Josh Peeno says digital transformation fails when manufacturers treat the factory like a math problem instead of a cultural one. From machine teaching and fractional tech leadership to workforce training and global competition, AI only works when it is built for the plant floor.
Most digital transformation failures in manufacturing are blamed on technology. The software was immature or the data wasn’t ready or the model wasn’t accurate enough. But Josh Peeno, P.E. and president of the JPeeno Innovation Group, argues that explanation misses the real issue.
The problem, he says, is usually human.
Peeno, a fractional technical director, industrial AI specialist and professor at the University of Toledo, has spent more than 25 years in manufacturing, much of it in the heat and complexity of glass production. That experience has made him skeptical of innovation strategies that look elegant in a conference room but collapse in front of operators, maintenance teams and aging equipment.
On a recent episode of the Ctrl+Alt+Mfg podcast, Peeno laid out a practical case for why digital transformation initiatives stall, why AI still unnerves the plant floor and what manufacturers can do differently. His answer is not to slow down on AI, but to deploy it with more realism. Start with operators, preserve expert knowledge, use outside technical leadership wisely and train the next generation for the messy reality of industrial environments.
Digital transformation is a cultural shift, not a math equation
Too often, manufacturers approach digital transformation like a technical exercise when it is really an organizational one.
“What I see a lot is companies want to deploy AI or some of the digital intelligence without even going onto the plant floor and talking to the people that are the experts in the industry,” Peeno said. “That’s where it tends to fall apart because it’s a people problem, too.”
That disconnect shows up when corporate teams pursue projects that make sense on paper but do little to solve actual operating pain points. The result is familiar across industry: long roadmaps, expensive platforms and limited buy-in from the people expected to use them.
For Peeno, the right starting point is not a software demo. Instead, it’s operator conversation.
“The first question that people should be asking when they’re thinking about digital transformation or deployment of AI is what are the problems? What are the actual problems on the plant floor? What are the actual problems that the operators are having?” he said. “Go talk to them, ask them specifically what we can do to actually make your job easier and solve those problems first. Then you’re going to get buy-in every time from the operations team and the experts.”
That is a deceptively simple prescription, but it reframes digital transformation as a change-management effort grounded in plant-floor credibility. In Peeno’s view, manufacturers get further when they solve one meaningful operational problem than when they launch a sweeping initiative no one on the line trusts.
Why machine teaching works better than fear-based AI
If digital transformation has a human problem, AI has a trust problem. Peeno said conventional AI often creates resistance because operators see it as a black box — or worse, as a threat. In many cases, they are told the technology will optimize decisions they have spent decades learning to make. That is not a recipe for adoption.
“It shouldn’t be viewed that way, it should be viewed as an enhancement,” Peeno said. “It allows them to do their job more effectively rather than replace them. It’s an Iron Man suit, if you will, for the operators.”
That is where his concept of “machine teaching” comes in. Rather than treating AI as something that learns abstractly from large amounts of data, Peeno favors the more grounded approach of teaching autonomous systems how experienced humans think about the process.
“What that really means is teaching machines to think the way that humans do about a process,” he said.
He compares the model to an apprenticeship. Instead of replacing the veteran operator, the AI learns alongside that person, absorbing constraints, preferences and process logic that may never appear cleanly in a spreadsheet.
“It’s almost like bringing in an apprentice,” Peeno said. “Instead of the human apprentice, you’re bringing in the AI apprentice, and you’re letting it work alongside an expert, a guy that’s been doing it for 20, 30 years.”
That approach helped shape a successful deployment in glass manufacturing, where Peeno and his team applied autonomous AI to the gob-forming process. With dozens of interacting variables and too many “knobs” for a human to optimize simultaneously, the system was trained in simulation using machine data and expert guidance. The key was not handing over control overnight, but using AI first as decision support, with the operator still in the loop.
That strategy lowers resistance because it respects expertise instead of dismissing it.
The rise of the “fractional” technology bodyguard
Peeno’s work also points to the growing reality in manufacturing that many mid-market companies need technical leadership, but not necessarily a full-time chief technology officer.
