RM 198: Sean Patterson on Making Al Useful in Manufacturing

Mike Konrad (0:00)

Looking for the latest in SMT assembly process knowledge?

Come to PCB West on September 29th and 30th for a dedicated program on electronics assembly, led by leading practitioners in solder paste, placement, reflow, stencil design, and cleaning, highlighted by Chris Shea's workshop on solder paste qualification.

All at the Santa Clara California Convention Center. Visit pcbwest.com for details.

Artificial intelligence is everywhere right now. It's in the headlines. It's being built into software tools, and it's showing up in more and more conversations about engineering, manufacturing, quality, and process improvement. But there's a big problem. The company can buy the software. A team can attend the training. Everyone can walk away impressed by what AI might be able to do.

And then, a few weeks later, very little actually changes. The tool is there. The potential is still there, but the habits never form. And that matters directly to reliability. Welcome, or hopefully welcome back, to Reliability Matters. In electronics manufacturing, reliability doesn't come from good intentions. It comes from repeatable processes, disciplined execution, clear documentation, good decision making, and the ability to recognize problems before they become failures in the field. That's where AI can become more than just another interesting tool. Used properly, AI may be able to enhance workflows such as reviewing ECOs, engineering change orders, and identifying downstream processes, documentation, material, or inspection impacts, analyzing AOI, SPI, test, and process data to detect recurring defect patterns, supporting root cause analysis by organizing failure data, inspection results, and corrective action history, monitoring process drift in areas such as reflow, cleaning, stencil printing, placement, and environmental conditions, assisting with corrective action reports, customer responses, and internal documentation, capturing tribal knowledge from meetings, shift reports, troubleshooting notes, and past problem-solving activity. Those are all reliability-related activities. They touch the way products are designed, built, inspected, documented, corrected, and improved. But AI only helps if it becomes part of the way work actually gets done.

My guest today is Sean Patterson, author of the Printed Circuit Design and Fabrication article, Why Your AI Training Isn't Sticking. Sean argues that AI adoption is not really a training problem. It's a habit problem. In other words, the goal isn't simply to teach people what AI can do. The goal is to connect AI to the work people are already doing, such as preparing for meetings, reviewing engineering change orders, working through technical questions, or organizing complex problems. For those of us focused on reliability, that distinction is important. AI doesn't replace engineering judgment, process knowledge, or verification. But when used properly, it may help strengthen the behaviors that support reliability, asking better questions, capturing knowledge, reducing overlooked details, and improving the consistency of technical decision making. So today, we're going to talk to Sean about why AI training often fails to stick, how small triggers can turn occasional use into practical habits, where AI can fit into engineering and manufacturing workflows, and how companies can approach AI in a way that is useful, responsible, and directly connected to reliability. And it all begins right now. Stick around.

SPEAKER_2 (3:45)

Welcome to Reliability Matters, a podcast for the Electronic Assembly Industry. Each episode covers topics related to reliability, best practices, and environmentally responsible assembly techniques, with insights from experts across the electronic assembly industry. Now, here's your host, Mike Konrad.

Mike Konrad (4:03)

Sean Patterson, welcome to Reliability Matters. I'm excited about our conversation today.

Sean Patterson (4:09)

Thanks, Mike, same to you.

Mike Konrad (4:12)

Okay. So, let's, as I mentioned in the introduction, let's talk about your article a bit. Your article makes the point that AI training often fails because people learn the tool, but they don't build the habit.

What are you seeing that led you to come to that conclusion?

Sean Patterson (4:33)

I would say that the past 30 or 40 years of new technologies coming out of technology companies into non-technology companies, right? It's not just AI, it's the history of patterns that we've had.

Whether it's HubSpot, Salesforce, ERP, X, Y, and Z, all these kind of systems that have been largely put out into the business world. I like to say leaders sort of buy the gym membership and expect everybody to get fit, right?

Mike Konrad (5:06)

Wait, you actually have to go?

Sean Patterson (5:08)

Yeah, you actually have to go.

Mike Konrad (5:10)

Well, that explains a lot, Sean.

Sean Patterson (5:13)

Well, you get the analogy. It's always happened and I left corporate America to go do this as a full-time job. Because I'm an operations C-suite guy for a long period of time. And just we tend to not introduce new technology in an appropriate way, largely in any business, quite frankly, but really these non-technology businesses buying these other things. And then you wind up with leadership, one doesn't understand what they're buying, largely, because it's a whole sales process and you truly get, you want the results of what you see, but you actually understand what you're buying.