How to get your team to use AI the same way

Why teams drift

Microsoft and LinkedIn's 2024 survey found that 78% of people who use AI at work bring their own tools. On a small team that means one person on ChatGPT, another on Claude, each with their own pasted instructions, each producing a different version of the same client email. Nobody chose this; it's what happens when everyone figures it out alone.

The fix pays off most for newer staff. In the NBER study of a support team, the biggest gains went to the least experienced agents, because the AI passed on what the best agents did. A team gets that effect only if the best agent's method is what everyone's AI is using.

A worked example

Illustration: a made-up business. A 9-person travel agency notices that three advisers send quote emails that look nothing alike. One is two lines, one is a page, one forgets the cancellation terms. Each uses AI, differently.

• The tasks they pick: the quote email, the pre-departure checklist, and the reply to a complaint.
• The method: the senior adviser, whose quotes convert best, explains on a 15-minute call how she writes one. The call notes become a procedure: what to confirm first, the order of sections, the cancellation line word for word, one real quote with names removed.
• One copy: the procedure becomes shared instructions that each adviser's AI uses, so they type "quote email for the Lisbon family trip, notes below".
• The owner: the senior adviser, with a review every quarter, or sooner when supplier terms change.

After a month, every quote has the cancellation terms, and a new adviser's first quotes look like the senior adviser's. When the airline changed its baggage rule, she updated one copy.

Ways to keep one copy

Plan details from OpenAI's and Anthropic's help pages, read 5 November 2026. More on the options: three ways to share prompts with a team.

A two-week plan

1. Day 1: list tasks. Ask everyone which tasks they use AI for. Pick the three that most people do and where results differ most.
2. Days 2 to 4: write each method. Have the person who does it best explain it out loud for 15 minutes, then turn the notes into steps, a template and one example.
3. Day 5: agree the rules. Which tools, what never gets pasted, who reads before sending. See what not to paste.
4. Week 2: one copy. Put the three methods in one shared place each person's AI reads, and test each on three real requests.
5. End of week 2: retire the old copies in notes and personal instructions, and name the owner of each method.

Next steps

The method you wrote is also your new hire's first-day material: teach a new hire your team's AI workflow on day one. When you're ready to turn a written method into a skill the AI picks by itself, make your first Claude skill from a doc you already have walks through it, and AI skills, explained covers the idea.

The AI skill below does step 2: give it a colleague's spoken or written explanation, or a screen-recording transcript, and your AI writes the procedure with steps, decision points, checks, an owner and a review date.

Questions

How do I get my team to use AI consistently?

Pick the few tasks everyone does, write down how your team does each one, and keep one copy of those instructions that every person's AI uses, with an owner who updates it. Consistency comes from sharing the method, not from training people to write better prompts.

Should everyone on the team use the same AI tool?

It helps, but it's rarely true in practice: people bring their own tools and accounts. Share the instructions in a way that reaches every tool your team uses, rather than forcing one tool first.

Who should own the team's AI instructions?

The person who'd own the task without AI: whoever is responsible for client emails owns the client-email instructions. Give each one a review date, and let anyone suggest a change.