That is where the fractional model comes in. Peeno described his role plainly: “Fractional really just means part-time.” But the value is larger than the word suggests. A fractional technical leader can evaluate current capabilities, pressure-test vendor claims, help build a roadmap and prevent companies from making oversized software bets before they understand what they actually need.
In practical terms, that means helping manufacturers take smaller, more disciplined steps. It also means helping software providers understand the factories they want to sell into.
Peeno said he often works on both sides of that divide, advising manufacturers on which tools are truly worth pursuing, while also helping software startups adapt their offerings to the “harsh environments that they’re going into that they probably don’t understand very well.”
For mid-sized firms, that kind of guidance can be the difference between measured progress and a multimillion-dollar mistake. It can also prevent digital transformation from becoming overwhelming. Instead of trying to modernize everything at once, plants can sequence investments, validate business value early and avoid feature-heavy systems that create more technical debt than operational benefit.
Why heavy industry breaks pristine AI models
Peeno’s credibility on industrial AI comes in part from his experience in glass, where high temperatures, legacy systems and incomplete data quickly expose the limits of textbook models.
“Factories aren’t clean environments, the stream of data isn’t as clean,” he said. “There’s missing data. There’s lots of information that is just not there.”
That reality matters, especially in lagging industries where modernization has been uneven. Engineers and developers outside manufacturing often assume plants are collecting more process variables than they really are, or that the signals they do collect are stable and usable enough to support sophisticated analytics. On the plant floor, the truth is usually messier.
The harshness of the environment is only part of the problem. The bigger issue is that process knowledge often lives with people rather than systems.
“I think the most important asset in any factory is not any of the machines, but the most important asset in any factory is those experts that have been there for 20-plus years,” Peeno said.
That makes AI deployment in heavy industry as much a knowledge-capture challenge as a technical one. The question is not just whether a model can survive the environment. It is whether the organization can extract enough real-world process understanding to make the model meaningful in the first place.
Teaching students what Industry 4.0 really looks like
That same realism shapes Peeno’s work in the classroom. At the University of Toledo, he teaches mechatronics and industrial automation, but his focus goes beyond PLC programming.
His goal is to teach students how to think like plant engineers. That means learning how to troubleshoot, work with legacy controls, navigate communication issues and understand the practical constraints of factory systems. In other words, he is preparing them for the reality that modern manufacturing is not a clean-sheet design problem.
That matters as the industry confronts a widening workforce gap.
“There is for sure a skills gap, and there’s a big wave of skills gap coming, as well, that needs to be addressed,” Peeno said.
He sees AI as part of the answer, not because it eliminates the need for engineers, but because it may help preserve knowledge as older experts retire.
“The biggest problem — the one that’s in front of us — is we are going to lose skills, and we need to find ways using technology to capture that data from the experts, capture that skills and that knowledge, and then use that to digitalize our factories,” he said.
North America can no longer treat automation as optional
Peeno’s final warning was about competitiveness. In his view, manufacturers outside the United States have moved faster on automation and digitalization because they recognized earlier that they had no choice.
“I think North America is falling behind in the fact that we have always treated automation and digitalization as a luxury rather than a necessity,” he said.
That mindset is getting harder to sustain. Labor shortages, retirements and global competition are turning technology investment from a strategic nice-to-have into an operational requirement. The plants that move successfully will be the ones that ground technology in human expertise, practical deployment and clear business problems.
Peeno’s message is not anti-AI. It is anti-abstraction. If manufacturers want digital transformation to stick, they must stop designing for ideal conditions and start building for the plant they actually have.
The Ctrl+Alt+Mfg Podcast
Make sure to check out other episodes of the Ctrl+Alt+Mfg podcast, where hosts Gary Cohen and Stephanie Neil discuss a range of digital transformation insights. The last five episodes are listed below:
Ep. 10: Modernization vs. digital transformation, with Dan Furrow and Luis Atencio of Wesco
Ep. 11: What plant engineers really want, with Amara Rozgus of Plant Engineering
Ep. 12: Why system integrators matter more than ever in the age of AI, with Adrienne Meyer of CSIA
Ep. 13: Bad data, broken maintenance, with Paul Ross of Limble and Ross Fergerson of RBC Bearings
Ep. 14: From data silos to smart factories, with John Dyck of CESMII and John Harrington of HighByte
