From the partner’s head to the firm’s method: turning know-how into skills people and AI can use

Two kinds of know-how

Most firms manage one kind of know-how and leave the other to chance.

• What lawyers draw on: precedents, standard clauses, template letters and practice notes. These are documents. Firms store them, version them and, at their best, give each one an owner and a review date.
• How the work is done: how a partner reviews a lease, what they check before exchange, their view on a limitation of liability clause, the moment they would ring the client rather than write. This is method. It lives in people’s heads, in old mark-ups and emails, and in “ask Sarah, she’s done hundreds of these”.

The second kind is what makes a firm good. It is also the kind most at risk: it leaves when people retire or move on, it varies from office to office, and juniors pick it up slowly, by watching, if they pick it up at all.

Why it matters more now

• AI applies whatever method it is given. Give it a precedent and no method, and it will apply a generic one, confidently and at scale. Give it your firm’s method, and it works the way your best people do.
• Juniors learn less by watching. As AI takes on more first drafts and first reviews, trainees see fewer files from the inside. The method has to be taught, not absorbed.
• Supervision needs something to supervise against. The SRA’s warning notice on AI makes clear that supervisors may themselves be in breach if inadequate supervision lets poor work through. A written method is what good looks like, for people and AI alike.
• Consistency across teams and offices. Clients expect the same standard wherever their matter sits.

What a skill is

A skill is a short, written method for one task. It is plain enough for a trainee to follow and structured enough for an AI assistant to use. A good one fits on a page or two and covers:

1. When to use it: the task, the type of matter, and when it does not apply.
2. The steps, in the order an experienced lawyer would take them.
3. The checks: what must be confirmed before moving on.
4. What good looks like: the standard the output should meet, ideally with a short example.
5. Red flags: the things that should make someone stop and think.
6. When to escalate, and to whom.
7. Where the approved precedents are. The skill points to the master documents in your document management system; it never copies them.

Take a tenant’s commercial lease review as an illustration. The precedent bank holds the firm’s preferred clauses. The skill holds what the partner actually does: which clauses they read first, how they test a break clause against its conditions, what they compare the repairing obligations with, which points they always raise with the client, and the handful of findings that mean the partner must see it before anything goes out.

From the partner’s head to the firm’s method

AT A GLANCE

From one person’s head to the firm’s method

Six steps to turn how your best people work into a skill that people and AI can both use.

1. 1

   Choose

   One task that is frequent, needs judgement and is done differently by different people.

   The right place to start

2. 2

   Capture

   Walk through recent files with the person who does it best: show me, don’t tell me.

   The method, in their words

3. 3

   Write

   A short, plain skill: when to use it, steps, checks, red flags, when to escalate.

   One page anyone can follow

4. 4

   Test

   Try it on past matters, with a trainee and with your approved AI, and compare.

   Proof it works

5. 5

   Own

   A named owner, a review date and an approved status, like any precedent.

   Know-how the firm can trust

6. 6

   Use

   Point trainees, supervisors and your approved AI at the approved set, and nothing else.

   The same standard on every matter

1. Choose the task. Start where the value is: work that is frequent, needs judgement, is done differently by different people, or causes rework at review. One team and three to five tasks is plenty for a first round.

2. Capture it from the person who does it best. Asking an expert to write down what they do rarely works; much of what they know they no longer notice. Instead, sit with them and walk through two or three recent files: what did you look at first, what made you pause, what would you have done if this had been different? Their mark-ups and emails are evidence of the method in action. AI can help here too: your firm’s approved assistant can interview them one question at a time and draft the structure, as long as client details stay out of the conversation.

3. Write it down plainly. Use the headings above, short sentences and the firm’s own terms. The partner reviews and corrects the draft; they should not have to write it from scratch. Keep it in plain text in the firm’s own systems, so any approved tool can read it and no tool owns it.

4. Test it. Give the skill to a trainee and to your approved AI assistant, with two or three past matters, and compare the results with what the partner did. Every gap is a missing step, check or red flag. Fix it and test again. This is the step most often skipped, and the one that turns a document into something the firm can rely on.

5. Own it. Treat each skill like a precedent: a named owner, a review date, an approved status and a record of changes. Retire skills that are out of date, because an AI will use a stale method as confidently as a current one.

6. Use it. Make the approved skills the starting point for everyone: trainees learning the task, supervisors checking it, and your approved AI assistant doing the first pass. Point AI at the approved set only, not at every draft and note on the shared drive.

What changes for the partner

Some partners worry that writing down their method gives away what makes them valuable. In practice it does the opposite. Their judgement reaches every matter of that type, not just the ones they touch; they answer the same questions less often; and juniors arrive at their desk with better first drafts. Put their name on the skill as its owner, and the credit stays where it belongs.

The hardest part is the knowledge people cannot easily put into words: “I just know when a deal feels wrong.” Examples and red flags capture more of it than rules do. Ask for the last three times they stopped a matter, and why.

Keep it safe

• Accountability stays with people. A skill guides the work; a named person still checks it and signs it off, as the SRA and the Law Society both expect.
• Respect confidentiality. A playbook written for one client may itself be confidential, and needs the same access controls as the client’s matters.
• Use only skills the firm has written and approved. Skills carry instructions. One from an unknown source is as much a risk as any other unvetted software.
• Review on a cycle. Law, practice and the firm’s appetite for risk all change; the skill must change with them.

Start this month

Pick one team, choose three tasks, and capture one skill a week with the person who does each best. Test each on past matters before anyone relies on it. By the end of the month the firm will have its first written methods, a clearer picture of where its expertise really sits, and something its approved AI can use safely. None of this needs new software. It is knowledge management, done for the age of AI.

It builds on two things we have written about before: designing professional judgement into your systems and AI, and building workflows around the results each team needs. Our free workflow capture tool uses the same interview approach for whole matter types.

We help firms capture, test and govern their know-how so that people and AI can use it safely, independently of any vendor.

If you think anything here is out of date, please tell us and we will correct it.