The knowledge problem inside your own company

A company's systems are a transcript of its convenient knowledge. The CRM keeps the close date and the deal size. A ticketing system keeps the resolution time. Each field exists because it was cheap to capture, so the data you hold ends up being the data that fit neatly into a column.

What the systems miss is the reasoning. The CRM records that a deal slipped. It does not record the account manager's quiet certainty, three weeks earlier, that the buyer had already chosen a competitor and was just being polite. The ticket says the issue was resolved. Nothing captures that the agent who resolved it has now seen the same failure four times this month and suspects a pattern no report is counting. The knowledge most likely to change a decision is usually the knowledge that was never worth a field.

This is not a failure of discipline or of software. It is a property of the knowledge itself.

Michael Polanyi described this in a single line: we can know more than we can tell^*. A person who is good at something often cannot state the rule they are following, because there is no rule. There is judgment, built from thousands of cases, that arrives faster than any explanation of it. That is why rules-based systems and dashboards miss it. You cannot enter into a form what an expert cannot put into words on demand.

If this were only a philosophical observation it would not be worth your time. What makes it a business problem is that the gap is invisible from inside the system. Your dashboards have no column for what they are failing to capture. The knowledge that would change a decision sits one conversation away from the people making the decision, and nothing in the stack is built to close that distance.

The clearest measurement of this comes from a study that was not trying to prove our point. Erik Brynjolfsson, Danielle Li, and Lindsey Raymond followed 5,179 customer-support agents^* through the staggered rollout of an AI assistant and published the results in The Quarterly Journal of Economics. Access to the tool raised productivity, measured as issues resolved per hour, by 14 percent on average.

The average hides the finding. For the newest and least-skilled agents the gain was {{claim:brynjolfssonNovice:about 34 percent}}, while the most experienced agents barely moved. That skew is the signature of tacit knowledge changing hands. The tool worked by surfacing the practices of the strongest agents, the moves that lived in their heads and had never reached a script, a macro, or the CRM, and putting them in front of everyone else. Experienced agents gained little because they already knew it. Novices gained a third because they were handed knowledge the company already owned but had never been able to see.

Read that result the other way and it measures the knowledge problem directly. The expertise was valuable enough to move a hard productivity number, and until something went looking for it, it sat outside every system the company ran.

It would be easy to overclaim from one study. This was a single firm, in customer support, measuring speed and customer sentiment rather than strategic foresight or the early read on a market starting to turn. The knowledge was also captured passively, mined from existing chat logs, rather than gathered by asking people what they know. No published study has yet measured what happens when you ask frontline employees directly, and at scale, about their own work and their own view of the business. The evidence that tacit frontline knowledge carries hard value is strong. The evidence for deliberately eliciting it across functions is something the field is still building.

We say this plainly because that missing evidence is what we are setting out to generate, and we would rather name the gap than have a careful reader find it. The research settles that the knowledge exists and that it has value. Our work is to show what happens when you ask for it on purpose.

Tacit knowledge has lived in people's heads because the one instrument that could draw it out, a real conversation that asks a follow-up question and adapts to the answer, never scaled. You could interview a handful of people and learn a great deal, or survey thousands and learn almost nothing about their reasoning. Depth and reach pulled against each other, so most companies chose reach and gave up the reasoning.

A system of record is faithful about the past. The judgment that produced the past sits in the other column, in the one form the system cannot store.

Your existing stack is a system of record. It is faithful about the past and silent about the judgment that produced it. The knowledge that would change your next decision is held right now by the people closest to the work, in the one form your systems cannot store.

The opportunity is not another dashboard. It is to treat that knowledge as the asset the research shows it to be. That means going and asking for it, in a way that keeps the distinct voices it came from intact rather than averaging them into a summary. And it means keeping that knowledge current as the business shifts underneath it. That is the layer your systems were never built to be. The knowledge is already in the building. The question is whether anything is listening for it.