Tribal Knowledge

In almost every organization there is knowledge that exists only in people’s heads. No one has written it down, but everyone experienced acts on it. Which customer never calls on Fridays. Which machine needs a minute to warm up first. Which button you must never press, because it throws the billing into chaos. This exact kind of unwritten experiential knowledge of a group is called tribal knowledge. It is not passed on through manuals, but through observation, asking questions, and years of familiarity.

Why companies fall apart when someone quits

Tribal knowledge is enormously valuable and, at the same time, extremely dangerous. Valuable, because it makes a team fast. A seasoned employee solves in five minutes a problem that a newcomer would spend half a day on. It is dangerous because this knowledge is stored nowhere. If the person retires or quits, it vanishes from the company along with them.

In business terms, this is called a concentration risk. A company can invest millions in software and still grind to a halt because the one colleague who alone knows how the old billing run works is out sick. In the industry, there is even a dedicated term for this: the bus factor. It describes how many people would have to be lost before a project collapses. A bus factor of one is an alarm signal.

For the tech industry, this point is especially topical. Anyone who wants to hand off work to software or AI must first be able to describe it. What no one can write down, no computer can take over either. That is why tribal knowledge regularly turns up today in analyses of why automation projects end up more expensive than planned.

How unwritten knowledge emerges and gets passed on

Tribal knowledge does not arise on purpose. It grows because reality is more complicated than any manual. A set of instructions describes the normal case. But everyday work consists of exceptions, workarounds, and stopgap solutions that have proven themselves. Nobody writes these workarounds down, because to those involved they are simply obvious.

Experts distinguish here between explicit and implicit knowledge. Explicit knowledge can be put into sentences, such as an instruction manual. Implicit knowledge shows itself only in action, such as the gut feeling for when a customer is about to walk away. Tribal knowledge is above all implicit, and that is exactly why it is so hard to capture.

Passing it on traditionally works through proximity: shared shifts, being trained alongside an experienced colleague, questions asked in the hallway. Companies try to make this more systematic. They maintain wikis, write checklists, document workflows, or have teams explain in interviews how they actually work. A common misconception is that a wiki would solve the problem. Documentation goes stale quickly, while the knowledge in people’s heads keeps adapting continuously.

From the workshop floor to the AI training dataset

In everyday life, the term shows up anywhere seasoned groups work together. In a restaurant kitchen, in a school administration office, in a craft business. Even a school class has tribal knowledge: which teacher tolerates lateness and which does not is written down in no school rulebook.

In business news, the term often comes up in connection with acquisitions and layoffs. When a corporation buys a company and many experienced people leave afterward, the technology remains, but the knowledge of how it is actually run is missing. Analysts then say the buyer has lost tribal knowledge. Critics make a similar argument regarding major layoff waves in the tech industry.

In the world of AI, the term has taken on an additional meaning. Language models learn from written text. Knowledge that was never written down is simply missing for them. That is why companies today pay experts to explicitly articulate their experiential knowledge, so it can flow into training data. Anyone who wants to deploy AI within an organization quickly realizes: the real work consists of finally writing down the unwritten.