A few years ago, you couldn’t sit through a manufacturing conference without hearing “Industry 4.0” at least a dozen times. Smart factories. Cyber-physical systems. The Fourth Industrial Revolution. It was a great story, and like most great stories in manufacturing, it was mostly told by people who don’t run a shop.

Walk the floor at most of the plants I work with today and nobody says “Industry 4.0.” Nobody says “digital transformation” either, at least not without a little eye-roll. What they say now is a lot more specific: “Can this thing quote a job faster than Dave?” “Will it tell me when the CNC is about to go down?” “Is this going to replace my third shift lead?”

That shift in language isn’t cosmetic. It tells you the industry moved from a destination framing to a tool framing — and that changes everything about how adoption actually happens, who resists it, and why.

The buzzword died. The technology didn’t.

Here’s the paradox: the terminology cooled off right as the actual capability got serious. Industry 4.0 was largely a sensors-and-connectivity story — IoT, cyber-physical systems, data exchange. What’s landed on the floor in the last two years is a different animal: AI that can look at a quote history and price a job, flag an anomaly on a line before it becomes scrap, or draft a maintenance schedule without someone building a
spreadsheet for it.

The numbers back this up. RSM’s 2026 Middle Market survey of manufacturers found 88% now have AI at least partially integrated into their operations, with 32% reporting it’s fully integrated across core processes. Automation World’s 2026 State of Manufacturing data shows AI adoption jumping from 53% to 72% of manufacturers in just two years. This isn’t a pilot anymore. It’s operational.

But — and this is the part that matters if you’re actually running a plant — adoption and scaled adoption are two different animals. Redwood Software’s 2026 outlook found that 98% of manufacturers are exploring AIdriven automation, but only 20% feel fully prepared to run it at scale. Automation World’s number is even blunter: 72% have deployed AI in some form, but only 10% are scaling it effectively.

That gap — between “we bought the thing” and “the thing actually changed how we work” — is where the real story lives. And it’s not a technology problem. It’s a people problem, all the way up and down the org chart.

Leadership isn’t as unified as it looks from the outside.

Here’s a stat that should make every executive team a little uncomfortable: RSM found that 84% of manufacturers agree leadership is more enthusiastic about AI than their employees are. On paper, that reads as “leadership is bought in, we just need to get the floor on board.” In practice, I don’t think that’s the whole story.

I’ve sat in enough leadership meetings across five subsidiaries to tell you that plenty of that “enthusiasm” is performative — it’s a box being checked because the board asked about AI strategy last quarter. A recent enterprise survey from Writer found 75% of executives admit their company’s AI strategy is “more for show” than real internal guidance, and 48% called their own AI adoption effort a “massive disappointment.” That’s not a workforce resistance problem. That’s a leadership clarity problem wearing a workforce resistance costume.

And it gets worse further down. There’s a concept some change-management writers have started calling the”frozen middle” — the idea that upper leadership pushes AI initiatives from fear of falling behind, frontline staff are often more curious about AI than anyone gives them credit for, and it’s middle management holding the whole thing still. Research on this pattern points out that traditional, operations-heavy industries like manufacturing see more of this resistance than tech companies do, partly because middle managers in our world have longer tenure and more identity wrapped up in “the way we’ve always run this cell.” When the thing being automated is coordination and judgment — the stuff a plant manager or a supervisor has built a career on — the resistance isn’t laziness. It’s self-preservation, and it’s rational.

That matches what PwC and the Manufacturing Institute found in their 2026 look at frontline leadership: when leaders were asked what concerns their teams actually raise, the top answers were insufficient training (40%) and lack of clarity on purpose and ROI (38%). Fear of job displacement came in third, at 25%. People aren’t primarily afraid of the robot. They’re afraid nobody’s told them what it’s for.

The floor is more ready than leadership thinks.

This is the part I find most interesting, because it inverts the story most executives tell themselves. Automation World’s 2026 data found that more than half of frontline manufacturing employees — 53% — describe themselves as generally receptive to AI. Only 22% call themselves resistant. That runs directly counter to how leaders characterize their own workforce.

