A senior machinist retired in the spring. Four months later, his plant had absorbed an estimated 2.3 million dollars chasing problems he used to solve before anyone knew they existed. That figure comes from aerospace trade reporting, and the number is almost beside the point. What matters is the shape of it, because the shape repeats in every plant I walk.
The money did not leave when he did. It leaked out afterward, in scrap on a setup he used to feel by sound, in a tolerance he held by habit, in the second-guess he made on certain jobs that nobody else knew to make. None of that was in a document. It was in his hands. And tribal knowledge capture services exist for exactly one reason: to get what is in those hands onto something durable before the hands are gone.
This is not a one-plant story
It is tempting to file that as a story about one unusually critical machinist. It is not. It is the structural condition of the industry right now.
According to the Manufacturing Institute, one in four US manufacturing workers is over 55, and the sector needs as many as 3.8 million workers by 2033, of which roughly 1.9 million jobs are projected to go unfilled. Industry estimates put the experience riding out the door over the next decade at more than 70 million combined years. The retirements are not a surprise. The plants can see them coming on a calendar. What surprises them every time is how much undocumented process leaves with each one.
The cost is real, and it is invisible by design
The reason this keeps happening is that the cost never appears on the P&L with an honest label. There is no line item that reads “we let one person’s experience substitute for a process.” Instead it shows up scattered: as a scrap spike on a particular setup, as overtime for the people covering the gap, as a quality finding attributed to a training opportunity, as a new hire who takes far longer to get productive than anyone budgeted for.
That last one is its own quiet tax. PwC reports most manufacturers onboard frontline employees in just five to ten days, while the Bureau of Labor Statistics projects roughly 963,400 production-occupation openings per year through 2034, largely from people leaving the work. A five-day onboarding against a thirty-year operator’s tacit process is not a training program. It is a hope.
Why the knowledge was lost long before retirement
Here is the part most plants get backward. They treat the knowledge as lost on the day the person retires. It was lost long before that, on the day capture stopped happening.
The senior operator’s expertise was never a risk while he was standing there to apply it. The risk was created the moment the plant decided his presence was a substitute for a written, validated, transferable process, and stopped doing the work of capturing what he knew. Retirement does not cause the loss. It just reveals a loss that has been accumulating, invisibly, for years. The exposure was there the whole time. The clock simply ran out.
Why the binder and the search bar never solved it
Plants have tried to solve this, and the market has sold them tools to try. Better binders. Document-management systems. A search bar over the binder. More recently, an AI that summarizes the binder. Every one of these improves access to the document. None of them touches the problem, because the problem was never that the document was hard to find. The problem is that the document never contained the tacit knowledge in the first place.
You cannot search a binder for the thing nobody ever wrote down. You cannot summarize what was only ever in someone’s hands. Aging workforce knowledge transfer is not a retrieval problem. It is a capture problem, and capture is harder, slower, and far more valuable than anything a search bar does.
Run this in your plant this week: the retirement-exposure back-of-envelope
You can size your exposure in an afternoon, without hiring anyone.
Start with a list of every operator and technician with 15 or more years of tenure. For each name, write the single process only that person can run clean, the one where everyone else gets a worse result or asks them for help. Then put a rough probability next to each name: how likely is this person to leave, retire, or rotate in the next 24 months. High, medium, low is precise enough.
Now you have a one-page map. The high-probability names sitting next to a critical, single-owner process are your live risk. Multiply it out however you like, but the honest version is simpler than any formula: every high-probability line on that page is a process that is one departure away from becoming undocumented, and the cost of rediscovering it the hard way is the 2.3-million-dollar shape from the top of this piece, scaled to your operation.
If that page makes you uncomfortable, it is doing its job.
What actually works
What works is capturing the process while the person is still there to validate it, and keeping it current after they are gone. That is the only problem we built SenseiLab to solve. The 30-day SOP Sprint puts a Lean practitioner and a structured AI capture process onto the floor, working next to your most experienced operators, capturing the five to ten procedures that carry the most risk, validating each one with the people who actually run it, and turning the result into Living SOPs that stay accurate as the work evolves. The back-of-envelope you just ran is the five-minute version. The Sprint is the 30-day version, scaled across the processes that matter most.
The retirement clock is real. The 70 million years of experience are real. The undocumented process on that one-page map is real. When the high-probability names walk out the door, they take it with them, unless you capture it before then.
About the author
Diego Echenique is the CEO and co-founder of SenseiLab, where the team captures the knowledge that actually runs manufacturing operations before it walks out the door. He has spent more than two decades in manufacturing operations across the Americas, Europe, the Middle East, and Asia.
