Insights

What AI can do in a law firm today, and what it should not

AI in a law firm can already take real work off a fee earner's desk, from first drafts to checking contracts. It works best when a lawyer reviews and approves what it produces before anything leaves the firm. This guide sets out where it helps, where it should stop, and how to judge what a vendor tells you.

What AI can do in a law firm today

The most useful legal AI is not a chat window on the side of the desk. It works inside the matter, with the matter's own information, and produces something a fee earner would otherwise have written or checked by hand. Four kinds of work stand out.

Drafting from the matter's own data

Letters, reports and contracts follow patterns. When the AI can see the parties, the property or the issues, the key dates and the firm's own templates, it can prepare a first draft that is already specific to the matter. The fee earner's time goes on judgement and finishing, not on retyping names and dates. The quality of the draft depends heavily on the quality of the matter record and the templates behind it, which is one more reason to keep both in good order. Our page on legal document management software covers drafting from templates in more detail.

Review and checking

AI is good at reading a long document against a list and flagging what does not fit: an unusual clause, a missing term, a date that does not match the rest of the file. Used this way it is a second pair of eyes that points a lawyer to the parts that need attention. It does not replace the lawyer's reading; it helps them spend it where it matters.

Assembling reports

Some reports draw together information that already sits in the file. A report on title, for example, pulls from the title, the searches and the enquiries. AI can assemble that material into a structured draft for the fee earner to review, correct and sign off. Our page on conveyancing case management software shows where that fits in a property matter.

Keeping matters moving

Not all useful AI writes anything. Prompting the next task when a stage completes, noticing that a matter has gone quiet, or putting the right piece of work in front of the right person all help a matter keep moving. This kind of help is easy to overlook and often saves the most time across a team.

What AI should not do

The limits matter as much as the capabilities, and a firm should be able to say clearly where its AI stops.

  • It should not decide. Advice, judgement and the choices that carry professional responsibility belong to a lawyer. AI can prepare the material for a decision; it should not make one.
  • It should not send unreviewed work. Nothing the AI produces should reach a client, another party, a court or a registry until a fee earner has read and approved it.
  • It should not hold client data outside the firm's controls. Client information used by AI should stay within systems the firm has chosen, secured and can account for, under the same permissions as the rest of the matter.

See AI reviewed and approved on a matter

Book a demo and we will show you drafting and review inside a matter, with a fee earner approving the work.

Supervision and the COLP's record

Firms generally treat AI the way they treat any other tool used in client work: someone is accountable for how it is used, there is a written approach to it, and there is a record of who checked what. In practice that means a few things.

  • Every AI-assisted action is recorded on the matter, alongside the fee earner who approved it and when.
  • The firm can decide which kinds of work AI is used for, and change that as it learns.
  • Supervisors can see AI-assisted work in the same way they see any other work on a file.
  • The COLP can show how AI is used and overseen, from the record rather than from memory.

Regulators and professional bodies publish guidance on the use of AI in legal services, and it continues to develop. It is worth reading the current versions directly and taking advice where the firm's position is unclear, rather than relying on any vendor's summary, including ours.

How to evaluate a vendor's AI claims

Most vendors now describe some form of AI. These questions help separate a demonstration from something a team would use every day.

  1. Show it on a real matter. Ask to see the AI work on a matter like yours, from the matter's own data, not on a prepared example.
  2. Who approves? Ask where the fee earner reviews the output, and what happens if they reject or change it.
  3. What is recorded? Ask to see the audit trail for an AI-assisted action: what was produced, by what process, and who approved it.
  4. Where does the data go? Ask where client data is processed and stored when the AI uses it, and under whose controls.
  5. Can the firm switch it off? Ask whether the firm can limit AI to certain work, teams or stages.
  6. What does it not do? A vendor who can describe the limits of its AI clearly usually understands it well.

Our checklist for choosing legal case management software places these questions alongside the rest of a system review.

Where tsuro fits

tsuro's AI does the work inside the matter: it drafts from the matter's data and your templates, reviews and checks, assembles the report on title and prompts the next step. A fee earner reviews and approves its output, and every AI-assisted action is in the audit trail with a record of who approved what, supporting the COLP's oversight. Our page on AI software for solicitors explains how that works in practice.

See AI working inside a matter

Book a demo and we will show you drafting, review and approval on matters like the ones your firm runs.