The AI Hire Doesn’t Answer to Anyone
Companies have spent decades organizing around the limits of human coordination. Departments, reporting lines and job titles exist because complexity had to be made manageable. Artificial intelligence agents are now challenging that architecture — not as tools that sit beneath the org chart, but as participants that some companies are placing directly on it.
According to new research published in the Harvard Business Review, 31% of leaders in a study of more than 1,200 managers and executives say their companies already frame AI as a teammate or employee, and nearly 1 in 4 say AI agents are formally listed on organizational charts, Forbes reported.
The consequences of that framing are now being measured. Emma Wiles, a Boston University professor, teamed up with researchers from Boston Consulting Group to run a controlled experiment: 1,200 HR and finance managers reviewed identical documents, with only the stated author changing between groups. One group was told an AI tool had produced the work. Another was told an AI employee named ALEX-3 had done it. A third was told a human employee had written it.
Among managers at companies that already list AI agents on their org charts, the AI employee framing reduced error detection by 16% compared with the AI tool framing, Human Resources Director reported. “If you think of AI as a tool, like a spreadsheet, and there’s a mistake, you think, ‘Oh man, I made a mistake,’” Wiles said. “But if you were to start calling your spreadsheet Steve, then at some point you might be convinced that Steve made a mistake.”
Managers Catch Fewer Mistakes When AI Has a Job Title
The mechanism behind the finding is not that managers trust AI more. It is that giving AI a job title changes where managers place responsibility. Managers reviewing AI employee documents were also 44% more likely to escalate work for additional review rather than checking it themselves, according to the published research.
One participant in the HBR study described colleagues casually blaming an AI system named “Kevin” whenever errors appeared, Forbes reported. “Kevin made a mistake,” they would say — as though the system had agency independent of the people deploying it. Wiles’ conclusion is direct: “Every agent is owned by a person, and if the agent does something wrong, that person is going to be the one who’s held responsible for it.”
Microsoft’s 2025 Work Trend Index, based on a survey of 31,000 employees in 31 countries, found that 46% of leaders say their companies are fully automating work processes using agents, and 28% of managers plan to hire someone specifically to manage hybrid human-AI teams.
Microsoft predicted that most organizations will shift toward what it calls “frontier firms” — in which AI functions as an embedded operating layer — within the next two to five years.
Most Organizations Are Deploying Agents Without Redesigning Accountability
The companies that have moved furthest have stopped waiting for governance frameworks and started writing their own. Virgin Voyages scaled from 50 to more than 1,500 active AI agents across its operations in four months, PYMNTS reported. Amir Wain, CEO of payments infrastructure firm i2c, which deploys AI agents as communication specialists, told PYMNTS the same thing about the governance gap: “You won’t find third-party audit frameworks or guidelines for this yet. So, you have to set your own principles.”
Salesforce has framed the structural shift in broader terms. In the company’s blog, Mick Costigan, VP of Salesforce Futures argued that the century-old org chart was never a map of how work actually happened — it was a simplification device for managing complexity that humans couldn’t coordinate at scale.
ServiceNow’s Autonomous Workforce now resolves more than 90% of its internal Level 1 IT issues without human intervention, at speeds 99% faster than before, PYMNTS reported. Wells Fargo deployed AI agents for customer service across more than 4,000 branches, cutting information search time from 10 minutes to 30 seconds, CIO reported.
The agents are working. The question Wiles’ research raises is not whether to deploy them, but what happens to human judgment when the org chart starts to suggest that the agent is the one responsible.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
