Sathish:
“How does this generative AI wave compare to previous technology shifts in manufacturing?”
Rick:
“Going back to early manufacturing, we had machines, people running them, people inspecting things, a lot of people handling stuff. In my time, we did computer-integrated manufacturing. We put in robots. At the Autolite spark plug plant, that probably replaced 500 people picking up spark plugs, looking at them, and setting them on the next tray. Those robotics made a huge impact on the physical handling. But we actually increased pretty significantly the amount of knowledge work, reports, and analyzing that had to be done. Generative AI is exactly attacking that. It is taking the overhead we put in during my time. It was the first thing to really help take the workload off the engineer’s head so that he has to think less about feeding systems and more about solving problems.”
Sathish:
“Where can manufacturers see the best use case and value with generative AI?”
Rick:
“Product development is anywhere from two to five to 15 percent of the total cost. Manufacturing is where the cost is. In manufacturing, a little mistake, a little lack of action at the right time, can generate huge amounts of cost. A lot of automotive companies, their profitability at a deep level is based on how they do with plant startups and model changeovers. When I was at GM, 80 to 90 percent of them went smooth, but 10 percent really had big issues that would literally define the profitability of the company. With generative tools, they can literally look at real-time data, see what is wrong, diagnose it, and give very detailed advice for somebody to go fix something. That is where it shows at the bottom line.”
Sathish:
“What is the biggest challenge when manufacturers try to adopt AI?”
Rick:
“You need to start on something small and make sure the value is really there. People are still skeptical of AI in manufacturing. You do not want to do an AI project that is really cool but either does not work or does not give a return, because that will slow everything down. The other big mistake is trying to do it yourself. You can get 80 to 90 percent right, but if you need to be 100 percent right, that cool thing you built is really hard to turn into something that runs everywhere. It would almost be like the guys on the shop floor writing their own PLM system. Who is going to maintain that? Experiment internally, but understand that rolling it out at scale typically takes expertise and productized platforms that most companies do not have in house.”
Sathish:
“Manufacturers generally have so many systems that are not connected. Can AI work if the data is fragmented?”
Rick:
“It is not all off the machines. It is on spreadsheets, emails, information coming in from the supply chain. We found during a startup in Valencia, Spain, there were about 300 spreadsheets in use. Over a three-month period, we had six sigma people collect them all and analyze them. It turned out that only 50 percent of the information on those spreadsheets was current and accurate. Yet those sheets were driving daily decisions. Gen AI can absolutely nail that problem. It can pull together structured and unstructured data, highlight inconsistencies, and help leaders see the current state faster than any manual process.”
Sathish:
“What about agentic AI, where AI takes action on its own? Is that safe for manufacturing?”
Rick:
“There is a study that should scare every manufacturing person. Somebody gave AI a set of options, good ones and bad ones, and told it clearly which were which. Without pressure, about 14 percent of the time it picked the bad answer. Under pressure, that went up to 80 percent. That is dangerous on the shop floor. As a controls engineer, I would certainly put another level of control around it. A framework where it simply cannot do the bad thing. AI should advise. It should help people make faster, better decisions. But letting it take autonomous action on critical systems without hard guardrails is a risk most manufacturers are not ready to manage yet.”
Sathish:
“Where do you see AI in manufacturing in three to five years?”
Rick:
“There was a study that said AI is used in about 50 percent of product development today. In manufacturing it is about 2 percent. Manufacturing is the laggard for understandable reasons: heavy equipment, safety concerns, complex training, high cost of mistakes. But if a manufacturer becomes fully capable with these tools, sometimes 10 percent factory efficiency, that is huge money. That will be very difficult to compete with. You need to be really fast but really careful. You need to accept that this is going to happen, get the right partners, give everyone in your company access to AI in a private secured manner, and envision how you want your plant to operate in the future, fully AI-enabled, and then work backwards to get there.”
