White-collar workers are quietly rebelling against AI as 80% outright refuse adoption mandates

There was a moment, not long ago, when “shadow AI” felt like a good-news story. Workers were sneaking ChatGPT and Claude past the IT department, using personal accounts to do what used to take hours in minutes. An MIT study published last year found that employees at more than 90% of companies were using personal chatbot accounts for daily tasks — often without approval — even as only 40% of those same companies had official LLM subscriptions. The shadow economy was booming. Management called it a governance problem. The workers called it getting the job done.

Now the data tells a different story. The tool that workers once raced to adopt covertly has become, for a large and growing share of the workforce, the tool they’ve stopped using altogether. Not because it doesn’t work. Because they’re afraid of what happens when it works too well.

A new global survey of 3,750 executives and employees across 14 countries, conducted by SAP subsidiary WalkMe for its fifth annual State of Digital Adoption report, finds that more 54% of workers bypassed their company’s AI tools in the past 30 days and completed the work manually instead. Another 33% haven’t used AI at all. Combined, roughly eight in 10 enterprise workers are either avoiding or actively rejecting the technology their employers are spending record sums to deploy. Average digital transformation budgets rose 38% year-over-year to $54.2 million — yet 40% of that spend has been underperforming due to adoption failures.

Executives are blind to how employees really feel

What the early enthusiasm obscured is now visible in the numbers. Only 9% of workers trust AI for complex, business-critical decisions, compared to 61% of executives — a 52-point trust chasm. Eighty-eight percent of executives say their employees have adequate tools; only 21% of workers agree — a 67-point gap on tool adequacy alone. Executives and their employees are, in the report’s language, “describing fundamentally different companies.”

The skeptics have data on their side, too. Steve Hanke, the Johns Hopkins economist, has been through enough technology cycles to know what hype looks like from the inside. “AI didn’t deliver,” he told Fortune recently. “Welcome to the real world. Forget the AI bubble. You know, it didn’t deliver. You look at all the surveys and yeah, everybody’s using it a little bit, but you dig into it and it hasn’t done much.” Hanke’s bottom line: “Productivity, by the way, it was weak. If AI delivered, productivity would be way up. You listen to these Silicon Valley guys and they say we’re gonna have GDP going to 5% of 6%. Productivity is gonna go up to six. It’s just not happening.”

That skepticism is, in its own way, consistent with what the WalkMe data is finding. Dan Adika, CEO and co-founder of WalkMe, has been tracking this divergence from the front lines. He meets regularly with CIOs and asks them a simple question: how many of your people are actually using AI to do meaningful work? “The numbers are sub-10%,” he said.

Adika used the metaphor, favored by this particular editor as well, that AI is like a sports car in terms of its speed. He said his favorite analogy is if you buy every employee a sports car, but they don’t know how to drive it—they don’t have the AI skills.

Part of the problem is structural, not behavioral. “You buy every employee that sports car, the Ferrari, but they don’t know how to drive,” Adika said. “They don’t have fuel sometimes, which is the context. Knowing how to drive is the prompting. And in some cases, there are not even enough roads — there’s no API or MCP server to actually do what you want to do.” What do you do when you have a Ferrari, but no driver, no fuel, and no roads? You don’t go very fast.

Brad Brown, Global Head of Tax Technology & Innovation for KPMG in the U.S., used almost the same exact metaphor in a separate interview with Fortune. “It’s like an F1 car driver,” he said. “The F1 car is amazing. But if you don’t have a skilled and talented driver, that tool’s not gonna do much for you.” The fact that two veteran technologists — one a founder, one a Big Four partner — converged on the same description unprompted suggests they are describing something they’ve both seen firsthand, repeatedly, at scale.

The chasm is costing companies

The downstream cost of that undriven Ferrari is now quantifiable. The WorkMe report found that workers lose the equivalent of 51 working days per year to technology friction — nearly two full months — up 42% from 2025. That’s 7.9 hours per week. Goldman Sachs economists reported this week that AI saves workers who use it correctly an average of 40 to 60 minutes per day. The math is almost symmetrical: the productivity AI gives to people who use it well is almost exactly equal to the productivity it destroys for people who can’t get it to work.

The old shadow AI story is still alive beneath the surface. Seventy-eight percent of executives say they want to discipline shadow AI use — yet only 21% of workers report ever being warned about AI policy, and 34% don’t even know which tools their employer has approved. Executives are threatening punishment for behavior that they’ve never explained is prohibited. The contradiction runs so deep that 62% of those same executives privately concede that the risk of unsanctioned shadow AI is overstated compared to the risk of not leveraging AI at all.

