Generative AI can help fill skills gap, attract younger workers

“When a new employee who maybe doesn’t have those years of experience is running that injection molder, or whatever piece of equipment, and runs into a problem, they can query using normal human language and say, ‘I found this problem, how should I think about it?’ ” said Tim Gaus, smart manufacturing business leader at Deloitte.  

The generative AI solution can then combine the interviews of experts with information from maintenance records, manuals and additional sources to provide a “richer and quicker way to address and resolve issues,” Gaus said. 

The availability of generative AI also might make manufacturing careers, like plastics processing, more appealing to Gen Z workers who might otherwise shun those jobs, according to Deloitte. While manufacturing has evolved dramatically since the industrial revolution, younger workers often still see the sector as grimy and outdated. 

“You have a lot of folks entering the workforce today that just expect to have these digital tools at their availability,” Gaus said. “They grew up in a very digital age with a phone in their hand from, some would argue, birth, but I won’t go that far.” 

Gen AI tools can dramatically speed the transition of new workers from entering the door and getting orientation to becoming an effective employee on the plant floor. In the past, employers worried about how to retain workers. However, today, many younger folks entering the workforce don’t really want to stay. 

“I’ve had a few folks tell me, ‘I’m not even worried about retention anymore. I’m worried about planning for turnover and doing that as effectively as possible,” Gaus said. 

Help with predictive maintenance, automation 

In addition to using generative AI for worker training, manufacturers increasingly are using it to assist with predictive maintenance on manufacturing equipment, like injection molding machines. Artificial intelligence has been available for a while to make predictive maintenance recommendations to users. However, generative AI can take those suggestions to a new level and instill users with confidence in the recommendations. 

“The challenge we’ve had with adopting a predictive maintenance recommendation from an AI engine is that it feels like a black box,” Gaus said. “It says to do it, but you don’t know why it tells you to do it, so, you are kind of asked to trust the machine. What you’re seeing now with generative AI is you can overlay that kind of natural language ability to communicate those predictive models that allow them to be much more explainable.” 

As an example, Gaus said if an AI model recommends a user replace a motor on a piece of equipment, the worker can now use natural language to ask the AI software why the motor should be replaced and what caused the AI software to predict a possible near-term failure. The generative AI software can then explain, for example, that a particular sensor was vibrating well out of tolerance.  

In addition, AI and generative AI are allowing robotic processes to learn and adapt to new situations without human intervention, enhancing flexibility and efficiency. An example Gaus cited was vision analytics used in conjunction with robotics in manufacturing processes. 

“Vision analytics traditionally have been a little victim to things like changes in lighting conditions and things of that nature, and that could actually send your process out of control,” Gaus said.  

However, advances in artificial intelligence allow users to program robots to respond to situations they might not have previously encountered. 

“You’re able to synthetically create data to replicate conditions that you haven’t been able to actually show it during the actual training process, and then, moreover, with that kind of installed base and understanding of how to react to things that it’s never seen, as it starts to detect that new condition occurring, it can automatically adapt to it without a human being involved in that loop.” 

As an example, Gaus said if a robot is involved in picking parts from an assembly line and placing them in a bin, if the bin is moved, the robot with its vision sensor can correct for the bin’s new location on its own. 

Simulation and digital twins 

Over the next year, the use of simulation in the manufacturing industry is expected to increase, especially given the potential for business disruptions, the need to control costs and the continued proliferation of AI tools.  

Examples of simulation in manufacturing cited in the Deloitte report include:  

• Causal AI, which is used to more effectively simulate cause-and-effect relationships, thereby enhancing decision-making capabilities.  

• Production line simulation, which can help eliminate bottlenecks and optimize workflows before any physical changes are made.  

• Process simulation, which can result in higher throughput and reduced costs. 

• Business scenario simulation, which simulates challenges, such as employee absences, raw materials that arrive with quality issues and supply chain disruptions, and potential actions to optimize the response. 

“Simulation, to me, has been an underutilized tool in the plastics processing toolkit,” Gaus said.  

Plastics processors using digital twins (virtual models of a physical process or piece of equipment) can use simulation to make better decisions, he said. A plastics processor, for example, can run simulations of how equipment would perform while processing resins with different properties. Running simulations ahead of time allows a processor to be prepared to respond to changes in resin quality rather than reacting while production is underway. 

“To me, simulation allows a significant derisking of your operations,” he said. “It allows a richer way to make decisions as you are in the moment.” 

Simulations can play a critical role in predictive maintenance accuracy by testing different operating conditions and predicting potential equipment failures before they occur. Digital twins can also simulate how different mold designs might perform prior to building the mold. 

Contact:  

Deloitte, New York, 212-492-4000, www.deloitte.com