The numbers don’t lie, and they don’t comfort. Only 22% of workers worldwide strongly agreed that their job was safe from elimination, according to a new report from ADP Research released Wednesday. The finding comes from one of the largest workforce sentiment surveys ever conducted—more than 39,000 workers across 36 countries—and lands with the force of a gut punch: The world’s workers are consumed by fear.​

“Despite three years of historically low global unemployment and steady economic growth, our data reveals widespread job insecurity expressed by workers worldwide,” said Nela Richardson, chief economist at ADP.​

The culprit is hiding in plain sight: artificial intelligence. As generative AI tools race into the workplace at breakneck speed, workers from Tokyo to Topeka are struggling to process what it means for their livelihoods—and they’re not reassured by what they see.​ “AI is not like the weather. It is not just going to descend upon us,” Richardson told reporters in a briefing on the survey results in New York City. “It really is hitting us at the task level—by augmenting and making certain tasks more high value.”

A world of anxious workers

The ADP Research Today at Work 2026 report, based on survey responses collected in late summer 2025, paints a portrait of a global workforce caught in the crosscurrents of technological disruption, demographic upheaval, and deep uncertainty. The anxiety cuts across borders and industries, but ADP found that it hits hardest at the bottom of the organizational ladder.​

Among individual contributors—the frontline workers who make up the bulk of most companies’ headcount—only 18% felt their job was safe. Frontline managers fared only slightly better at 21%. Confidence rose predictably with seniority: Middle managers came in at 23%, upper managers at 31%, and C-suite executives at 35%. In other words, the higher your perch in the org chart, the less afraid you are of falling.​ But even then, only barely more than a third of top executives feel like they have job security, according to the results.

The geographic divides are equally stark. In Japan—a country long defined by lifetime employment culture—only 5% of workers felt their jobs were secure, the lowest reading of any market in the survey. Nigeria, by contrast, registered the most confident workforce, with 38% of workers expressing job security, driven largely by a young, tech-savvy population and booming AI adoption. In the United States, the figure was 28%.​

AI is making it worse—and better, sort of

The paradox at the heart of the report is this: AI use is correlated with higher engagement and less stress, yet it’s also making workers feel dramatically less productive. Daily AI users were four times as likely as nonusers to say they weren’t as productive as they could be.​

The explanation, researchers suggest, is psychological. AI has taken over the small, checklist-style tasks that gave workers a sense of daily accomplishment—answering emails, summarizing documents, generating first drafts. Without those quick wins, people feel as if they’ve done less, even when they’ve arguably done more. “AI does the things that we used to say make us feel productive,” Richardson explained to reporters. “It writes our emails for us. It responds. It summarizes documents for us. And so we have to kind of remeasure productivity in a different way; it’s shifting from productivity based on volume of work to value [of work].”

The flip side: 30% of daily AI users were fully engaged at work, versus just 14% of those who never use AI. Heavy AI users were also significantly less likely to experience negative stress: 11% reported feeling overloaded, compared with 23% of nonusers. The data suggests AI may be a powerful tool for worker well-being, if companies can figure out how to deploy it without triggering dread.​

The creeping, unpaid extension of the workday isn’t helping with this angst over productivity: 62% of global workers said they put in up to five hours of unpaid work per week, while 38% reported working off the clock for six hours or more; 12%—disproportionately executives and upper managers—said they work without pay for 16 hours or more weekly.​

The data reveals a troubling paradox: Workers logging the most unpaid hours are simultaneously the most engaged and the most likely to be looking for another job. They’re dedicated enough to give their time for free, yet burned out enough to be quietly interviewing elsewhere. “Free work comes at a cost,” the report concludes. “People who put in unpaid hours are more likely to feel unproductive and stressed. They’re also more likely to quit.”

“AI is entering a workforce that is anxious,” said Jay Caldwell, ADP’s chief talent officer. “And to me, that’s very, very risky. And the importance for HR professionals right now is not as much about the technology. It’s more around, ‘How do we lead through the technology? And how do we bring our workforce along?’ And the transformation that comes with that.”

Five generations, one nervous breakdown

Compounding the AI anxiety is a demographic collision unlike anything the modern workplace has ever seen. For the first time in history, five generations are working side by side—from teenagers to great-grandparents. And they are not on the same page.​

Young workers ages 18 to 26 were the most optimistic, with 29% saying they had the skills needed to advance. But older workers ages 55 to 64 told a bleaker story: Only 18% felt similarly equipped, and just 12% believed their employer was investing in their development. Meanwhile, 20% of young workers strongly agreed AI would positively impact their jobs in the next year—a figure that dropped to just 10% among workers ages 55 to 64.​

“Younger workers are definitely more optimistic about their skill set,” Richardson told reporters. “Older workers are also, you know, more likely to say that they’re financially unprepared. Which is interesting. They make more money, but they feel more stretched financially. They’re more likely to say that they’re less productive and less engaged than younger workers. Youth and optimism go hand in hand.” She connected these findings back to Japan, which has a culture famously respectful towards elders. Fortune has previously covered the madogiwazoku, or “window workers,” who are so named because they sometimes just stare out the window. The upshot is that jobs for younger workers are harder to come by.

The data also exposes a troubling engagement crisis hiding beneath the surface calm of low unemployment. Only 19% of workers globally were fully engaged on the job in 2025—unchanged from the year before—meaning more than 80% of the world’s workforce is, by some measure, just going through the motions. Workers who find meaning in their jobs are 12.5 times as likely to be fully engaged as those who don’t.​

What employers must do

ADP researchers are emphatic that the anxiety gripping workers is not inevitable; it is, in large part, a leadership failure. Workers who feel their employers are investing in their skills were 5.3 times as likely to feel their jobs were secure. Among those who strongly agreed their employer was investing in them, 53% were fully engaged. Among those who didn’t feel that investment, only 12% were.​

The prescription is clear: Communicate, upskill, and stop treating workers as passive recipients of technological change. “Upskilling isn’t just a strategy,” Richardson said at the briefing. “It’s a reassurance. It’s a trust pact between the employer and the worker.”

Caldwell urged HR professionals to help workers reframe what productivity even means in an AI-driven world—away from task volume and toward judgment, creativity, and long-term impact. “Workers who clearly see the role their existing skill sets will play in an organization’s future,” Caldwell said, “will be more engaged, productive, and have the confidence to thrive in this next era of work.”

For now, though, the world’s workers are mostly not thriving. They’re waiting, watching, and wondering whether their jobs will survive the machines. The survey suggests that the answer—for most—remains maddeningly unclear.