Imagine walking into work, only to discover a new team member has joined your ranks – not a person, but an AI tool. This digital newcomer, perhaps named Alex, comes with a job title and defined responsibilities, framed by your company as a true “employee.” How do you envision your collaboration with this artificial colleague?
If recent research is any indication, treating Alex as a genuine coworker rather than a sophisticated software tool might actually hinder your performance. A fascinating study by Boston University business professor Emma Wiles revealed a surprising truth: people caught 18% fewer errors when work was presented as coming from an “AI employee” compared to a mere chatbot. It turns out, the way we frame AI dramatically impacts our interaction and effectiveness.
The Rise of Digital Colleagues: A New Workplace Reality?
This unsettling insight offers a glimpse into a future aggressively promoted by Silicon Valley. Last year, Nvidia’s CEO Jensen Huang spoke of workplaces populated by “digital humans,” and major tech players like Microsoft, OpenAI, Anthropic, and Google have since released new tools designed to manage teams of AI agents. Many of these are explicitly marketed as digital colleagues, boasting the flexibility and cognitive power we usually associate with actual people.
The concept isn’t just theoretical; it’s already infiltrating corporate structures. Nearly a third of the 1,261 managers participating in Wiles’s study reported that their companies already frame AI agents as employees, with 23% even listing them on organizational charts. While the technical progress of agentic AI — AI tools programmed to work in a loop to achieve a goal — is genuinely impressive, labeling these tools as “coworkers” or “employees” is a significant leap.
This branding risks setting unrealistic expectations for AI capabilities, potentially leaving the human employees supposedly responsible for them in a worse position. Such framing can obscure the true nature of these tools, which, despite their sophistication, remain algorithms performing predefined tasks. The distinction between a tool and a team member is crucial for effective collaboration and accountability.
The Hidden Perils of Shared Responsibility
Wiles’s research suggests that framing AI as an employee fundamentally inverts our sense of who is in charge. When participants in the study believed an AI tool was a “coworker,” they perceived themselves as less responsible for its output. This shift in accountability has tangible, negative consequences for workflow and oversight.
The study found that participants were also 44% more likely to escalate questionable work produced by an AI “employee” to a human manager for further review, rather than trusting their own corrections. This not only adds unnecessary steps and burdens human supervisors but also negates the very time-saving purpose for which AI agents are often implemented. The efficiency promised by AI can be undermined by ambiguous roles.
The implications extend far beyond typical office dynamics. As AI agents become increasingly embedded in critical sectors like healthcare, warfare, education, and government, there’s a growing risk that they become a convenient scapegoat for failures. These failures are often, in reality, the product of flawed human decisions, misaligned incentives, or insufficient oversight, as exemplified by cases where AI has been blamed for outcomes rooted in human error.
Redefining the Human-AI Partnership for Success
“AI agents right now are being marketed as things that can replace humans, and I think that’s just a losing proposition,” asserts Daron Acemoglu, an MIT economist and Nobel laureate who studies AI’s impact on the economy. He advocates for a fundamental shift, arguing that AI should be optimized to improve human capabilities, rather than attempting to mimic or outright replace them. This collaborative model promises a more beneficial integration of technology.
What could this look like in practice? Consider a recent effort at Stanford, where researchers engaged 1,500 workers across 104 different jobs. Instead of dictating AI’s role, they provided information on potential AI tasks and then directly asked what would genuinely be most helpful and productive for the workers themselves.
The findings were insightful: while workers did desire automation in certain areas, such as law clerks wanting AI to track case progress, their preferences often diverged from tech expert assumptions. Tasks deemed highly suitable for AI by specialists, like verifying customer credit ratings for sales representatives, were frequently identified by the actual workers as things they explicitly did not want or need an agent to do. This highlights the importance of user-centric design.
Ultimately, calling an AI an “employee” is primarily a branding exercise, convenient perhaps when things go awry, but not a functional improvement. As Emma Wiles’s compelling research demonstrates, this framing doesn’t make the tool more effective; instead, it tends to make the humans working alongside it less effective at their own jobs. True progress lies in empowering human capabilities, not diluting human responsibility.
Source: MIT Tech Review – AI