If it’s not resistance, what is it? The research points to mistrust, not defiance — and a lot of that mistrust is earned. Fifty-three percent of manufacturing employees in that same survey believe AI could replace them, which is a completely reasonable thing to believe when your employer hasn’t said otherwise. The people running the equipment aren’t rejecting the technology. They’re rejecting being kept in the dark about what it means for them.

There’s a comparison in that Automation World piece I keep coming back to: no plant manager would hand a brand-new hire the keys to a machine on day one and walk away. You’d shadow them, build up their responsibility slowly, coach them through mistakes. Most AI rollouts skip that whole arc. Straight from pilot to production, no mentoring period, then leadership is surprised when adoption stalls. That’s not an AI failure. That’s a training system failure we’ve made a hundred times before with new equipment, and we’re making it again with new software.

The trust math is measurable, and it’s not close.

If you want a single number that explains the gap between “explored” and “scaled,” Prosci’s change management research gives you one. On their scale for measuring how smoothly transformations go, organizations with strong AI rollouts scored leadership support at +1.65. Struggling organizations scored -1.50.That’s not a nuance — that’s more than a three-point swing on a four-point scale, and it tracks almost
perfectly with whether the rollout succeeds or gets quietly shelved.

The same research found frontline workers trust AI at +0.33 while executives trust it at +1.09 — a gap of more than three quarters of a point on that scale. WalkMe’s 2026 study of enterprise workers found an even sharper version of that gap on high-stakes decisions: only 9% of workers trust AI for complex, business critical calls, versus 61% of executives. That confidence gap isn’t the workforce being backward. It’s the workforce having a more accurate read on the risk, because they’re the ones standing next to the machine when it’s wrong.

So what actually works

None of this is a reason to slow down. It’s a reason to be disciplined about how you speed up, which is a distinction our industry has never been great at. A few things I’d put weight behind, based on where the data and my own experience line up: Say what it’s for, in plain language, before you deploy it. The single biggest predictor of a smooth rollout isn’t the sophistication of the tool — it’s whether leadership can articulate the purpose clearly enough that a supervisor can repeat it to their team without a slide deck. PwC’s data on frontline leaders backs this up directly: lack of clarity on purpose and ROI outranks fear of job loss as the actual objection.

• Treat AI rollout like you’d treat training on new equipment. Shadowing, staged responsibility, a real ramp period — not “here’s the tool, go.” This is Theory of Constraints thinking applied to change management: you don’t push more volume through a system before you’ve proven the new step can actually hold the load.
• Get honest about the leadership gap before blaming the floor. If 84% of your org believes leadership is more enthusiastic than employees, and three-quarters of executives privately admit theirAI strategy is more show than substance, the fix doesn’t start on the shop floor. It starts with leadership actually using the tools themselves and being willing to say out loud when a rollout isn’t working.
• Watch the middle, not just the front line. The floor is more receptive than most leadership teams assume. The friction is more likely sitting with the supervisors and plant managers whose coordination role is exactly what’s being automated. Naming that plainly — and giving those managers a real role in the new system instead of asking them to hand it over — matters more than another all-hands presentation.

We spent years arguing about what to call this era of manufacturing. Turns out the name mattered a lot less than whether the people running the equipment believed anyone had thought about them before hitting “deploy.”

---

Sources Referenced:

• Automation World, “Scaling AI in Industrial Automation: 2026 Data on Workforce Buy-In” (2026 State of Manufacturing Survey)
• RSM US, “Here’s what AI for manufacturers looks like in 2026” (RSM Middle Market AI Survey 2026)
• Redwood Software, “Manufacturing AI and Automation Outlook 2026”
• Writer / Workplace Intelligence, “Enterprise AI adoption in 2026” survey
• PwC and The Manufacturing Institute, “Frontline leadership in manufacturing’s AI adoption” (2026)
• Prosci change-management benchmarking data, cited in “Change Management for AI Adoption: A 2026 Playbook”
• WalkMe, 2026 study of enterprise workers
• SecureWorld, “Frozen in the Middle: The AI Bottleneck” (2026)
• AI Assembly Lines, “How to Overcome Middle Management Resistance to AI” (2026)

---