Framing AI agents as employees can weaken oversight

Companies are increasingly presenting AI agents as digital colleagues, with tools from Microsoft, OpenAI, Anthropic, and Google aimed at managing teams of agents. Boston University professor Emma Wiles found that the label changes how people supervise these systems: participants caught 18% fewer errors when work was described as coming from an agentic “AI employee” rather than a chatbot.

In Wiles’s study of 1,261 managers, nearly a third said their companies already frame AI agents as employees, and 23% said the systems appear on org charts. When framed as employees, AI tools made managers feel less responsible for their outputs and 44% more likely to escalate questionable work to a manager instead of trusting their own corrections.

The concern extends beyond workplace efficiency as agents move into health care, warfare, education, and government. MIT economist Daron Acemoglu, who won the Nobel Prize in 2024, said AI agents are being marketed as human replacements, but should instead be designed to improve human capabilities. Stanford research with 1,500 workers across 104 jobs also found gaps between tasks experts saw as suitable for AI and tasks workers actually wanted automated.