Supervisors were far less likely to trust an employee’s report when they knew AI had been involved: trust fell from 68 percent without AI to 24 percent with it. Their judgments weren’t based on the quality of the work itself. That’s the workplace bind highlighted by a new study: employees can follow instructions to use AI and still hurt their standing with the people who decide their next promotion.
The study, published in The Leadership Quarterly, was led by Samuel Meyers, a PhD student at the Hebrew University of Jerusalem. Researchers found that employees who doubled down on using tools such as ChatGPT were trusted less by employers, actively hurting their chances of promotion. The penalty appeared even in workplaces that actively encouraged AI use.
In a series of experiments, supervisors were asked to decide whether to trust an employee’s report and told whether the employee had used AI, without basing their judgment on the quality of the work itself. The percentages reflect those experimental decisions, not promotion rates across the workforce. Still, the gap points to a problem that better output alone may not solve: knowing how a report was produced can change a manager’s assessment of its author.
“AI is changing not only how we work, but also how we judge the people we work with,” study coauthor Yochanan Bigman, a senior lecturer at the Hebrew University of Jerusalem, said in a statement accompanying the research. “An employee may use AI to produce better work more efficiently yet still lose their manager’s trust simply because AI was involved.”
The findings come as some bosses change their attitudes toward AI-generated work. As Futurism has reported, many managers have spent years pushing employees to use AI, despite research questioning its productivity benefits and the tech industry’s promises of a transformed workplace. Now, some are pushing back against the low-value output often called “workslop”.
According to Business Insider, the CEO of a tech recruiting company was inundated with employee Slack messages she believed were clearly AI-generated. She responded by strictly banning AI-generated internal communications. The account names neither the CEO nor the company.
Another anonymous CEO went further. As Inc reported earlier this year, the executive threatened to “fire the next person” who sent them unedited ChatGPT output. Futurism interprets these reversals as bosses recognizing that extensive, often counterproductive AI use could do more harm than good, rather than improve white-collar work.
Costs are another concern. Executives have reported that AI agents and coding tools sent spending soaring, hurting their companies’ bottom lines and often prompting cuts and reversals. Meanwhile, a Forbes discussion of the AI productivity trap describes the lack of a meaningful economic windfall, and polling covered by Futurism indicates that the vast majority of CEOs haven’t seen productivity gains. Futurism calls that combination an expensive distraction, particularly as the job market deteriorates, with or without an AI-related productivity boost.
For workers, the immediate problem is contradictory expectations: follow their supervisors’ instructions to use AI, or earn their trust by ignoring those instructions. Bigman said organizations need to recognize that AI adoption brings psychological and managerial challenges as well as technological ones. Encouraging adoption doesn’t, by itself, remove the evaluation penalty described in the study.
Meyers and his colleagues worry that this growing tension could give rise to “shadow AI”: using the technology at work without disclosing it. That could further damage their career prospects, leaving workers exposed both when they acknowledge using AI and when they hide it.
