Every shop has knowledge that lives in only one person’s head, and then you find out just how much when that person takes a vacation or leaves the company. The line starts acting up, the yield slips, and everyone realizes that they should have captured that person's knowledge.
The cost is not limited to the parts that must be reworked or scrapped. Production slows down as people try to replicate the decisions the experienced operator made without much thought. Supervisors start calling around for answers, engineering gets pulled into the line, and the new operator becomes afraid to touch anything without asking permission. One missing person can quickly turn into five people losing time.
It’s not like you didn’t already know this. So, to prevent this situation, you try handing the veteran plater a work-instruction template to fill out. A few weeks later, the template comes back half-completed, or it contains the same basic steps already posted beside the line: check the temperature, verify the chemistry, start the cycle, record the readings. All are correct, but none of them capture the experiential knowledge base that makes that plater so valuable. Yet, this has been the standard half-hearted and wholly ineffective attempt to do just that.
So, why doesn’t the template approach work? The problem is that most templates are built to document a process after the important decisions have already been made. They are good at holding limits, steps, and sign-offs. They are not good at uncovering why an operator checks one thing before another or why two readings that appear acceptable can still point to a problem.
The missing information is in the small decisions of what he looks at first, which reading he doesn’t quite trust, the sound that makes him turn around, and the condition that tells him to wait five minutes before making an adjustment. After 20 years, those decisions don’t feel like decisions anymore; they’re simply part of running the line. If you ask how he knows a bath is starting to drift, he may say, “You can just tell.” That information has to be drawn out before it can be organized. AI can help.
Start With a Conversation
Instead of trying to pull that knowledge out of your plater with a form, use AI to start a conversation. You just need one experienced operator, one interviewer, and a small block of time. The interview can happen in one sitting or over several shorter conversations on different days. With the operator’s knowledge, record each conversation so nobody has to take notes. Explain that you’re trying to capture how the process is actually run, so the next operator can learn faster and make fewer mistakes. Most experienced people are happy to teach what they know. They just don’t want to spend an afternoon staring at empty boxes on a computer screen.
To continue reading this article, which originally appeared in the October 2026 I-Connect007 Magazine, click here.