The Illusion of AI Coworkers: Why Treating AI Agents as Employees Backfires

The notion of AI agents as "coworkers" is gaining traction in the tech industry, but new research indicates this framing can be detrimental. A study by Boston University business professor Emma Wiles revealed that when participants believed work originated from an "AI employee" rather than a mere chatbot, they caught 18% fewer errors. This highlights the significant impact of how we name and perceive these advanced tools. This trend is not merely theoretical; it's actively being pushed by Silicon Valley giants. Nvidia's CEO has envisioned workplaces with "digital humans," while Microsoft, OpenAI, Anthropic, and Google have all launched tools promoting AI agents as flexible, cognitively powerful "digital colleagues." Alarmingly, nearly a third of managers in Wiles's study reported their companies already categorize AI agents as employees, with 23% even listing them on organizational charts. While the technical advancements in agentic AI — tools programmed to achieve goals in a loop — are real, equating them to human coworkers is a substantial leap. Such a framing sets unrealistic expectations for AI's capabilities and, crucially, negatively impacts human employees. Wiles's research suggests this approach inverts the sense of responsibility, making human participants feel less accountable for the AI's output. They were also 44% more likely to escalate questionable AI work to a manager instead of correcting it themselves, undermining the very efficiency AI is supposed to provide. The implications extend far beyond office dynamics. As AI agents become integrated into critical sectors like healthcare, warfare, education, and government, there's a growing risk that they will become convenient scapegoats for failures that are, in fact, products of human error, poor decisions, or inadequate oversight. Daron Acemoglu, an MIT economist and Nobel laureate, emphasizes that marketing AI agents as human replacements is a "losing proposition." Instead, he advocates for optimizing AI to enhance human capabilities, a goal currently not being met. So, what does a more constructive approach look like? Researchers at Stanford explored this by asking 1,500 workers across 104 jobs what AI tasks would genuinely be most helpful. While workers welcomed automation in specific areas — like law clerks wanting AI to track case progress — they often rejected tasks that tech experts deemed suitable for AI, such as sales reps not wanting agents to verify customer credit ratings. This suggests a disconnect between what AI can do and what humans need it to do to be truly augmentative. Ultimately, labeling an AI tool an "employee" might be convenient for branding or blame deflection, but it doesn't improve the tool's fitness for the job. As Wiles's study clearly demonstrates, it makes the human workers around it perform worse. Humans possess the agency that AI attempts to mimic, and they deserve systems designed to empower, not diminish, their capabilities. The future of AI should be about collaboration and augmentation, not an illusion of digital coworkers.