Talk Title: AI Agents as Employees
Abstract: Organizations are increasingly experimenting with “AI agents” that perform work alongside employees. Anecdotally, some firms are even beginning to describe these systems as “AI employees” or “teammates” and to place them on organizational charts—yet there is little systematic evidence on how such framing affects productivity, oversight, or beliefs about accountability. We provide new evidence on both the prevalence and consequences of organizations positioning AI as an employee using a sample of managers, directors, and executives in HR and Finance recruited through a professional expert network. Over a fifth of respondents report that their company already has an AI agent listed on its organizational chart, with adoption highest in large firms and in tech and financial services. We run a randomized experiment where we provide these managers with documents to review, holding the underlying work product fixed while varying whether documents are described as produced by an AI tool, a human employee, or an AI employee. We find that AI employee framing modestly worsens review performance, and it meaningfully shifts oversight behavior: among managers in organizations that already have "AI employees'', being exposed to the AI employee framing increases requests for additional review relative to AI tool framing. Together, our results show that seemingly subtle choices about whether AI is framed as a “tool” or an “employee” can shape monitoring effort, escalation decisions, and attitudes in AI-augmented work.