AI Colleague 'Alex' or Nameless Chatbot? The Name Affects Your Error Rate, Warns Researcher

Imagine you arrive at work and find out that you have a new colleague. However, this "person" is not a human but an AI tool. Your company calls it "Alex," and it has a job title and defined tasks as an "employee." How well do you think you would collaborate with Alex?

If you resemble the managers recently surveyed by Emma Wiles, an economics professor at Boston University, treating Alex as a "colleague" rather than a software tool would lead you to perform worse. Wiles noted in a paper that people identified 18 percent fewer errors when the work was supposedly done by an agentic "AI employee" rather than a regular chatbot. This shows that how we address the AI tool makes a significant difference.

Teams of AI Agents Enter the Workforce

This is an alarming insight into a future that Silicon Valley is currently steering us towards. Last year, Jensen Huang, CEO of Nvidia, spoke about work environments with "digital humans." Since April, Microsoft, OpenAI, Anthropic, and Google have gradually released tools aimed at managing teams of AI agents; many of these are explicitly marketed as digital colleagues with the flexibility and cognitive abilities of real humans. Almost a third of the 1,261 executives who participated in Wiles’ study indicated that their companies already consider AI agents as employees (23 percent even include them in organizational charts).

The technological advancement in agentic AI is, of course, not just hot air. Agents, which can essentially be thought of as AI tools programmed to work in a loop until they achieve a goal, have indeed become measurably better at complex tasks. However, it is a huge leap to label these programs as colleagues or employees, which raises unrealistic expectations about AI's capabilities while human employees, who are supposedly responsible for them, are left in a worse position.

Study Shows How Employees Shift Responsibility to AI Agents

Part of the reason for this is that, as Wiles’ research suggests, it upends our understanding of who holds authority in a company. When an AI tool was portrayed as an employee, study participants felt less responsible for its outcomes. They were also 44 percent more likely to forward questionable AI-generated results to a supervisor for further review, rather than trusting their own corrections (which would negate the time-saving purpose of using AI agents in the first place).

This has implications far beyond office culture: As AI agents are embedded in healthcare, education, government institutions, and even warfare, the risk grows that they become a convenient scapegoat for failures. Otherwise, these would have been the result of poor human decisions, incentives, and negligence, but now it’s “the AI.” (Just remember how the bombing of a girls' school in Iran was publicly blamed on Claude, despite many signs pointing to a chain of human errors.)

"AI agents are currently marketed as something that can replace humans, and I consider that a futile endeavor," says Daron Acemoglu, an economist at MIT who received the Nobel Prize in 2024 and studies the impact of AI on the economy. "They should instead be optimized to enhance human capabilities – which is not the case at the moment."

Survey: What Tasks Could AI Potentially Take Over?

What might this look like? Insight comes from a new project at Stanford University, where researchers presented information to 1,500 employees in 104 professions about which tasks AI could potentially take over in their work. The scientists then asked them what would actually be the most helpful and productive. Employees indeed wanted automation in certain areas: for example, legal interns were convinced that AI could help make meaningful progress in many legal cases. However, often it was precisely the tasks that technology experts deemed "best suited for AI" – such as credit checks by sales representatives – that employees definitely did not want an AI agent to handle.

This brings us back to the AI agent Alex. Referring to Alex as an employee is easy – and convenient, especially when something goes wrong. But it’s pure branding. It doesn’t make the tool better suited for the tasks, and as Wiles’ research shows, it worsens the performance of people in their actual work environment. And the "human employees" should remember: they are also the ones whose agency the AI is trying to mimic. They deserve better than an Alex.