A recent study from Boston University has revealed a counterintuitive effect of how companies present artificial intelligence tools. When an AI agent is labeled as an 'employee' or 'coworker', humans tend to reduce their vigilance and make more errors in checking its work. The research, conducted by professor Emma Wiles, found that participants caught 18% fewer errors when the work was attributed to an 'AI employee' rather than a simple chatbot. This finding raises significant questions about the future of AI integration in workplaces.
Names matter: the psychological effect of terminology
The study involved 1,261 managers, nearly a third of whom reported that their companies already frame AI agents as employees, with titles and defined responsibilities. 23% even list them on organizational charts. Wiles discovered that treating a software tool as a human coworker inverts the perception of responsibility. Participants felt less responsible for the output and were 44% more likely to escalate questionable work to a supervisor, thus negating the time-saving purpose of using the agent. This phenomenon, Wiles explains, is dangerous because it reduces the sense of control and the ability to correct errors.
Sponsored Protocol
Tech companies push AI agents as 'digital colleagues'
Major names like Microsoft, OpenAI, Anthropic, and Google have recently released tools oriented towards managing teams of AI agents, often advertised as digital colleagues with human-like cognitive abilities. Nvidia CEO Jensen Huang has also talked about 'digital humans' in the workplace. But according to Daron Acemoglu, an MIT economist and 2024 Nobel laureate, this marketing is misleading. Acemoglu argues that AI agents should be optimized to enhance human capabilities, not to replace them. Technical progress in agents is real, but calling them 'coworkers' creates unrealistic expectations and harms human performance.
A Stanford study confirms: workers know what they want from AI
Another study, conducted at Stanford with 1,500 workers across 104 occupations, asked them which tasks they would actually like to automate. The results show a gap between what tech experts deem suitable for AI and what workers desire. For example, law clerks wanted assistance monitoring case progress, while credit rating verification, considered ideal by experts, was among the tasks workers definitely did not want to delegate. This suggests that a collaborative approach, where AI supports without replacing, is more effective.
Sponsored Protocol
The risk of a new 'blame culture'
Labeling an AI agent as an 'employee' could become a convenient scapegoat. There have already been incidents where AI was blamed for human errors, such as the bombing of a girls' school in Iran initially attributed to Claude. Wiles' research shows that labeling influences human behavior: if the tool is perceived as a coworker, people feel less responsible. For a comparison of different AI chatbots and their practical applications, read our article Which AI chatbot suits you? Comparing ChatGPT, Claude, Gemini, Grok and Perplexity for specific needs. Additionally, the ethical debate around AI use intensifies, as seen in the controversy over Netflix using AI-generated Gene Wilder voice.
Sponsored Protocol
Conclusions: call things by their name
Ultimately, the research suggests that calling an AI agent 'Alex the employee' is just a branding exercise. It does not make the tool more fit for the job and, on the contrary, worsens the performance of the humans working with it. As Acemoglu emphasizes, the goal should be to augment human capabilities, not replace them. For more on the concept of intelligent agents, see the Wikipedia entry. Humans deserve tools that help them do better, not fake coworkers that make them less attentive.
Source: https://www.technologyreview.com/2026/06/29/1139849/ai-agents-are-not-your-coworkers
