Start With An Interview, Not A Draft
Most weak AI answers come from a simple problem: the AI starts writing before it understands what you actually need. If you type “write an email to my team,” it has to guess the tone, the purpose, the audience, the deadline, and what counts as success. Those guesses may sound polished, but they are still guesses.
A better first move is to make the AI interview you before it answers. Use this prompt: “Before you draft anything, ask me one question at a time until you understand my goal, audience, constraints, and any missing details. Then summarize what you learned and ask for permission to draft.” This turns the chat from a vending machine into a short planning conversation.
Use One Question At A Time
The “one question at a time” part matters because it keeps the process easy. If the AI asks eight questions at once, you may answer vaguely or skip the hard ones. One focused question forces useful detail. For example, before writing a complaint email, it might ask, “What outcome do you want from the company?” That is better than immediately guessing whether you want a refund, apology, repair, or record of the issue.
Answer briefly, then let it ask the next question. Good questions usually cover four areas: goal, audience, constraints, and missing facts. Goal means what the result should accomplish. Audience means who will read it and what they care about. Constraints are limits like length, tone, deadline, budget, or format. Missing facts are details the AI needs but you have not provided yet.
Make It Summarize Before It Writes
Once the AI has asked enough, do not let it rush straight into the final version. Ask for a short summary first: “Summarize my goal, audience, constraints, and key details in bullets. If anything is still unclear, ask one more question. Otherwise, wait for my go-ahead.” This creates a checkpoint, which is a pause where you can catch errors before they spread into the draft.
This step prevents a common failure mode: the AI confidently writes the wrong thing. Maybe it assumes your email should sound formal when you need warm and direct. Maybe it builds a travel plan around museums when your real priority is keeping kids rested. Correcting a short summary is easier than repairing a full answer built on the wrong assumptions.
Know When To Stop Interviewing
The interview should be useful, not endless. For a simple task, two or three questions may be enough. For a higher-stakes task, such as a resignation letter, a budget plan, or a message to an upset client, let the AI ask more. The test is whether another question would change the answer in a meaningful way. If not, tell it to draft.
You can also set a limit: “Ask no more than five questions unless something critical is missing.” This keeps the AI from over-planning. The point is not to make prompting complicated. The point is to slow down for a minute so the AI can stop filling gaps with generic guesses and start reflecting your actual situation.
Key takeaways
- Use the prompt: “Before you draft anything, ask me one question at a time until you understand my goal, audience, constraints, and missing details.”
- One-question interviews work better than long question lists because you are more likely to answer clearly.
- Have the AI summarize what it learned before it drafts so you can correct bad assumptions early.
- Use more interview questions for higher-stakes writing, planning, or research tasks.
- Set a question limit when speed matters, such as “Ask no more than five questions unless something critical is missing.”