A new study from Boston University business professor Emma Wiles reveals a counterintuitive effect of personifying AI tools. When an AI agent named Alex was presented as an employee rather than a chatbot, managers caught 18% fewer errors in its work. This finding challenges the Silicon Valley push to treat AI agents as digital coworkers.

What You Need to Know

The study involved 1,261 managers and tested how they reviewed work from an AI tool labeled either as a chatbot or as an AI employee. Nearly a third of managers said their companies already frame AI agents as employees. The research suggests that calling AI agents coworkers reduces human accountability and oversight.

Study Details

Wiles designed an experiment where managers reviewed work attributed to either a standard chatbot or an AI employee named Alex. The results showed a clear pattern of diminished human performance when the AI was personified.

  • Error detection drop: Participants caught 18% fewer errors when the AI was called an employee.
  • Responsibility shift: Managers were 44% more likely to escalate questionable work to a supervisor when it came from an AI employee.
  • Org chart inclusion: 23% of companies in the study list AI agents on organizational charts.

This pattern inverts the intended benefit of AI agents. Instead of saving time, the framing creates more managerial overhead. The trend is accelerating, and companies like Microsoft, OpenAI and Anthropic have released tools for managing teams of AI agents.

Why This Matters

The implications extend beyond office productivity. As AI agents enter health care, warfare and government, the framing of these tools as employees creates a dangerous diffusion of responsibility. Blame for failures may be conveniently shifted to the AI, obscuring human errors in design and oversight. MIT economist Daron Acemoglu, a Nobel laureate, warns that marketing AI as a replacement for humans is a losing proposition. He argues that AI should be optimized to augment human capabilities.

Stanford researchers found a similar mismatch in worker expectations. Law clerks wanted AI to track case progress, but sales reps did not want AI to verify customer credit ratings. The mismatch highlights the danger of imposing AI roles without human input.

This brings us back to Alex. Calling an AI tool an employee is a branding exercise that makes the humans around it worse at their jobs. They deserve better than Alex.