almost everyone is like that Utilize AI at work now. Very few people use it successfully.
That’s not a hot take. That’s what three major studies published in the past six months have consistently shown. AI has taken off faster than almost any other workplace technology in history. But the gap between having access to it and actually getting value from it could be what separates professionals in 2026.
Here’s what the data shows and that people on the right side of that gap are behaving differently.
numbers look better than reality
According to Gallup Q3 2025 Employee Survey Of more than 23,000 U.S. employees, 45% now use AI in the workplace at least several times a year, up from 40% last quarter. Frequent use, defined as several times a week or more, increased from 19% to 23% over the same period.
These are real benefits. However, upon closer inspection, the situation becomes more complex.
Only 10% of the U.S. workforce uses it every day. And a staggering 23% of employees say they don’t even know if their organization has one. AI technology not at all. Those furthest from that knowledge are individual contributors, 26% of whom report being unsure, compared to just 7% of leaders. AI is spreading across organizations, but its spread is uneven and its purpose is less clear than the number of deployments would suggest.
Gallup data also reveals what most employees are actually using AI for: integrating information (42%), generating ideas (41%), and learning new things (36%). These are useful applications. They are not transformative.
5% Matters — and Why It Matters to You
Here is EY 2025 Rethinking Work SurveyThe company, which has mobilized 15,000 employees and 1,500 employers in 29 countries, is uncomfortable.
88% of employees report using AI at work. However, the majority limit themselves to surface-level tasks such as searching, summarizing, and drafting. Only 5% are leveraging AI in ways that fundamentally change how they work and what they produce. EY research found that this gap is holding organizations back up to 40% of potential AI productivity gains.
It’s not a technology issue. It’s human.
The study identified a clear cause: Only 12% of employees say they are adequately compensated AI training To unlock something beyond the basics. If your employees don’t know how to break through the first layer What can AI do?,it’s not. They use it as something safe and familiar: a smarter search engine, a faster way to draft their first paragraph, and move on.
So what does this mean for you? If using AI feels like a more convenient version of what you were doing before, you’re probably in the 88%. 5% are not using radically different tools. They are using the same tools for fundamentally different purposes.
The element of fear that no one talks about enough
EY data also reveals that leaders rarely mention it directly. That is, fear is actively inhibiting how deeply people engage with AI.
38% of employees surveyed said they fear losing their jobs to AI. The same proportion are concerned that over-reliance on AI will undermine their own skills, expertise and judgment over time. These two concerns are not contradictory. In fact, they have the same anxiety but in different directions. Fear of being replaced by AI and fear of leaving it to us.
As a result, defensive behavior occurs. Workers use AI enough to keep it up to date, but refrain from deeper experimentation that would actually change outcomes. They are hedging. And that comes at the expense of the interests they are trying to protect.
According to EY research, in organizations where leaders are successful in driving full-scale AI adoption, 75% of employees report that their leadership teams are clearly aligned on the vision for AI and, importantly, communicate not only what AI will do, but also the human behaviors that AI will make more valuable. This message is the difference between an experimental workforce and a risk-averse workforce.
What an “AI power user” actually looks like
Here are some discoveries that always surprise people: The workers who use AI the most are not alienated from human connection.
Gensler’s 2026 Global Workplace SurveyThe survey, which gathered responses from more than 16,400 office workers in 16 countries, identified 30% of employees as “AI power users.” This is defined as people who regularly incorporate AI in both their professional and personal lives. When researchers compared this group to workers with lower AI adoption rates, the differences were striking.
AI power users spend less time working alone: 37% of their weekly hours compared to 42% of late adopters. They spend more time learning: 12% versus 8% of their working week. They spend more time socializing: 11% vs. 9%. The narrative that AI forces people into digital isolation does not hold true. Workers who are most integrated into AI workflows can spend more time on the distinctly human parts of their jobs.
Janet Pogue McLaughlin, global director of workplace research at Gensler. put directly: “Employees who are most integrated into AI workflows are also the ones who are most engaged in learning and have better team relationships. This shift signals the creation of important new roles in the workplace.”
AI will not replace their collaboration. It handles solitary, repetitive cognitive tasks so well that you have more leeway to do everything else.
The damage caused by the “shadow AI” issue
Another thing the data reveals is when organizations aren’t actively providing information. AI tools And with training, employees are already starting to find their own way.
According to EY research, between 23% and 58% of employees across industries have implemented their own AI solutions in the workplace, tools that have not been vetted, approved by their employer, or integrated into a consistent workflow. This shadow AI issue is not a sign of fraud. It shows that while demand exists, supply lags.
The risk isn’t just data security, but it’s real. That is, the use of unsanctioned and uncoordinated AI further exacerbates the depth problem. People use whatever tools they have at hand, in ways that instinctively occur to them, for tasks that seem obvious. This gives you 88% utilization and 5% impact.
EY data identified threshold worthy of attention: Employees who receive 81 or more hours of AI training per year report an average increase in productivity of 14 hours per week. This is nearly double the median 8 hours reported by untrained employees. The return on investment for structured AI is not theoretical. It is measurable and steep.
How to cross the line from 88% to 5%
Research across all three studies points in the same direction. Widespread use of AI is no longer a challenge. The depth is.
-
Audit how you are currently using AI. Be honest about whether your current usage is at a convenience level or a transformation level. If AI primarily accelerates tasks that it already knows how to do, that’s a starting point, not an upper limit.
-
Push one workflow from the surface. Choose one recurring task where you currently use AI in your first draft or outline, and go one layer deeper. Ask them to question their assumptions, pressure test their logic, and generate counterarguments they haven’t thought of. Composite values in AI appear in iterations rather than in single-pass output.
-
Treat training time as non-negotiable. EY’s data on training time is unusually accurate for a study of this type, and the implications are clear. In other words, casual experiments produce casual results. If your organization doesn’t offer structured AI development, look elsewhere. A productivity increase of 14 hours per week at an 81-hour training threshold represents a return that makes nearly every investment in learning worthwhile.
-
Let’s stop trying to protect what AI is trying to change anyway. Thirty-eight percent of workers concerned about skills decline are primarily looking to maintain parts of their jobs that have already been improved elsewhere by AI. The workers on the right side of this transition are not adhering to the old workflow. They are building new things that are better because they have AI built in.
The 5% are neither lucky nor superhuman. These are workers who decided that “using AI” and “obtaining results from AI” are not the same thing, and took action.
Featured image from Supavadee Butradee/Shutterstock
