IPC International Community magazine an association member publication
Issue link: https://iconnect007.uberflip.com/i/1546580
SEPTEMBER 2026 I I-CONNECT007 MAGAZINE 111 check first when a lot comes back bad?" The operator is now having a conversation. It is not a writing assignment. The tool takes the answers and organizes them into a draft that could become a work instruction or a training page. He corrects what is wrong, and for the first time, what is in his head is on paper. Are You New to the Factory? If you are the 21-year-old who just got hired, I am talking to you. You are probably the least experi- enced person in the building, and not the one who will make the engineering call. However, you are also the best-positioned person in the building to start this work. I mean that literally. The senior engineers already know everything, and for years, they've been saying that the defect data is a mess and somebody should clean it up and write down what the plater knows. They are right, but they are also buried. They have a customer escape and a quote due Friday. The press is down again, and the cleanup never gets to the top of the list. You have two things they do not: time that is not yet fully claimed, and a relative comfortability with such tools, having "grown up" with them. You already know how to talk to an AI. But you don't yet have the process knowledge of the shop. That's fine, because the plater has it and he will talk to you if you ask with respect. You are the person who can gather the evidence in one place, utilizing the person with the required expertise. How to Start Now Do not start with a big project. Take one problem that has been an issue for a long time, like the de- fect that comes and goes, that nobody has ever nailed down. Ask a supervisor which one it is, and they will (should) know immediately. Then go collect every place that the problem has been written about: scrap codes, spreadsheets, shift notes, photos of the notebook pages, and the emails from the last time it got escalated. Do not clean anything yet, but gather all of it. Now, hand it all to an AI tool and ask it to do two things. First, find all the different words people used for this problem and group them. Second, have it tell you what is missing. Where are the gaps in the dates? Which lots have a scrap code but no note? The gaps are often where the answer is hiding. Again, do not ask the AI tool about the root cause, though you will be tempted to. Ask it only to orga- nize the evidence and show where each piece came from. Take that output to the person who has run that process the longest (the expert). Show them the pattern. What do they notice that never made it into a log? Interview them, using the tool to help capture what they say. Then bring the whole package to an engineer and have them make the call. When they reach a conclusion, make sure it points back to the records. If it does not, it is a guess with a nice font. That is the whole method: Gather it, put it in one set of names, find the gaps, interview the expert, let the engineer decide, and keep the provenance. Repeat on the next "stuck" problem. One More Time Inspection is an important part of the overall quality process. Keep your AOI and catch what you can at the end of the line. But do not confuse better inspection with better quality. A machine that finds a bad board faster is useful. A person, with a tool in their hands, who figures out why you keep making that bad board in the first place is worth far more. Deming said it 70 years ago, and the industry is still catching up: "Quality comes from under- standing and improving the system, not from catching mistakes at the end." AI finally gives a normal person on a normal shop floor a way to better understand the system. The only question is whether your shop creates the right tools (with AI) and puts the tool in the right hands. If you are new, those hands just might be yours. I-CONNECT007 Sean Patterson is the founder of CrossGen AI. He previously served as COO and CTO at Summit Interconnect, held senior multisite roles at TTM Technologies, and was CRO of Nano Dimension, a former Amazon executive, and U.S. Navy submarine veteran.