leave their COLP exposed if a regulator asks how the deployment decision was made.
Should a solicitor buy an AI subscription or build custom automation? Buy a subscription when a single, well-defined task fits an off-the-shelf product. Move to custom automation when the problem spans systems, such as intake that crosses phone, email, and your practice management system. Most firms need both.
If your firm has outgrown what a subscription tool can do, talk to us about AI automation built around how your practice already works.