# The Tribal Knowledge Thesis ## In process-manufacturing plants, nobody wrote the most important operation knowledge A senior engineer looks at a drifting yield curve. The engineer knows which valve, which supplier lot and which handover between operator groups to examine first. None of this knowledge is in the ERP (enterprise resource planning system) or the manufacturing execution system (MES). When this person retires or stops work, the plant keeps the machines but loses the diagnostic memory. The thesis behind Foreman is that AI agents can collect this knowledge, put it in a structure and make it possible to find it. As a result, a junior operator can do work that is nearer to the work of a veteran. > [!abstract] The claim in one line > Knowledge is a *stock* that decreases when persons stop work. AI agents can change it into a *flow* of decisions. The flow stays after the persons who created the knowledge stop work. ## How strong are the indications? | Pillar | Strength | Information | | --- | --- | --- | | Unplanned downtime has a very high cost | **Strong** | Research by an institute that has no relation to the company shows this, see [[What the Downtime Number Actually Means]] | | Missing skills and missing knowledge increase the recovery time | **Strong** | The same research shows this relation clearly | | The supply of new technicians is small | **Unverified** | Decks say that a small number of persons start vocational training in some regions. Nobody checked a primary source for this claim | | Plants will pay for the knowledge that the company collects | **Unproven** | The pain exists. The open question is if plants will pay at the necessary ticket sizes | | The collected knowledge changes decisions | **Unproven** | Data from before and after a deployment in production are necessary, see [[Claims Versus Evidence]] | > [!warning] The difference between the pain and the software > All persons agree that the pain exists. At this time, nobody external to the company can see if the collected knowledge is correct. Nobody can see if the customer uses it. Nobody can see if it is better than a retrieval tool that costs less. See [[The Evaluation Gap]]. ## Why it matters A pain that exists and that many persons feel is necessary. But it is not sufficient. The investment decision needs the pain to change into software that the company can use again and again. This is the question in [[Platform or Solutions - The Central Question]]. ## Related - [[Foreman MOC]] - [[What the Downtime Number Actually Means]] - [[Agent Skills as Codified Domain Expertise]] - [[Knowledge Graphs for Industrial Data]] - [[Industrial AI MOC]]