Not yet human-reviewed
This source record was written automatically by the research pipeline and has not yet been reviewed by the maintainer or integrated into a topic page. Treat it as a provisional summary and verify claims against the original source.
When AI Becomes an Agent of the Firm: Examining the Evolution of AI in Organizations Through an Agency Theory Lens
Citation. Humberd, B. K., & Latham, S. F. (2025). When AI Becomes an Agent of the Firm: Examining the Evolution of AI in Organizations Through an Agency Theory Lens. Journal of Management Studies, 63(2), 668–694. https://doi.org/10.1111/joms.13274
Summary. This peer-reviewed conceptual article uses agency theory to model how an AI system may progress from executing routines to functioning as an autonomous agent with consequential decision rights in a firm. It argues that agency mechanisms must evolve with AI capability because incentives and monitoring designed for human managers do not straightforwardly transfer to AI agents. The authors propose staged monitoring and alignment mechanisms intended to preserve principal–agent alignment as AI systems become more autonomous and interconnected.
Key claims
- The article theorizes four stages of AI evolution—routine, machine-learning, generative, and agentic AI—whose increasing autonomy and decision rights progressively alter the agency problem within firms.
- Monitoring should develop from verification and examination to cross-boundary cooperation and transparency as AI systems become more capable, interconnected, and difficult for human principals to observe directly.
- Incentive alignment may shift from training-based reinforcement at the generative stage to resource-provision mechanisms at the agentic stage, because wealth-based incentives that align human agents do not apply to AI agents.
- Firms should introduce agency mechanisms before AI reaches agentic autonomy rather than treating oversight as a response after consequential authority has already been delegated.
Evidence & limitations
- This is a peer-reviewed theoretical exercise that develops propositions by integrating agency theory with a staged model of AI evolution; it does not report a new empirical test of the mechanisms.
- Its account of agentic AI is prospective, so the feasibility and effectiveness of its proposed monitoring and alignment mechanisms remain empirical questions.