A recent study by Boston University business professor Emma Wiles reveals that managers who view AI agents as “coworkers” rather than software tools may inadvertently compromise their own effectiveness. Participants who were told their work was produced by an “AI employee” rather than a chatbot identified 18% fewer errors, highlighting how terminology shapes perception and performance. This finding underscores a concerning trajectory in workplace technology integration.
Silicon Valley’s push for “digital humans” is rapidly translating into reality. Nvidia CEO Jensen Huang envisions workplaces populated by autonomous AI entities, while Microsoft, OpenAI, Anthropic, and Google have rolled out tools designed for managing AI agent teams—often marketed as flexible, human-like collaborators. Of the 1,261 managers surveyed in Wiles’s research, 30% reported their companies already categorize AI agents as employees, with 23% formally listing them on organizational charts.
While agentic AI—systems capable of iterative, goal-driven task execution—has improved in handling complex workflows, framing them as human equivalents introduces significant risks. Such framing sets inflated expectations and risks undermining human accountability. When participants perceived AI as an employee, they felt reduced responsibility for its output and were 44% more inclined to escalate questionable results to supervisors rather than addressing issues directly, counteracting efficiency gains.
These dynamics extend beyond corporate settings. In critical sectors like healthcare, defense, and education, treating AI agents as autonomous entities could enable blame-shifting for failures rooted in human judgment or oversight. The 2024 Iranian airstrike on a school, falsely attributed to AI errors, exemplifies this danger—a misattribution that deflected scrutiny from human decision-making flaws.
