Nearly everyone is using AI at work now. Almost nobody is using it well.
That’s not a hot take. It’s what three major studies published in the last six months consistently show. AI has spread faster than almost any workplace technology in history. And yet the gap between having access to it and actually getting value from it may be the defining professional divide of 2026.
Here’s what the data says and what the people on the right side of that gap are doing differently.
The Numbers Look Better Than They Are
According to Gallup’s Q3 2025 workforce survey of more than 23,000 U.S. employees, 45% now use AI at work at least a few times a year—up from 40% just the quarter before. Frequent use, defined as several times a week or more, rose from 19% to 23% over the same period.
Those are real gains. But look closer, and the picture gets more complicated.
Daily use sits at just 10% of the American workforce. And a striking 23% of employees say they don’t even know whether their organization has implemented AI technology at all. The people furthest from that knowledge are individual contributors, 26% of whom reported being uncertain, compared to just 7% of leaders. AI is spreading across organizations, but it’s spreading unevenly, and the clarity about its purpose is thinner than adoption numbers suggest.
The Gallup data also reveals what most workers are actually using AI for: consolidating information (42%), generating ideas (41%) and learning new things (36%). These are useful applications. They are not transformative ones.
The 5% Problem—and Why It Matters for You
Here’s where the EY 2025 Work Reimagined Survey, which drew on 15,000 employees and 1,500 employers across 29 countries, gets uncomfortable.
88% of employees report using AI at work. But the vast majority are limiting themselves to surface-level tasks: search, summarization, drafting. Only 5% are using AI in ways that fundamentally change how they work and what they produce. EY’s research found that organizations are leaving up to 40% of their potential AI productivity gains on the table because of this gap.
That’s not a technology problem. It’s a human one.
The research identified a clear culprit: Just 12% of employees say they receive sufficient AI training to unlock anything beyond the basics. When workers don’t know how to push past the first layer of what AI can do, they don’t. They use it for what feels safe and familiar—a smarter search engine, a faster way to draft a first paragraph—and move on.
So what does this mean for you? If your use of AI feels like a more convenient version of what you were already doing, you’re likely in the 88%. The 5% aren’t using fundamentally different tools. They’re using the same tools with a fundamentally different intent.
The Fear Factor Nobody’s Talking About Enough
The EY data also surfaces something leaders rarely address directly: Fear is actively suppressing how deeply people engage with AI.
38% of employees surveyed said they fear losing their job to AI. The same proportion worry that relying too heavily on AI will erode their own skills, expertise and judgment over time. These two anxieties aren’t contradictory. They’re actually the same anxiety in different directions: Fear that AI will replace you, and fear that you’ll let it.
The result is defensive behavior. Workers use AI just enough to stay current but pull back from the deeper experimentation that would actually change their output. They’re hedging. And it’s costing them the upside they’re anxious to protect.
EY’s research found that in organizations where leaders have successfully driven genuine AI adoption, 75% of employees report that their leadership team is clearly aligned on an AI vision and, crucially, communicates not just what AI will do, but what humans will do that becomes more valuable because of it. That message is the difference between a workforce that experiments and one that hedges.
What ‘AI Power Users’ Actually Look Like
Here’s the finding that most consistently surprises people: The workers who use AI the most are not the ones retreating from human connection.
Gensler’s 2026 Global Workplace Survey, which gathered responses from more than 16,400 office workers across 16 countries, identified 30% of employees as “AI Power Users,” defined as people who integrate AI regularly in both their professional and personal lives. When researchers compared this group to workers with lower AI adoption, the differences were striking.
AI Power Users spend less time working alone: 37% of their work week, compared to 42% for late adopters. They spend more time learning: 12% of their work week versus 8%. They spend more time socializing: 11% versus 9%. The narrative that AI drives people into digital isolation doesn’t hold up. The workers most embedded in AI workflows have more time for the distinctly human parts of their work, not less.
Janet Pogue McLaurin, global director of workplace research at Gensler, put it directly: “The employees most embedded in AI workflows are also the ones most engaged in learning and have better team relationships. That shift signals a new and important role for the workplace.”
The AI isn’t replacing their collaboration. It’s handling enough of the solitary, repetitive cognitive work so they can afford to do more of everything else.
What the ‘Shadow AI’ Problem Is Costing You
One more thing the data makes clear: If your organization isn’t actively providing AI tools and training, your people are already finding their own.
EY found that between 23% and 58% of employees across different industries are bringing their own AI solutions to work—tools their employers haven’t vetted, approved or integrated into any coherent workflow. This shadow AI problem isn’t a sign of rogue behavior. It’s a sign that demand exists and supply is lagging.
The risk isn’t just data security, though that’s real. It’s that unsanctioned, uncoordinated AI use compounds the depth problem. People use whatever tool they grabbed, for whatever task felt obvious, in whatever way came instinctively. That’s how you get 88% usage and 5% impact.
EY’s data identified a threshold worth noting: Employees who receive more than 81 hours of AI training per year report productivity gains averaging 14 hours per week, nearly double the median eight hours reported by those with less training. The return on structured AI investment isn’t theoretical. It’s measurable, and it’s steep.
How to Cross the Line From 88% to 5%
The research across all three studies points in the same direction. Breadth of AI use is no longer the challenge. Depth is.
Audit how you’re using AI right now. Be honest about whether your current use is convenience-level or transformation-level. If AI is mainly accelerating tasks you already knew how to do, that’s the starting point, not the ceiling.
Push one workflow past the surface. Pick a single recurring task where you currently use AI for a first draft or a summary, and go one layer deeper. Ask it to challenge your assumptions, pressure-test your logic or generate the counterargument you haven’t considered. The compound value of AI shows up in iteration, not in single-pass outputs.
Treat training time as nonnegotiable. The EY data on training hours is unusually precise for this kind of research, and the implication is clear: Casual experimentation produces casual results. If your organization doesn’t offer structured AI development, seek it out elsewhere. The 14-hours-per-week productivity gain at the 81-hour training threshold represents a return that makes almost any investment in learning worthwhile.
Stop protecting what AI is going to change anyway. The 38% of workers who fear skill erosion are largely trying to preserve the parts of their work that AI is already improving everywhere else. The workers on the right side of this transition aren’t protecting their old workflows. They’re building new ones that are better because AI is in them.
The 5% aren’t lucky, and they aren’t superhuman. They’re the workers who decided that “using AI” and “getting results from AI” are not the same thing and acted accordingly.
Featured image from Supavadee Butradee/Shutterstock
