A plant schedules a knowledge-transfer project ahead of a senior operator's retirement. There is a spreadsheet template with a column for "critical knowledge," a weekly interview slot on his calendar, and a deadline six months out. The meeting gets moved twice because the line needs him more than the interview does.
When it finally happens, most of what comes out of him is "you'd just know." He is not being difficult. The judgment he has is not the kind that sits still long enough to be asked for — it lives in the shape of a fault he has seen forty times, not in a sentence he has ever needed to say out loud. The interviewer writes down what she can. He retires. The folder holds four pages, and none of them is the thing that actually took years to learn.
Why we build it this way
The plan failed for a reason that has nothing to do with how well it was run. Writing down what you know is not a task a busy person will ever choose over the task in front of them, no matter how important someone tells them it is. It has no deadline that matters more than the machine that is actually down right now, and it asks for something a form is bad at holding in the first place: judgment earned over years, requested cold, in the fifteen minutes somebody could spare.
The obvious fix is a better knowledge-management project — a wiki, a template, a mandate from above to document more. That does not change the shape of the problem. It is still a second job stacked on the first one, competing for the same hour, and it will lose every time, because the first job is the one with a machine waiting on it.
The knowledge was never the problem. Asking someone to stop and write it down was.
So we do not run a capture project. We build the writing-down into tools people already have a reason to use today, for their own job, with no form in sight. The record accumulates because the work happened, not because anyone was asked to document it.
How it works, roughly
Nothing here asks a person to sit down and recall anything. It reads what people were already doing:
- A technician confirming or correcting a diagnosis a machine proposed
- An operator correcting a step in a procedure that was slightly wrong for this line
- An engineer rejecting a design option, with the reason for the rejection
- A deviation someone measured and flagged, that nobody had a place to log before
Each of those is an action the person was going to take anyway to get through their day. What changes is that it gets written down, in their own words, at the moment it happens, instead of relying on someone remembering to mention it before they leave. When a person disagrees with what the record suggests, that disagreement is kept too, reason attached — a "no" is exactly as useful as a "yes," and it is what makes the next answer better.
A year on, the plant is not holding four pages from an exit interview that came too late. It is holding a record that grew while everyone kept doing their actual jobs — a diagnosis here, a corrected step there, a rejected option with its reason attached, all in words the plant already uses. Nobody was asked to write any of it down, and a person still decides everything it feeds into. The record itself sits on hardware the plant owns, so when someone retires, it does not retire with them.
Read the vision →