Tacit Knowledge Management: Drawing Out What People Cannot Easily Say

Reading Time: 12 minutes

Quick Answer: Tacit knowledge management is the practice of surfacing, sharing and keeping the judgement people build through experience but rarely put into words. Because much of it cannot be fully written down, the work is less about storing documents than drawing knowledge out: asking better questions, learning from real cases and making it safe to share. AI now helps by learning patterns directly from examples of the work.

The philosopher Michael Polanyi built his account of tacit knowledge on a simple observation: we can know more than we can tell. Most advice on tacit knowledge management quietly starts from the opposite assumption. It treats what people know as something waiting to be converted into documents and filed in a knowledge base, as if the only obstacle were finding the time to write it all down.

That approach answers the easier half of the problem. The harder half is getting the knowledge out in the first place, and part of it will never fit in a document at all. Good tacit knowledge management starts from that limit rather than ignoring it: what kind of knowledge is at stake, which method will surface it, and what should stay with people. Recent advances in AI have also moved the line between what can be captured and what cannot.

What Is Tacit Knowledge Management?

Tacit knowledge is judgement built through experience: knowing what matters, noticing when something is off and deciding what to do next, often without being able to explain exactly how. A planner can tell a forecast is wrong before checking the numbers. An account manager hears risk in the way a client phrases a routine email. A support agent knows which question to ask first, and it saves twenty minutes.

Tacit knowledge management is the deliberate effort to surface that judgement, spread it to the people who need it and keep it inside the organisation when individuals move on. Traditional knowledge management grew up around repositories: intranets, wikis, procedure libraries and search. Tacit knowledge management centres on people, practice and conversation instead, because that is where this kind of knowledge lives. It sits within the wider discipline of capturing institutional knowledge, but deals with its hardest part: the know-how people use every day and rarely say out loud.

Why Tacit Knowledge Resists Being Written Down

The standard advice in tacit knowledge management is to convert tacit knowledge into explicit knowledge. That is partly right. Some of it converts well, and a good share of what people call tacit is simply unrecorded. The mistake is treating all of it the same way, when several things work against writing it down:

• It is triggered, not recalled. Know-how surfaces when a situation calls for it. Ask an expert to describe their job and they describe the routine; the judgement appears when something unusual happens.
• Experts compress. Steps that feel obvious to an experienced person drop out of their explanation altogether, because they stopped noticing them years ago.
• Some of it is perception, not rules. Noticing that a figure looks wrong, or that a customer is about to leave, comes before any rule about what to do next.
• It loses its context on paper. A written rule rarely carries the conditions under which an experienced person would break it.

So the useful question for tacit knowledge management is not how to write everything down. It is which part can be drawn out, by what method, and which part will stay with people however carefully it is documented. Answering that starts with recognising that unwritten knowledge is not one thing.

Three Kinds of Unwritten Knowledge

Tacit is often used as a label for everything that is not written down. In practice that label covers three different situations, and each responds to a different approach. Treating them as one is why so many tacit knowledge management efforts produce plenty of documentation and very little change.

Knowledge Nobody Has Asked For

Much of what organisations describe as tacit is knowledge people could put into words if anyone asked. The finance lead who always runs two extra checks before month-end could list them in a minute. Nobody has ever prompted her, and a blank template headed ‘describe your role’ will not either. This is the easiest kind for tacit knowledge management to surface. Direct, specific questions draw it out, while general requests to document a job invite people to describe the routine rather than the exceptions.

Knowledge That Only Shows in the Doing

The second kind is judgement that genuinely resists being reduced to a rule. An experienced customer manager can tell which complaint signals a real risk of losing the account and which is simply a bad day, but cannot write a checklist that would let someone else make the same call reliably. This knowledge becomes visible in real cases, in watching the work and in comparing many examples side by side. It will never be fully written down, but it can be partly learned from examples, which is where AI has started to change what tacit knowledge management can achieve.

Knowledge People Choose to Keep

The third kind is not unknown but held back. People keep workarounds and hard-won insight to themselves for understandable reasons: it makes them valuable, they are unsure how it will be used, or they doubt that sharing it will benefit them. That is often a rational choice rather than a difficult attitude, and treating it as a compliance problem tends to make it worse. It surfaces when capture is confidential, the purpose is clear and contributors get credit. It is also the knowledge most likely to leave with someone, which makes it central to preventing knowledge loss.

