I’ve been thinking about how teams actually get better with LLMs. I do not think the answer is one perfect prompt. It is a set of working habits that make the tool part of a disciplined process.

1. Start your work with the frontier tools.
2. Give the model a real brief.
3. Ask it to help you think, plan, review, test, and improve.
4. Use the right model for the job.
5. Delegate repetitive work that drains time and attention.
6. Treat the model as a working partner rather than a search box.
7. Do not use it as a replacement for judgment.
8. Do not blindly trust the output.
9. Use the approved tools available to you.
10. Review the result and own the final work.

The list is intentionally practical. Start with the work in front of you, learn by building, and keep confidential, restricted, personal, or regulated information out of unapproved tools.

The model can help create the work. The person using it still owns what leaves the room.