According to Kence Anderson, founder and CEO of AMESA, the answer begins with understanding exactly why those operators become experts in the first place.

“It's because of variation,” Anderson says. “It's because of variation in the input material. It's because of variation in the equipment that changes over time. Ambient temperatures change over time. Humidity changes over time. It takes a lot of practice to get it right across all those different variations.” 

Rather than relying solely on production data, Anderson says manufacturers must also capture the decision-making process experienced operators use as conditions change throughout production.

His company's "Data to Autonomy" methodology combines plant-floor data with operator expertise to train AI agents in simulation before they're introduced into manufacturing environments.

“Every manufacturing process has operating regions,” Anderson says. “Startup is different from steady-state operation, which is different from shutdown. Those are different skills that have to be learned.” 

Looking for quick answers on assembly and manufacturing topics? Try Ask ASM, our new smart AI search tool. Ask ASM

In adhesives and sealants manufacturing, that expertise often extends beyond information collected by production equipment.

Quality measurements may come from laboratory testing performed after production, requiring manufacturers to combine real-time process data with quality results to fully understand how production decisions affect finished products.

“The quality data needs to be pulled together because you need a complete picture,” Anderson says. “Did the things I did lead to the right thickness? Did they lead to the right uniformity? But then, did you pass the quality tests?” 

Simulation also allows manufacturers to model complex physical and chemical interactions that would otherwise be difficult to reproduce through trial and error alone.

“You can combine machine learning techniques with the physics,” Anderson says. “You get a quick-to-develop simulation, but you also get some of that first-principles accuracy.” 

Despite advances in AI, Anderson believes experienced operators remain at the center of successful manufacturing.

“Seventy percent of the time, companies never intended to do anything beyond decision support,” he says. “They're trying to empower their people to do better at their job.” 

For manufacturers wondering where to begin, Anderson recommends focusing first on the production processes where expertise is most at risk of disappearing.

“The use case where you should start,” he says, “is where, if you don't do this in the next year, you're going to lose the expertise completely.”