Treat AI like a partner, not a vending machine.

A vending machine takes one input and gives one output.

Type a request, get a result, move on. That works for simple lookups. For anything with nuance, it produces average work, because the assistant has nothing to be specific about. The output is only as particular as the input.

A partner needs to know what you know.

Tell it who the work is for, what good looks like, what you have already tried, and what you do not want. Two sentences of context change the answer more than any clever phrasing. Think of how you would brief a new teammate on their first day.

Ask it to push back.

Try: “What is weak about this plan? What am I assuming that might be wrong? What would a skeptical customer say?” An assistant is often more useful as a second set of eyes than as a first-draft machine.

Divide the work honestly.

Give the assistant the heavy lifting of reading, sorting, drafting, and comparing. Keep for yourself taste, relationships, priorities, ethics, and the final word. A partnership works because each side does what it is good at.

Stay the owner.

You are accountable for what goes out under your name. Read it, check the facts that matter, and edit until it sounds like you. The goal is not to disappear from your own work. It is to spend your attention where it counts.