Industry is abuzz with all that artificial intelligence (AI) stands to offer—what it can do to create smarter, more capable, more efficient manufacturing and logistics operations. Sure, AI has been in use for quite some time in manufacturing, especially for predictive maintenance. But the value that it promises has changed, particularly with the advent of generative AI and intelligent agents.
At this year's AI & Smart Automation Conference, put on for the third time by the Association for Advancing Automation (A3), the attitude was one of pragmatism.
Two words often used to describe AI are "hype" and "magic," noted Robert Huschka, vice president of education strategies for A3, as he kicked off the conference. AI is not magic, he conceded. "But it's not completely hype either."
Despite referring to the "superpowers of AI" in the title of his introductory presentation at the conference, Kence Anderson, CEO and co-founder of Composabl, did much to bring the topic back down to earth. In fact, he countered the idea of AI's magic with this perspective: "When you combine AI innovations with domain expertise, you do get things that look like magic," he said. "But that's not magic. That's engineering."
By combining machine learning and AI with traditional automation, we get layers of intelligence and decision-making capabilities that were not previously possible, Anderson added.
Composabl creates tools that help engineers teach AI agents to work alongside them in real-world settings. The superpowers of intelligent agents are perception, learning, strategy, planning, and deduction, Anderson explained, and they all need to work together to make intelligent agents effective.
In one case study that Anderson detailed, PepsiCo created an intelligent autonomous agent to control an extruder that makes Cheetos. It takes an operator about 12 years to learn all the ins and outs of controlling some 25 parameters on that extruder. But the agent took nine months to build and one month to train before it was certified as an expert operator.
An important aspect of AI, though, is that it is not lined up to make existing engineering systems obsolete. The nail didn't become obsolete just because the screw came along, Anderson pointed out. Likewise, new algorithms do not replace the old algorithms. AI innovations must be combined with domain expertise and applied by engineers.
Real Opportunities for AI
In a panel discussion during the conference, speakers from UPS, 3M, and Tyson Foods detailed the opportunities they see for using AI to improve operations on several different levels.
UPS is building out the AI of the future, said Gregory Brown, vice president of strategy and R&D. "We want it to be as autonomous as possible," he said. "If we have the right data, those autonomous decisions can be applied to anything."
3M is trying to understand how to use AI not only to do what it's currently doing but better, but also how to do things now that it hasn't been able to do in the past. In the machine vision space, that might mean finding subtle, ambiguous defects that 3M couldn't find in the past, said Jeffrey Adolf, senior specialist of manufacturing technology. "AI helps you solve problems that were unapproachable before."
Tyson Foods is trying to bring some insight and organization into the chaos of day-to-day production. Unplanned downtime is a huge part of that for a company that has 103 manufacturing facilities around the world, noted Sean Lawrence, associate director of automation. Tyson is not looking to use AI for autonomous control, but rather to help workers make better decisions. As Lawrence noted, not everyone is an expert, and AI can help to upskill team members.
With regard to generative AI specifically, there have been questions about whether it could actually work. "We found the answer to be yes," UPS's Brown said. "It's not a solution. It's a tool that goes along with all the other ways you can solve the problem." Gen AI has been useful in many ways for UPS, he added.
In the manufacturing space, though, Adolf thinks gen AI is not yet where it ought to be. "We're just scratching the surface, as probably many people are," he said. There are a lot of opportunities, though, which is why he was at this conference. "We're trying to understand where people are winning in this space."
Where gen AI can really help on the manufacturing floor, Lawrence said, is providing operators the insights they need from the dashboards they just don't have time to dig into. Operators know what questions to ask, and gen AI gives them a way to ask those questions and find the answers they're looking for.
Overcoming Challenges in Food Variability
From a machine vision perspective, there is considerable opportunity to put AI to work in the food sector. With a huge amount of variability in food manufacturing, especially in proteins and produce, AI can go a long way to making vision systems more effective, according to Harley Green, vice president of strategic accounts at Oxipital AI.
Key benefits of AI-based vision include reduced dependence on labor, increased yield, increased quality, and reduced waste and rework.
Green presented a use case involving a company that processes chicken breasts, where saving even just a gram of product can have a massive impact by the end of the year. Switching from high trim to low trim, for example, human operators are very likely to over trim or under trim. That can equate to millions of dollars for each of these producers.
In another case, Green showed how a company making eggrolls could save close to $7 million a year just by putting the AI-based vision system on the line before the relatively expensive filling is dispensed—scrapping about $187,200 worth of eggroll wrappers rather than more than $7 million worth of filling.
Just Get Going
"We're at the cusp of something tremendous," said Dianne Eldridge, worldwide leader of business development and GTM, industrial AI/Gen AI, at Amazon Web Services, who spoke of transforming business and manufacturing with AI.
Eldridge provided a compelling use case with a major paper company that was seeing issues with a particular machine throughout its facilities. It's a $250,000 machine, but every time it breaks down, it costs them $50,000. By applying AI to the problem, the company was able to cut downtime by 20%, saving it $1 million per machine, with multiple machines across multiple plants. Improving overall equipment effectiveness (OEE) by 10% also meant a savings of $13 million.
But beware: Data is the differentiator, Eldridge warned, and you're not going to be able to get the most out of AI if you don't have the right data. So, your data strategy will have to be your first step.
Nonetheless, just get going, Eldridge urged. "Just iterate and get going."
Contact Us Today
Whatever your marketing goals may be, TECH B2B Marketing can help with them all. To learn more about how we can help with your next integrated marketing campaign, contact us today.
