Calling AI tools 'coworkers' makes managers perform worse
managers caught 18% fewer errors when ai was presented as an 'employee' versus a tool, suggesting anthropomorphizing agents undermines oversight.
a study by boston professor emma wiles reveals that managers caught 18% fewer errors when ai was framed as an 'ai employee' with a title compared to a standard chatbot. this finding challenges the current industry trajectory where firms like microsoft, openai, and anthropic are aggressively marketing 'ai agents' as digital coworkers or teammates. anthropomorphizing these tools appears to lower user vigilance, creating a new category of human error in human-ai collaboration. as companies deploy fleets of autonomous agents, this research suggests that framing ai as subservient tools rather than peers may be critical for maintaining quality control.