AI agents aren’t your coworkers – here’s why
▼ Summary
– People caught 18% fewer errors when work came from an “AI employee” rather than a chatbot, according to a study by Boston University professor Emma Wiles.
– Nearly a third of managers surveyed said their companies already frame AI agents as employees, with 23% listing them on org charts.
– Treating AI tools as coworkers sets unrealistic expectations and leaves human employees worse off, according to the research.
– When AI was framed as an employee, participants felt less responsible for its output and were 44% more likely to escalate its work to a manager.
– Framing AI agents as employees risks creating a way to blame failures on AI instead of human errors, especially in high-stakes fields like health care and warfare.
Imagine you’re a manager, and you’re handed work supposedly done by an AI employee rather than a simple chatbot. According to new research from Boston University professor Emma Wiles, you’d likely catch 18% fewer errors in that work. Why? Because calling an AI a “coworker” changes how we interact with it , and not for the better. The label matters far more than most of us realize.
This finding offers a troubling preview of the workplace future Silicon Valley is pushing toward us. Nvidia CEO Jensen Huang has already predicted offices filled with “digital humans.” Since April, major players like Microsoft, OpenAI, Anthropic, and Google have unveiled tools designed to manage teams of AI agents, often marketed as digital colleagues with human-like flexibility and reasoning. In Wiles’s study, nearly a third of 1,261 managers reported that their companies already treat AI agents as employees, with 23% even placing them on official org charts.
The technical strides in agentic AI are real. These tools, programmed to loop through tasks until they reach a goal, have become measurably better at handling complex assignments. But calling them coworkers or employees is a dangerous leap. It sets unrealistic expectations for what AI can actually do , and it leaves the humans supposedly overseeing them worse off.
Part of the problem, Wiles found, is that the label flips our sense of responsibility. When an AI tool was introduced as an employee, study participants felt less accountable for its output. They were 44% more likely to escalate questionable results to a manager rather than trust their own corrections. That defeats the whole purpose of using an AI agent to save time in the first place.
This dynamic extends far beyond office culture. As AI agents are rolled out in health care, warfare, education, and government, the risk grows that they’ll become a convenient scapegoat for failures that actually stem from flawed human decisions, incentives, or oversight. Consider the tragic bomb strike on a girls’ school in Iran, widely blamed on Claude , even though all evidence points to a cascade of human errors. The real danger isn’t that AI will replace us; it’s that we’ll let it take the fall for our own mistakes.
(Source: MIT Technology Review)
