Most people struggle with AI in the same way. They type a request, get a bland, generic answer, and decide the tool “isn’t that smart.”
The tool is usually fine. The problem is that it’s guessing. You asked for “a marketing plan for my bakery,” and the AI has no idea whether your bakery is in Chicago or a village, whether you have $100 or $10,000, or whether you want customers online or walking through the door. So it fills the gaps with the safest, most average assumptions.
You can fix this with one sentence.
The trick: flip the roles
Instead of trying to cram every detail into your first message, tell the AI to ask you first:
“Before you answer, ask me the questions you need to give me the best possible result. Ask one at a time.”
That’s the whole technique. The AI stops guessing and starts interviewing you. By the end, it knows your situation, your goals, and your limits, and the final answer feels written for you rather than for “a person.”
See the difference
The usual way:
“Write a weekly meal plan for me.”
→ You get a generic plan with salmon, quinoa, and roasted vegetables, which is fine for nobody in particular.
The interview way:
“I want a weekly meal plan. Before you write anything, ask me the questions you need, one at a time.”
→ The AI asks: How many people are you cooking for? Any allergies or foods you hate? How much time do you have on weekdays? What’s your budget? Do you have a slow cooker or an air fryer?
→ You answer in a few words each. The plan you get fits your kitchen, your schedule, and your wallet.
Two minutes of answering questions replaces twenty minutes of editing a mediocre result.
Where it works best
- Business tasks: business plans, pitch decks, job descriptions, customer emails, pricing strategies. The more context matters, the bigger the gain.
- Learning: ask the AI to quiz you about what you already know before it explains a new topic, so it skips the basics you’ve mastered.
- Big decisions: choosing a laptop, planning a trip, comparing job offers. The AI’s questions often make you realize what you actually care about.
- Writing: speeches, cover letters, and posts, where your voice and your story are the point.
Three upgrades
1. Limit the questions. If the AI gets carried away, add: “Ask no more than five questions.” It keeps the process fast.
2. Let it skip what it can infer. Add: “If you can reasonably assume something, tell me your assumption instead of asking.” This saves time on obvious details.
3. Ask for a summary before the answer. Say: “When you’re done asking, summarize what you understood in three lines, and I’ll confirm.” Misunderstandings get caught before the AI writes 1,000 words in the wrong direction.
A ready-to-use template
Copy this and swap in your own task:
“I need help with [your task]. Before you begin, interview me: ask one question at a time, no more than six in total. Then summarize what you understood, wait for my confirmation, and only then give me the full answer.”
Why this works
A language model can only work with what’s in front of it. A vague request gives it almost nothing, so you get an average answer. The interview method draws out the details you didn’t think to mention, including some you didn’t realize mattered, and hands the AI a proper brief.
It’s also a more natural way to work. You wouldn’t hire a designer who started drawing before asking what the logo is for. Treat AI the same way and the results get noticeably better.
Try it today: pick one task you’ve been putting off, paste the template, and answer the questions honestly. Then compare the result with what you’d normally get. Most people don’t go back.