Ask someone to explain how they ride a bike.
Which way they lean. How far. When they correct. They've taken a thousand rides and they still can't say it in words that would let a stranger get on and ride away.
That's not a communication problem. That's how expertise works.
A philosopher named Michael Polanyi wrote it down in 1966: we know more than we can tell. He wasn't being poetic. He'd found something real, and every serious study of expertise since keeps landing in the same place.
The better you get, the less you can say
Here's the part nobody says out loud. The more fluent you get at something, the less of your own judgment you can produce as an explanation on demand.
Ask a beginner to walk you through a decision and you'll get every step, because they're still doing it step by step. Ask someone twenty years in and you'll get "it depends" or "I just know." They're not hiding anything. The steps stopped being steps a long time ago. They became one motion.
You've felt the other half of this too. Once you know something well, you can't fully remember what it felt like to not know it. Ask an expert to teach a beginner and they skip steps they don't realize they're taking, because those steps disappeared from view the moment they stopped needing them.
Watching yourself work gets harder exactly as the work gets better.
Five fields that don't read each other, saying the same thing
This isn't one researcher's pet theory. Over sixty years, five separate fields arrived at the same conclusion, working apart, in different languages, with no reason to agree with each other.
A philosopher studying how we know things (Polanyi). Two brothers studying how people move from beginner to expert (Dreyfus and Dreyfus). A sociologist standing inside real laboratories, watching scientists work (Collins). Researchers inside Japanese manufacturing firms, tracking where knowledge actually lived (Nonaka and Takeuchi). A psychologist studying firefighters making life-or-death calls in seconds (Klein).
Every one of them found the same thing: the best judgment a person has runs on knowledge that person cannot state.
When five fields that don't talk to each other keep landing on the same finding, that's not a theory looking for support. That's the evidence.
Why you can't just ask yourself
The obvious fix is to sit down and think harder about what you do. It doesn't work, and there's a reason.
Self-report is unreliable as a starting point, on its own. What you say you did and what you actually did diverge, and the only way to close that gap is to check the explanation against the real decisions and the real work, not against how the explanation feels while you're saying it.
This is why "just document your process" produces documents nobody can use. You wrote down what you remembered doing. Your actual work holds the rest, the steps that already left conscious memory before you sat down to write.
The proof, from a real advisor
One advisor I worked with slowed the conversation down every single time a client got excited about a decision. Nine of his last eleven calls. He had no idea he did it.
It was the real reason clients trusted him. Slowing down at the exact moment someone else would speed up, right when the client was ready to say yes, read as patience and care instead of a sales tactic. He'd been doing it for years. He'd never once said it out loud, because he'd never once seen it.
Someone read his actual calls and handed the move back to him with a name on it.
What actually surfaces it
Casual explanation isn't the tool for this. Structured methods are: reviewing real cases, watching for the moment a person changes strategy mid-decision, comparing many instances of the same kind of call. That's what turns up a pattern reliably, where "tell me what you do" turns up a guess.
Once a pattern like that has a name and evidence behind it, it becomes something else entirely. It can be taught. It can be priced. It can be hired against. Before it has a name, it's just something you're good at that nobody, including you, can put a price on.
Where AI actually helps
AI can read years of someone's real work, calls, decisions, client wins, at a scale no person could sit through. It can find the moves that repeat.
It cannot tell you which of those repeated moves actually matters, or what to call it, or how to use it on purpose. That part still takes a person who's done this before, reading what the AI surfaces and checking it against the real record.
AI reads. I interpret. You decide.
The thing that was already true
You weren't winging it. You built something real, and you built it so well that it stopped needing your conscious attention, which is exactly why you lost the ability to describe it on demand.
That's not a flaw in you. It's what mastery costs everyone who reaches it.
Your best work already proves what you're made of. It's just never been read back to you.
Curious what your own work would show? [See how it works →](/method)