A new annual survey on how employees actually use artificial intelligence at work suggests the office culture around AI has shifted in the past year, and not entirely in the direction companies might expect. Fewer workers now admit to bluffing their way through meetings or quietly passing off AI generated work as their own, a sign that the taboo around faking AI competence appears to be fading.
But the same research points to a different, more practical problem taking its place: a lot of employees who say they are comfortable with AI tools are not actually getting good results from them, and are burning time trying to make the technology work rather than simply doing the task themselves.
Less Faking, More Confidence
According to the survey, the share of workers who admitted to pretending they understood AI tools during a meeting fell sharply compared with the year before, and the number who confessed to presenting AI generated work as their own also dropped by a similar margin. Both figures suggest that AI use has become normalized enough that employees no longer feel as much pressure to hide it.
That shift also shows up in how workers describe their own skills. The vast majority now say they feel confident using AI tools on the job, a marked change from the uncertainty and stigma that surrounded early workplace AI adoption.
Confidence Without Competence
The trouble, the survey suggests, is that confidence has outpaced results. Only about a quarter of workers said an AI tool actually delivered what they needed on the first attempt, meaning most people are having to retry, rephrase or abandon their prompts before getting something usable. Roughly half of respondents went further, saying they had spent more time trying to get an AI tool to complete a task than the task would have taken to do manually in the first place.
That gap between how capable workers feel and how well the tools actually perform points to a subtler drag on productivity than outright dishonesty. Instead of workers hiding poor results, many appear to be spending real time chasing AI output that never quite lands, without necessarily recognizing the lost time as a cost.
Why It Matters
For employers who have poured money into AI tools expecting quick productivity gains, the findings are a reminder that access to the technology and comfort with it are not the same as using it well. A workforce that no longer feels the need to fake AI fluency is a step forward for honesty inside companies, but if a large share of that same workforce is quietly wasting hours wrestling with tools that do not deliver, the productivity payoff many executives are counting on may be smaller and slower to arrive than expected.
The findings suggest that training and workflow design, not just access to AI tools themselves, may end up being the deciding factor in whether companies actually see a return on their AI investment.