A study by Boston University business professor Emma Wiles found that managers caught 18% fewer errors when work was attributed to an AI employee rather than a chatbot. The research highlights how naming AI tools as coworkers, such as calling them "Alex" with titles and responsibilities, affects human oversight. Nearly a third of 1,261 managers surveyed said their companies already treat AI agents as employees, with 23% listing them on organizational charts, according to technologyreview.com.
Wiles’s study involved managers evaluating work outputs labeled either as coming from an AI employee or a chatbot. The results showed that when AI was framed as a digital colleague, managers were less vigilant in spotting mistakes. This suggests that anthropomorphizing AI agents may impair human judgment. The trend toward presenting AI as digital coworkers is growing, with companies like Microsoft, OpenAI, Anthropic, and Google releasing tools designed to manage teams of AI agents with human-like flexibility and cognitive abilities.
The findings underscore concerns about the evolving role of AI in workplaces, where tools are increasingly personified and integrated into organizational structures. Nvidia CEO Jensen Huang has described future workplaces featuring "digital humans," reflecting this shift. The study signals potential risks in overtrusting AI agents when they are perceived as employees rather than software tools, which could impact error detection and decision-making quality in business environments.
Emma Wiles’s research was published in The Algorithm newsletter by technologyreview.com on June 29, 2026. The study’s data comes from a survey of 1,261 managers and experimental evaluations of error detection rates tied to AI agent framing.