You've been doing your job for 15 years. Someone asks you to write instructions so an AI can handle part of it. You sit down, open a blank document, and realize you can't. Not because the work is complicated. Because you stopped thinking about it years ago.
This moment is happening in every organization right now. Companies are investing in AI tools, running pilots, hiring consultants. And they keep hitting the same wall: their best people can't explain how they do what they do.
Most leaders treat this as a training problem. Teach people to write better prompts. Give them templates. Run a lunch-and-learn.
It is not a training problem. It is a knowledge problem.
The four stages of not knowing what you know
Skills develop through four stages. Stage 1: you don't know what you don't know. Stage 2: you know what you don't know (the painful part). Stage 3: you can do it, but you have to think about every step. Stage 4: the skill has gone underground. You just do it.
Stage 4 is where your most experienced professionals live. That senior analyst who "just knows" when numbers don't look right. The project manager who senses a stakeholder problem before anyone has said a word. They are running pattern-recognition engines built over years of experience. But those engines run below conscious awareness.
Have you ever watched an expert try to train a new hire? They skip steps they don't realize they're taking. They say things like "you just have to develop an eye for it." That is not laziness. That is Stage 4 talking.
The delegation paradox
Here is the uncomfortable truth: expertise and explainability move in opposite directions. The better you get at something, the worse you get at describing how you do it.
This is why new hires struggle even after "thorough" onboarding. It is why process documentation never captures the whole picture. It is why knowledge transfer projects produce binders that collect dust.
Organizations have always absorbed this cost through longer ramp times, more supervision, and occasional failures chalked up to "experience you can't teach." The gap has been there forever.
AI made it visible.
When you delegate to a human, they bring their own judgment. They fill in your gaps. They ask clarifying questions. They watch how you react and adjust. AI does none of that. It takes your instructions literally. Every gap in your description becomes a gap in the output. Every unstated preference gets ignored.
Stop and think about the last time you looked at AI output and said "that's not what I meant." The AI did exactly what you asked. The problem is that what you asked was not what you meant, because what you meant was never in words.
This is why better prompting only gets you so far. The bottleneck is not the interface between you and the AI. It is the interface between your conscious and unconscious knowledge.
The bottom line
The barrier to AI adoption is the same barrier that has always existed for delegation, training, and knowledge transfer: your best people can't describe what makes them good. AI just raised the stakes. The organizations that invest in extracting and articulating that expertise will outperform the ones chasing better tools. The knowledge is already there. It is just buried.