In recent years, some companies have begun incorporating artificial intelligence (AI) agents as formal members of their workforce, even listing them in organizational charts. While this trend is intended to boost productivity and position firms at the forefront of technology, emerging research suggests that the implications may be more complex and problematic than anticipated.

Emma Wiles, a professor at Boston University specializing in the impact of AI on labor, along with collaborators from Boston Consulting Group, conducted studies that highlight potential pitfalls of treating AI agents as employees. Their research revealed that managers responsible for reviewing work attributed to AI employees were significantly less likely to detect errors than those reviewing human-produced work or outputs described as generated by AI tools. This suggests a diminished sense of accountability toward AI-generated tasks among managers.

Dr. Wiles theorizes that managers may believe mistakes made by AI employees are not their responsibility, possibly attributing faults to technical teams or executives who advocate for AI integration. “But it’s not your problem,” she summarized as a reflection of managerial attitudes regarding AI oversight.

The rise of AI in workplaces has brought increased awareness of certain known flaws, including biases within AI models, inaccurate or misleading responses from chatbots, and inadvertent disclosures of sensitive information. However, research indicates there are more subtle, underexplored issues emerging as companies push AI deeper into everyday operations.

One such concern involves an implicit bias within AI systems favoring AI-generated work over human-produced content. A 2025 study published in The Proceedings of the National Academy of Sciences documented that several large language models exhibited a marked preference for text created by AI, which may have far-reaching implications for hiring and evaluation processes. In a follow-up study, researchers found that AI tools commonly used in resume screening are prone to favor applications crafted with AI assistance, raising alarms for recruitment fairness.

Jane Yi Jiang, an operations professor at Ohio State University and co-author of the follow-up paper, noted that while some recruiters are beginning to address these issues, many companies have yet to fully consider the broader consequences of rapid AI adoption. “People are moving so fast to use large language models without thinking too much about the implications, biases,” she said.

Another emerging concern involves AI’s decision-making frameworks, which tend to adopt a rigidly rational approach rooted in game theory. This mindset may lead AI-driven pricing strategies or market expansion decisions to prioritize aggressive competition, risking adverse outcomes like prolonged price wars that harm all parties involved. Jiannan Xu, a Ph.D. candidate at the University of Maryland and collaborator with Dr. Jiang, explained that AI systems often assume humans behave more rationally than they do, which can result in recommendations that prove counterproductive.

Though these biases and flaws can, in principle, be mitigated through careful design and accountability measures, many corporations remain unaware or unprepared. A survey conducted by Dr. Wiles and her team of more than 1,000 corporate managers found that about one-third of organizations referred to AI tools as “teammates or employees,” and nearly one-quarter included AI agents within formal organizational structures.

The research also included an experimental exercise in which managers reviewed documents containing deliberate errors, with the work attributed either to AI employees, AI tools, or human workers. Results indicated that managers in firms that treated AI as employees were less vigilant in spotting mistakes when told the work came from AI employees, compared to other groups.

Dr. Wiles emphasized that long-established management practices developed for human teams cannot be easily transplanted to AI. The psychological dynamics are fundamentally different, and organizations are “going out there blind” in managing these new entities. She expressed concern about future developments, including companies reportedly planning for AI employees to supervise human workers.

As companies continue integrating AI in increasingly complex roles, experts caution that understanding and addressing these novel challenges will be critical to realizing AI’s benefits without compromising organizational effectiveness or fairness.