The scarcest skill in the AI era: saying it clearly
“What can be said at all can be said clearly; and whereof one cannot speak thereof one must be silent.” — Wittgenstein
Have you had this:
Twenty rounds with an AI. The conversation drifts. You finally close the window and start over.
You think the model is dumb. The more uncomfortable truth: the problem is not the model. You did not know what you wanted in the first sentence.
1. Most people do not know what they want
“Help me write a plan.”
I hear that a lot. That sentence hides at least five undefined variables:
- what kind of plan?
- who is it for?
- what problem does it solve?
- what is the budget?
- when is it due?
Every word you were too lazy to define, the model will define at random. Then you spend ten turns correcting a direction that should never have existed.
Worse, the bill is quiet. No error. No red text. The output looks “pretty much like a plan” and is not what you wanted. You still have to spend time locating what is wrong. That hidden cost is higher than rewriting from scratch.
A real case
A friend told me he wanted to do “self-media.” I asked a few questions:
- which platform?
- who is it for?
- what problem does it solve?
- how many hours a week can you put in?
After five questions he went quiet. “I guess I am not ready.”
He is not incapable. He had been driving himself with a fuzzy wish, treating “I want to do self-media” as if it were a plan. The sentence contains no concrete information.
2. Wittgenstein’s prophecy
A hundred years ago Wittgenstein said something that reads like a prophecy for the AI era:
“What can be said at all can be said clearly; and whereof one cannot speak thereof one must be silent.”
The cut is this: “cannot say it clearly” is treated as the same thing as “have not thought it clearly.”
Language is not wrapping around thought. Language is thought. How precise your words are is how precise your thinking is.
Socrates did the same job — he asked into the places you thought you already understood:
- “You said justice. What is justice?”
- “You said a good education. Who defines good?”
By the end you discover you do not know what you are talking about.
Two thousand years later the human bug is unchanged: we always think we have already thought it through.
3. AI is an amplifier of how clear you are
When you talk to a person, they fill in.
You say “handle this.” A colleague can usually guess. A manager says “tweak this.” You can read about 30% of the intent.
The unclear middle gets absorbed by manners, “chemistry,” and rework. People treat that as normal communication loss.
A model will not fill in.
It has no mind-reading, no reading the room, no “you know what I mean.”
- it takes what you said
- if you are vague, it stays vague
- it is precise only when you are
Every conversation with an AI is a live exam of how clearly you think.
If you fail it, it really is not the model’s fault.
4. The same bug shows up with people
In a meeting someone says “we need to improve user experience.”
Six people in the same room hear six things:
- the PM hears interaction flow
- the designer hears visual style
- engineering hears load time
- the boss hears DAU
The meeting ends. Everyone thinks there was consensus. Everyone goes and does a different job.
Two weeks later the direction is wrong. Another meeting.
A lot of communication cost is not disagreement. It is two sides discussing different objects without knowing it.
Wittgenstein again, a line I like:
“The limits of my language mean the limits of my world.”
How clearly you can speak is how much world you can reach.
5. How to practice saying it clearly
1. A template before you talk to an AI
Fill this before you open a chat:
Context: [what is true right now]
Goal: [what result I want]
Constraints: [time / budget / tech]
Audience: [who uses this / who reads this]
Acceptance: [what counts as done]2. A review habit
After every meeting, three questions:
- which sentence did they clearly not catch?
- which concept did I think I had explained, and they were still lost?
- what do I change next time?
Unclear speech is rarely a vocabulary problem. You did not organize language in the other person’s coordinate system.
3. Train definition sharpness
Every time you say an abstract word (“optimize,” “improve,” “upgrade”), ask:
Can I give three concrete measures?
If not, you have not thought it through.
6. Language is a craft
I have been in the study-abroad industry for almost seven years. The longer I do it, the more the job is “say it clearly.”
A practical example:
When a writing coach helps a student on a research proposal, the hard part is almost never prose. It is getting an academic claim across to an undergrad who may have no research experience.
You say “you need a more focused research question.” They hear “make the title shorter.”
The same sentence can put a galaxy between speaker and listener.
Or talking through a school list with parents — reach / target / safety:
- some parents need numbers: admit rates, past cases
- some need the logic of why the gradient is designed that way
- some need a story: a student in a similar situation last year, and where they landed
Language is a craft. The question is not what you want to say. It is what the other person can catch.
7. One underlying skill
Talking to an AI, running a meeting, doing study-abroad work, managing a team, raising a child — the bottom skill is the same:
Can you turn the blur in your head into an expression another person can catch.
That skill decides:
- how much leverage AI gives you
- how well your team executes
- how much clients trust you
- whether you can even talk well with the people closest to you
A lot of people pay interest on “I cannot say it clearly” for a lifetime and never notice the bill.
8. The point
AI will not replace people who can use tools. AI replaces people who cannot think clearly. And the first symptom of unclear thinking is unclear speech.
While everyone studies prompt tricks, memorizes templates, and hunts shortcuts, the real lever is not technique. It is clarity.
Clarity is not a gift. It is a skill you can practice on purpose.
Before you open your mouth, ask one more question:
“In this sentence, is there a word I cannot define myself?”
Not many people pass that gate.
A list you can run
If this landed, start today:
- Before the next AI chat, spend five minutes on the template (context / goal / constraints / audience / acceptance)
- After every meeting this week, two minutes: which sentence did they not catch?
- When you say three abstract words in a day, immediately: can I give three concrete measures?
In 30 days, conversations with AI should be at least 3× more efficient.
More important: conversations with people get more efficient too.
Clarity is a transferable underlying skill.
Inspired by a post from @Jaden_riku, combined with my own practice.
Comments