Companies are spending more on artificial intelligence than ever, but most of their employees are quietly refusing to use it. A global survey of 3,750 enterprise workers, published in April 2026 by digital adoption platform WalkMe, found that 54% had bypassed their employer’s AI tools at least once in the previous 30 days, choosing to complete the task manually instead. Another 33% said they had not used AI at all. Those two categories likely overlap: some workers who bypassed AI once may also fall into the broader non-usage group. But even under conservative assumptions about that overlap, the combined figures suggest that roughly four in five white-collar workers are either sidestepping or ignoring the AI tools their companies want them to adopt. That estimate treats the 54% bypass group as the larger circle and adds only the portion of the 33% non-users who are not already counted among the bypassers. If, for example, half of the non-users also bypassed a tool at least once, the remaining unique non-users would add roughly 16 to 17 percentage points to the 54%, producing a total in the high-60s to low-70s range. If the overlap is smaller, the figure climbs closer to 80%. The precise number depends on data WalkMe has not published, which is why the four-in-five framing should be read as an approximate upper bound rather than a hard count.
The scale of the pushback
The WalkMe study surveyed workers across 14 countries and focused on employees at enterprise-scale organizations. Its central finding is not just that people skip AI tools occasionally but that the pattern is widespread enough to carry a measurable cost. The company estimates that “technology friction,” its term for time lost to struggling with, working around, or ignoring employer-provided digital tools, drains the equivalent of 51 workdays per employee each year. That figure covers all enterprise software, not AI alone, but it captures the broader environment in which AI resistance is playing out: workplaces already saturated with tools that many employees find more frustrating than helpful.
Independent research supports the general trend. A Pew Research Center survey published in February 2025 found that a majority of U.S. workers reported using AI minimally or not at all in their jobs. Pew’s study measured general usage rather than resistance to specific mandates, and it predates the most recent wave of enterprise AI rollouts. Still, it establishes a baseline that aligns with WalkMe’s newer numbers: across different methodologies, sample sizes, and geographies, most office workers are not regularly engaging with AI on the job.
WalkMe has a commercial interest in these findings. The company sells software designed to help organizations smooth digital adoption, so framing technology friction as a large, solvable problem serves its business model. That does not invalidate the survey data, but it means the company’s interpretation of its own results deserves scrutiny. The directional agreement with Pew’s nonpartisan, noncommercial research is what gives the core finding its weight.
What workers and observers are saying
The survey data released so far does not include direct testimony from individual workers explaining their reasons for avoiding AI tools. Neither WalkMe nor Pew published named worker quotes alongside their findings. Similarly, no named HR leaders or independent workplace-technology analysts have been quoted in the primary sources underlying this story. That is a significant gap. Until researchers or journalists publish on-the-record interviews with employees who are bypassing AI and with the managers trying to drive adoption, the human dimension of this trend remains underreported. Readers should treat the narrative of “quiet resistance” as a pattern visible in aggregate survey data, not as a claim verified through first-person accounts.
The absence of an independent academic or analyst voice is also worth noting. The original version of this article cited Steven Hanke of Johns Hopkins University as an expert on technology adoption, but Hanke’s published work focuses on currency policy and inflation, not workplace AI. No more directly relevant academic source has been identified in the available evidence. Until a researcher whose expertise sits squarely in organizational technology adoption, human-computer interaction, or digital workplace strategy comments on these specific findings, the article lacks the independent expert perspective that would strengthen its analysis.
Why workers may be pushing back
The landscape of plausible motivations is broad and well-documented in adjacent research, even though no single cause has been isolated in the WalkMe or Pew data.
Training gaps are one likely factor. Many companies have rolled out AI tools, from Microsoft Copilot to internal chatbots, without investing equivalent time in teaching employees how to use them effectively. When a tool feels clunky or unreliable, the rational response is to fall back on a familiar workflow that gets the job done.
Trust is another barrier. Generative AI systems can produce confident-sounding output that turns out to be wrong, a problem widely discussed under the label “hallucinations.” For workers in fields where accuracy matters, such as finance, legal, or healthcare administration, the risk of an AI error can outweigh the time savings.
Job security fears may also play a role, though in a counterintuitive direction. Some workers worry that demonstrating AI can do their tasks efficiently will make their own positions look expendable. In that framing, resistance is not technophobia but self-preservation.
None of these explanations has been isolated as the dominant cause in the available research. Without sector-level breakdowns or qualitative interviews tied to the WalkMe or Pew data, the specific mix of motivations remains an open question.
What the data does not show
Several gaps in the evidence are worth flagging. The term “mandate” implies a directive with consequences for noncompliance, but neither the WalkMe study nor the Pew survey distinguishes between companies that strongly encourage AI use and those that tie it to performance reviews or employment conditions. A suggestion to experiment with a generative tool is a different thing from a policy that docks ratings for non-adoption, and the research does not separate the two.
Sector-level variation is another blind spot. The WalkMe survey spans multiple industries but has not published granular breakdowns by profession. Whether resistance is concentrated in specific fields, say, legal and compliance versus marketing and sales, or spread evenly across the white-collar workforce is unknown.
Geography likely matters, too. Regulatory environments, data protection rules such as the EU’s AI Act, and cultural attitudes toward automation differ sharply across regions. But without country-level analysis tied to the adoption numbers, it is impossible to say whether workers in some markets are more willing to embrace employer-provided AI than others.
Perhaps most importantly, the data captures a snapshot, not a trajectory. The 54% bypass rate reflects what workers reported in one 30-day window. AI tools are improving rapidly, and corporate training programs are still maturing. Today’s resistance rates may look very different a year from now.
The bigger pattern
Workplace technology adoption has never moved as fast as procurement. Companies bought PCs for years before most employees used them for anything beyond email. Enterprise software suites like SAP and Salesforce went through long, painful adoption curves marked by workarounds, shadow IT, and quiet noncompliance. AI appears to be following the same arc, just at higher stakes and under a brighter spotlight.
The gap between executive enthusiasm and frontline reality is not new, but the scale of investment makes it more consequential. If the workers those dollars are meant to empower are still doing tasks by hand, the return on that investment will remain elusive.
For employers, the WalkMe and Pew findings suggest that buying AI tools is the easy part. The harder work, redesigning workflows, investing in hands-on training, building trust in AI outputs, and giving workers a reason to change habits, is where adoption will be won or lost. Until that slower, messier process catches up to the spending, many employees will keep doing what workers have always done when handed a new system that feels risky or confusing: work around it, quietly, and get the job done the way they already know how.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
