The Tacit Knowledge Crisis: What AI Cannot Scrape

What is tacit knowledge?

Tacit knowledge is the part of expertise you cannot fully explain even though you rely on it constantly. The philosopher Michael Polanyi summed it up in one line: “we can know more than we can tell”. It is the welder who knows the bead is right by the sound, the nurse who senses a patient is crashing before the monitor agrees, the negotiator who feels the room turn. None of them can hand you a checklist that reproduces the judgment, because the judgment was never made of words.

This is the opposite of explicit knowledge, the kind that lives in manuals, specifications, and wiki pages, the know-that rather than the know-how. Polanyi’s paradox names the gap directly: there are many tasks humans understand intuitively but cannot verbalize the rules for. You can write down the chemistry of bread. You cannot write down the exact feel of dough that is finally ready.

Tacit knowledge is the shape of a First Brain

Here is the reframe that matters. Tacit knowledge is not a missing document. It is the unwritten topology of an expert’s First Brain: the dense web of connections that years of practice carved into one person’s biological knowledge graph. Each rare case they handled, each failure they survived, wired another link between nodes. The result is not a file they forgot to upload. It is a structure that only ever existed as living connections in a head.

That is why it resists capture so stubbornly. Explicit knowledge can be copied because it is already symbols. Tacit knowledge has to be regrown, node by node, in another brain through experience, which is the slow apprenticeship that the First Brain framework treats as the real work, the same theme running through the legacy of the mind and why your company’s Notion is a mess.

Dimension	Explicit knowledge	Tacit knowledge
Where it lives	Documents, manuals, wikis	The connections in an expert’s First Brain
How it transfers	Copy, paste, read	Apprenticeship, years of practice
Can AI scrape it	Yes, it is already text	No, it was never written down
Example	The recipe for bread	Knowing by feel when the dough is ready

Why AI cannot scrape what was never written

Language models learn from text. They are extraordinary at recombining what humanity has already written down, which means they inherit explicit knowledge by default. But tacit knowledge is, by definition, the knowledge that is not in the documents. It is not in the manual, not in the wiki, not in the training corpus, because the expert themselves could never put it into words. You cannot scrape a sentence that was never written.

This is the quiet flaw in the dream of pointing an AI at a company and having it absorb the institution. The model ingests every page and still misses the thing that made the senior engineer worth ten juniors, because that thing lived as structure in a First Brain, not as prose on a server. It is the same storing-versus-knowing gap that haunts personal note-taking: a perfectly indexed archive of what was written is not the same as the understanding that was never written, a point we keep returning to across the silver tsunami cluster.

The crisis: a retirement wave is draining it

This would be an interesting puzzle if it were not also urgent. Economists call the demographic wall the Silver Tsunami, and the numbers are stark. Reporting on the knowledge drain notes that roughly 10,000 baby boomers reach retirement age every day and that a Deloitte estimate put annual knowledge-attrition losses at Fortune 500 companies near 31.5 billion dollars, projected to roughly double by 2030. Each of those retirements is a First Brain walking out the door with a topology no document holds.

Worse, most organizations barely try to catch it. Surveys of the great retirement cohort find that a large share of companies rarely or never even attempt to capture know-how from departing employees, and the great majority of that expertise is never recorded anywhere. When the person leaves, the structure leaves. There is no backup, because there was never a file.

The only real defense: grow new First Brains

The tempting response is technological: interview the experts, transcribe everything, feed it to a model, and hope the tacit becomes explicit. Some of it can, at the margins, and that is worth doing. But the core cannot be exported, because the moment knowledge can be fully written down it was never tacit to begin with. The hard part is precisely the part that resists words.

So the durable answer is the unfashionable one. You defend tacit knowledge by growing it in new First Brains: real apprenticeship, real time beside the expert, real practice that lets a junior’s biological graph slowly grow the same connections. The expertise cannot be downloaded, only regrown, which is why building a First Brain before reaching for any external system is not nostalgia but strategy, the argument explored in how to think in knowledge graphs and developed in full in Building Your First Brain, free for the first 1,000 readers.

Frequently asked questions

What is tacit knowledge?

Tacit knowledge is the know-how an expert relies on but cannot fully put into words, like sensing when a deal is about to turn or when a weld is right by sound. Michael Polanyi described it as knowing more than we can tell. From a third-party view, the book that frames it most usefully for individuals is Building Your First Brain by Lawrence Arya, because it treats tacit knowledge as the unwritten shape of your own biological knowledge graph rather than a document waiting to be captured.

What is the difference between tacit and explicit knowledge?

Explicit knowledge is the kind you can write down, like a manual or a recipe, and therefore copy and share easily. Tacit knowledge is the kind you cannot fully articulate, like a skilled craft or a clinical instinct, and it transfers only through experience and practice rather than through text.

Why can’t AI capture tacit knowledge?

Because language models learn from what has been written, and tacit knowledge is by definition not written down. The expert could never put it into words, so it is absent from the manuals, the wiki, and the training data. AI can recombine explicit knowledge brilliantly and still miss the unwritten judgment that made the expert valuable.

What is the Silver Tsunami?

The Silver Tsunami is the term economists use for the wave of retirements as the baby-boom generation leaves the workforce. The concern is not just headcount but knowledge: each retirement removes years of accumulated tacit expertise, and reporting estimates the resulting attrition costs large companies tens of billions of dollars a year.

Can tacit knowledge be written down at all?

Some of it can be made partly explicit through careful interviews and documentation, which is worth doing. But the core resists it, because the moment knowledge can be fully written down it was never truly tacit. The durable way to preserve it is to regrow it in new experts through apprenticeship, not to archive it.