2 min read
The short version
- Describe the situation, not just the request.
- Ask for options rather than one answer.
- Tell it what was wrong and ask again — that is the whole skill.
The blank box is the hardest part. Not because the tool is complicated, but because it gives you no clue what it is for. Here is a five-minute starting point that produces something useful on the first try.
Start with a real problem from today
Not a test. Something actually on your list: what to cook on Thursday, a message you are avoiding writing, an afternoon to fill. Real problems produce answers you can judge.
Describe the situation, then the request
Most disappointing answers come from requests with no context. Compare:
“Give me some dinner ideas.”
“Two adults and a three-year-old. I have 25 minutes on weeknights. Nobody eats fish. Give me five dinner ideas using ordinary supermarket ingredients.”
Same question. Completely different answer. The second one you can cook from.
Ask for a number of options
“Give me five” is better than “give me one”, because you get to choose rather than accept. It also makes a weak suggestion obvious, since it sits next to four others.
Then say what was wrong
This is the actual skill, and it is the step people skip. “Too complicated, I have 15 minutes not 25.” “She will not eat anything with sauce.” “Cheaper.” You are not starting again — you are steering. Two or three rounds of this gets you somewhere genuinely useful.
What to expect in five minutes
Not magic. A decent first draft of something you would otherwise have done from scratch, arrived at faster. That is the whole proposition, and it is enough.
What these tools do badly is covered in our piece on where AI falls down.