Part 1. Teach your AI how you do things here

Pick one piece of work your team already does well. Sit down with the person who does it best and capture five things.

• The purpose: what this piece of work is for, and who reads it.

• Two examples: one you're proud of, and one that missed the mark, with what was wrong with it.

• The judgement calls: where that person stops and decides, and what they weigh up.

• Your standards: tone, length, and what must always be included.

• The lines: what it must never do, such as promise a date, quote a price or share customer details.

Save it as standing instructions in the AI tool your team uses: a Claude skill or project, ChatGPT custom instructions or a custom GPT, a Gemini Gem, or Copilot's saved instructions. Test it on next week's real work and compare the result with what your best person would have produced.

Part 2. Three lists before you book any training

General training produces general fluency, which isn't the same as getting better at the job itself. Asking your team what training they need rarely helps either, because most people believe they're already fine. The training call is yours, and it comes down to three lists.

1. What everyone needs. Smaller than you think: what AI is for in this business, what must never go into it, and that it's fine to say you used it.

2. What only some need. The roles where AI changes the work this year. Deep and narrow, on their own tasks with their own data, with someone sitting beside them.

3. What you won't train yet. The list nobody writes. Writing it is what makes the other two affordable.

The rest, from what's already happening in your business to keeping it going when you're not in the room, is in Leading Your Team's AI.