Tacit Knowledge Management in Practice: How to Draw It Out

In practice, tacit knowledge management works best when the method matches the kind of knowledge. The table sets out what tends to draw each kind out, and what tends to fail.

Kind of knowledge	Typical example	What draws it out	What tends to fail
Never asked for	The two checks a finance lead always runs before month-end	Specific, direct questions	Blank templates and ‘document your role’ requests
Only shows in the doing	Knowing which complaint signals a real risk of losing a customer	Walking through real past cases, observation, many examples	General questions about the process
Kept back	A workaround someone relies on but has never shared	Confidentiality, a stated purpose, credit for contributors	Mandatory capture drives and rushed exit handovers

Across all three, the quality of the questions matters more than the format of the output. A few techniques consistently draw out more than general interviews do:

• Ask about the last specific case. ‘Walk me through the last time an invoice looked wrong’ surfaces judgement that ‘how do you process invoices?’ never will.
• Ask what they noticed first. Then ask what would have changed their decision. Cues and turning points are where judgement sits.
• Ask what a capable new starter would get wrong. People find it easier to describe mistakes than their own expertise.
• Work from real artefacts. The actual email, report or ticket prompts far more detail than memory alone.
• Speak to several people who do the same job. Where their answers differ is usually where the judgement lives.

Record answers in people’s own words and check them back with the person, since a summary written by someone else tends to smooth out exactly the detail tacit knowledge management exists to keep. Knowing where this knowledge sits across teams is also a useful input to an AI readiness assessment, because it shows which work is understood well enough to hand over.

What AI Changes About Tacit Knowledge Management

For most of the history of computing, software needed explicit instructions. If nobody could state the rule, a computer could not follow it, which is why tacit knowledge sat out of reach and tacit knowledge management leaned so heavily on codification. Machine learning works differently. It infers patterns from examples rather than rules, so knowledge that nobody can fully articulate can still be learned, at least in part, from records of the work itself.

A study in the Quarterly Journal of Economics shows what that looks like in practice. The researchers followed 5,172 customer-support agents at one firm after the introduction of an AI assistant trained on past conversations, weighted towards the firm’s top performers so that it picked up behaviours they applied tacitly. Issues resolved per hour rose 15 per cent on average and around 30 per cent for less skilled and less experienced agents. Agents with two months’ tenure and access to the assistant performed as well as agents with more than six months’ tenure without it. The difference from documentation is clearest side by side.

	Written rules and documentation	Learning from examples
What it needs from experts	Time to write down what they do	Examples of the work itself
What it captures well	Stable, explicit steps	Patterns in judgement, tone and sequence
What it misses	Cues, exceptions and when to break the rule	Rare cases with few examples
How it stays current	Someone rewrites it	New examples keep feeding in
Main risk	Goes out of date unread	Copies habits, good and bad

In effect, examples spread tacit practice from the best performers to everyone else in a way documentation had not. It is one firm in one occupation, so it shows a direction rather than a guarantee, but it widens the realistic scope of tacit knowledge management. The repeatable part of work learned this way can now be carried out by an AI agent rather than written up for people to read.

The Catch: Experts Still Have to Keep Contributing

The same study contains a warning. The most skilled agents also came to follow the AI’s suggestions, and the researchers noted that fewer original contributions from top performers could make future versions of the model less effective at new problems. Examples only teach what is already in them. That makes tacit knowledge management with AI a continuous practice rather than a one-off extraction: new judgement has to keep flowing in from the people who have it, and the result should be judged on outcomes over time rather than on how much was captured.

Drawing Out Tacit Knowledge Across Every Team

GrowthNation is built around this kind of work. It interviews every team member through a confidential voice or text agent, asking about the specific cases and exceptions that written templates miss, which reaches both the knowledge nobody has asked for and the knowledge people would otherwise keep. The repeatable work that surfaces becomes agents served centrally, so the know-how stays with the organisation rather than in individual accounts. Hours saved are tracked per agent, so tacit knowledge management is judged by what it changes.

See what one team knows but has never written down. Book a walkthrough: we will interview a single team, show you the scorecard and hand you a working agent. No deck.

Frequently Asked Questions