A project you can finish before lunch
A finished checklist gives you something concrete to inspect. Every item has to come from somewhere, missing details are easy to spot, and you can decide whether the result would make your day easier. That makes it a useful place to begin.
Our suggested first project is smaller: take a fictional pile of notes and turn it into a checklist. “Replace the hallway bulb. Ask Sam about the ladder. Measure the shelf before buying brackets.” You can see immediately whether the assistant lost an item, invented a purchase, or made the plan easier to follow.
Give the task an edge
A good boundary is part of the assignment. Ask for a checklist using only those notes. Put missing details under “Questions.” Ask it not to suggest products yet. You have separated organizing the work from deciding what to buy, and that makes the result easier to inspect.
Count the cleanup
A fast response is only one part of the cost. Count the time spent explaining the task, checking the answer, and correcting errors. If the AI saves ten minutes but creates fifteen minutes of review, you have learned something valuable. It may still be useful for brainstorming, but it has not earned a permanent place in that workflow.
Make one change at a time
If the first result is too vague, ask for a concrete verb at the beginning of every item. If it invented details, tighten the evidence boundary. Keep the same notes while you change the prompt. Changing the model, prompt, and input together makes it hard to see what helped.
Keep the useful bit
Keep a prompt when it helps on several examples, and save a note about the details you still check yourself. Start a folder with the task, the prompt, one useful result, and one mistake to watch for. You can build a practical reference from those small records. This exercise is a suggestion; results vary with the model, input, and your review.