What changes
AI capability means the firm understands the tools well enough to choose them, use them and explain them to a client. It isn’t bought as a black box from a supplier and then left alone. In practice that means an approved set of tools, a written policy that fee earners actually follow, and enough training that every solicitor reviews AI output critically, the way they’d review a trainee’s first draft.
The foundation is confidentiality. Client information shouldn’t go into consumer AI tools that may keep prompts or use them for training. Use business or enterprise versions with contractual terms on data use, retention and storage location, and check them as you would any outsourced supplier. Think about legal professional privilege too: keep privileged material inside tools and accounts the firm controls, and record what was shared with which tool and why.
Why it matters now
Every later move depends on this one. You can’t offer a fixed-fee check of a client’s AI-drafted document if your own people can’t spot what AI gets wrong: invented case references, clauses from the wrong jurisdiction, and confident answers built on out-of-date law. SRA continuing competence asks solicitors to reflect on their practice and address their learning needs each year. For most solicitors, AI is now one of those needs, and learning it in a structured way is far cheaper than learning it from a complaint.
Checklist
- Choose one approved general AI assistant on business terms and one legal research tool. Check data handling, retention and where data is stored before anyone signs a contract.
- Write a one-page AI policy covering approved tools, what client information may go into them, how output is checked and who supervises. Share it with your indemnity insurer.
- Book every fee earner onto AI Fundamentals for Solicitors, then Legal Research with AI and Compliance & Risk Management, and have each one note it in their annual reflection on competence.
- Name an AI lead, ideally a partner, who owns the tool list and the policy and gives the partners a short update every month.
- Run a fortnightly half-hour session where fee earners share one task AI helped with and one output it got wrong, and add the lessons to the policy.
Illustrative example
A firm of 30 staff buys business licences for an AI assistant at £25 per user per month for its 20 fee earners, which is £500 a month or £6,000 a year. If each fee earner saves one hour a week on research, summaries and first drafts, that’s 20 hours a week across the firm. Over 46 working weeks, that comes to 920 hours a year. The firm doesn’t try to bill those hours. It uses them to take on fixed-fee work, answer clients faster and give solicitors time for the supervision the policy requires.
Common mistakes
- Letting each person pick their own tool. Client data ends up in a dozen places, and the firm has no practical way to supervise how it’s used.
- Training once and stopping. The tools change every few months, so build a short, regular habit of practice and review instead of a single event.
- Trusting citations. Check every case and statute reference against a primary source before it goes anywhere near a client, an opponent or a court.