A 2025 US survey, reported via Fortune, asked working adults about their AI use on the job. The headline number is the uncomfortable one: nearly half of employees who use AI at work admitted they'd hidden or downplayed how much they actually rely on it.
Nearly half of employees who use AI at work admit they've hidden or downplayed it. (WalkMe survey via Fortune, 2025)
The reasons people give aren't complicated: a worry that visible AI use reads as "not really doing the work", a fear that admitting reliance on a tool signals a skills gap, and, for some, a straightforward concern about job security if a manager starts wondering what the tool could replace. None of these worries require AI use to actually be a problem, they just require an audience who might judge it as one.
That's a real cost, separate from whether the AI use itself is any good. If people are quietly hiding how they work, they're also quietly not asking for help improving it, which means the gap between confident use and effective use (the exact gap the METR studies measured) goes unaddressed for longer than it needs to.
A cohort course or a shared workshop doesn't fix this, it can make it worse: now the "am I doing this wrong" question has to survive being asked in front of other learners. A recorded course removes the audience but also removes the feedback, so mistakes and bad habits go uncorrected.
1-to-1 coaching removes the audience without removing the feedback. No cohort, no forum, no colleague watching you ask the question everyone actually has. Just the work, and someone whose job is to help you get better at it, not to judge how much of it you're already doing with AI.
Source: WalkMe survey via Fortune, 2025.