Asking Questions Is the Scarcest Skill of This Era
Ever had this experience?
Open ChatGPT, type "make me a website," and AI responds with a bunch of stuff you don't understand. You think: "AI isn't that great after all."
The problem isn't that AI can't deliver. The problem is you don't know how to ask.
This might sound harsh, but it's the truth: Same AI, same topic, different phrasing — the results can be worlds apart.
What's the difference? Whether your question can "be executed."
One Example That Explains Everything
Let's say you want to build a calculator.
Approach A:
Make me a calculator
What will AI give you? Probably the most basic thing — four buttons, an input field, maybe it works, maybe it doesn't. It doesn't know what kind of calculator you want, so it can only guess.
Approach B:
Make me a calculator that:
- Can add, subtract, multiply, and divide
- Supports chained operations (can keep pressing numbers after an operator)
- Records the calculation history (shows each step to the user)
- Works in a web browser
- Minimal, clean style
Same AI — Approach B gives you a usable product straight away.
What's the difference? It's not that you typed more words. It's that before asking, you thought clearly about what the end result should look like.
Begin with the End in Mind
Stephen Covey said in The 7 Habits of Highly Effective People: Begin with the end in mind.
This isn't management fluff. It's the core methodology for communicating with AI.
How most people talk to AI:
I have a vague idea
↓
Throw it at AI directly
↓
AI gives a vague answer
↓
"AI sucks"
The correct approach:
What should the end result look like?
↓
Is it technically feasible?
↓
How much will it cost?
↓
Is it worth doing?
↓
Think all of this through, then ask AI
↓
AI gives you an actionable answer
You're not asking "make me X" — you're asking "make me an X that [specific description of the outcome]."
This difference is the line between amateur and professional.
Why Most People Don't Know How to Ask Questions
Because our education system never teaches it.
Asian education trains us to do one thing: Find the correct answer. And the metric for how well you find it is your score.
100 points = you're great. 60 points = you're not.
The entire system encourages "don't make mistakes" rather than "try more." But real learning comes from the number of failures, not from scores.
This creates a serious consequence: most people are afraid to ask questions. Because asking = exposing what you don't know = possibly getting penalized. So we learn a strategy — ask less, guess more, wait for someone to give us the answer.
From childhood, education trains us to answer questions: teacher asks, you answer. Exam poses questions, you respond.
But nobody ever taught you how to ask questions.
The workplace is the same. Boss says do something, you do it. Rarely does anyone ask: "Why are we doing this? What should the end result look like?"
The result: when AI gives you an "ask me anything" opportunity, most people don't know what to ask.
But the good news: asking is a skill, and skills can be practiced. You don't need to be "smart enough" to ask good questions — you just need practice. And AI is the best practice partner — it won't laugh at you, won't penalize you, and you can ask a hundred times.
The Four-Layer Structure of Asking
Before every AI conversation, run through these four questions in your mind:
Layer 1: What Do I Want? (Outcome)
Not "I want to make an app," but:
- What problem does this app solve?
- Who will use it?
- After using it, what should the user get?
The more specific, the better. If you can't articulate what you want, AI definitely can't help you.
Layer 2: What Are the Constraints? (Conditions)
- What's the budget? ($0? Or can you spend a little?)
- How much time? (Need it today? Or within a week?)
- Technical constraints? (Must run on mobile? On web?)
- Style preference? (Minimal? Professional? Cute?)
Constraints aren't bad. Constraints help AI narrow the scope and give more precise answers.
Layer 3: What's the Background? (Context)
- Is this a new project, or modifying something existing?
- What approaches have you tried before? Why didn't they work?
- Any reference examples? ("I want something like XX")
The more context, the less likely AI gives you an answer completely irrelevant to your situation.
Layer 4: How to Verify? (Criteria)
- After it's done, how do you know it's correct?
- What counts as "complete"?
- Are there quantifiable standards?
Many people skip this layer, but it's important. If you don't define "what is good," AI will just give you something "close enough."
Real-World Examples: From Bad Questions to Good Questions
Example 1: Writing Copy
❌ Bad question:
"Write me a marketing copy"
✅ Good question:
"Write me a Threads post about AI automation.
Target audience: office workers who want to start a business.
150-250 words, start with a counterintuitive hook.
Tone should be like friends chatting, not like selling.
End with an open question."
Example 2: Building a Product
❌ Bad question:
"Make me a website"
✅ Good question:
"I want to build a personal brand website:
- Single page, scrollable
- Sections: About me, Services (3), Portfolio, Contact form
- Style: Dark background, techy feel, sans-serif fonts
- Tech: React + Tailwind, deploy to Vercel
- Form submission should email me a notification
Plan the architecture first, confirm with me, then start building."
Example 3: Solving a Problem
❌ Bad question:
"My code is broken"
✅ Good question:
"I'm getting this error when running npm run build:
[paste error message]
My environment is Node 18 + Vite + React.
I just modified [specific file]'s [specific part].
It was working before the change."
See the pattern? Good questions always include: Outcome + Constraints + Context.
An Overlooked Practice Method
Most people only practice "asking" when using AI. But actually, you have opportunities every day.
Start asking questions to people around you.
Next time a colleague says "this project is kind of stuck," don't just reply "mm." Try asking:
- "Stuck at which step?"
- "Is it a resource shortage or unclear direction?"
- "What do you think ideal progress should look like?"
These questions are essentially the same as what you ask AI: figure out what the problem is, where the constraints are, what the expected outcome is.
Practice asking people, and you'll get better at asking AI too. Because the core skill is the same: turning vague things into clear things.
Prompting Checklist
Before every AI conversation, quickly run through this:
□ Do I know what the end result should look like?
□ Have I stated my constraints clearly?
□ Have I provided enough background context?
□ Do I know how to judge if AI's answer is good?
□ Can my question "be executed"?
(Can AI start working immediately, or does it need to ask me ten more questions?)
If all five are checked, your question quality already exceeds 90% of people.
One Last Thing
AI won't replace you. But people who can use AI will replace people who can't.
And the core of "being able to use AI" isn't memorizing prompt templates — it's whether you can ask questions.
Tools change, models get updated. But the ability to ask — breaking down vague ideas into clear instructions — that skill stays with you for life.
Starting today, before every AI conversation, spend 30 seconds thinking: What should the end result actually look like?
Those 30 seconds will save you 30 minutes.
This is part two of the "Getting Started" series. Previous: AI Development for Beginners: From a Smartphone to Shipping Products
Want more free resources? Join the Solo Lab Discord.
Discord: https://discord.gg/ewS4rWXvWk
Originally published on Ultra Lab — we build AI products that run autonomously.
Try UltraProbe free — our AI security scanner checks your website for vulnerabilities in 30 seconds: ultralab.tw/probe
Top comments (0)