“The use of shadow AI isn’t a behavior to penalize — rather, it’s an opportunity to address a systemic gap,” said Keith Kirkpatrick, Vice President and Research Director of Enterprise Software Digital Workflows at The Futurum Group. “When employees use unapproved AI tools, they’re compensating for performance or efficiency gaps left by sanctioned tools and unclear governance.”

AI disengagement

What’s new — and what the data is only beginning to capture — is the layer beneath shadow AI. Workers who aren’t sneaking around the rules. Workers who aren’t doing anything.

Adika was asked what he’d call this dynamic. He paused. “They have pride in what they do,” he said, about workers who are resisting AI adoption. “They won’t let some AI bot take over, and they will always find and show the flaws in that tool compared to them.” It sounds, unmistakably, like quiet quitting — the pandemic-era phenomenon in which workers stopped going above and beyond without formally resigning. It could also be a very understandable frustration with AI tools that just won’t stop hallucinating, wasting as much time as they promise to save.

“The organizations that get this right won’t be the ones that just automated the most tasks,” Adika said. “They’ll be the ones that figured out when the human should act, when the agent should act, and how the handoff between them works. That handoff is where trust lives. And right now, most companies haven’t even started thinking about it.” To this point, the MIT study found that 90% of workers still prefer humans for mission-critical work, a clear reluctance to dive into the deep end.

Oracle has announced layoffs of tens of thousands of workers, following a similar announcement from Block, although critics see this as “AI washing,” or disguising over-hiring with a convenient excuse that happens to boost the stock price. The logic is not lost on the rank and file. “We will be in a certain point of time when we will feel uncertainty, fear, we’ll see layoffs,” Adika said. “So I think it’s kind of a transition period that will happen over time. But again, at the end of the day, people are not using it yet.”

Adika was also clear that workers staying away from AI are not wrong to sense something real — they’re wrong about the conclusion. “You wouldn’t see any CEO of a bank or insurance company go tomorrow and lay off a lot of people, because who will do the work?” He said he sees a “big issue” coming to a head because claims that AI will replace everyone will have to confront the fact that “it’s just not happening right now.”

The skilled driver problem

Brown said he’s spending more time than ever thinking about what it actually takes to close the gap between the Ferrari and the driver. At KPMG, he has begun categorizing the workforce into what he calls builders, makers, and power users — distinct tiers of AI capability with explicit career paths attached. “Our focus right now is to craft incentives and career paths to get all our people to that level,” he said. “It’s time for the humans to catch up to where the tech is.”

The critical insight in that framing is that the problem isn’t intelligence, nor is it even training in the traditional sense. “I think with your sort of human skills that you bring to the table in terms of critical thinking and judgment,” Brown said, “that’s going to lend people into being makers” — workers who can leverage AI tools fluidly, including using them to build new tools themselves. The workers most at risk, in his view, are not the ones who lack technical skill. They’re the ones whose employers haven’t given them a safe space, a path, or an incentive to try.

A third of the enterprise workforce has never used AI tools at all — and they report the lowest levels of support, the least training, and the highest anxiety about disruption. They are not, the WalkMe report notes carefully, resisting AI. They have simply not been reached. As to whether the evolution of these tools is outpacing workers’ ability to catch up, Brown acknowledged that he definitely feels a gap.

Evolving is possible—and important

What brought Hanke back around was all the time saved, once he figured out what he wanted to use AI for. “AI to me is kind of like another research assistant,” he said, “and it saves a hell of a lot of time because if I had a research assistant doing this stuff, I’d have to send them to the library. They’d be screwing around over there for a week doing something I can do on AI in about an hour.” The caveat: “You have to know what they’re good for.” And, crucially, you have to know enough about the subject matter to catch the errors. “I know what to ask AI. I know how to structure what I want done,” Hanke said, pointing to his decades of domain expertise across economics, commodities, and international finance.

His own trajectory — from outright banning student use to cautious skepticism to daily reliance — tracks the arc many serious thinkers have traveled. He said he went from “‘no’, to ‘maybe’, to ‘this is great—but some of these tools suck.'” His verdict on the tools themselves is characteristically blunt: “There are all kinds of AI. And some of it’s really crap. It depends on what you need.”

Brown’s view is that this is ultimately an optimistic story — but only for those who move. “The winners are the ones where you have your workforce effectively leveraging the capabilities of AI,” he said. “A workforce that’s not leaning into AI is going to be challenged. And a work environment that is overly oriented to AI without the value of the human workforce is going to struggle.”

This story was originally featured on Fortune.com