Agency theory
Theoretical framework analyzing conflicts of interest between a principal and their agent, and the resulting economic costs when their objectives diverge.
Agency theory is an institutional microeconomics framework, formalized notably by Jensen and Meckling in their landmark 1976 article in the Journal of Financial Economics. It models the relationship between a principal who delegates a task and an agent, when information is asymmetric and interests diverge. The principal cannot perfectly observe what the agent does (moral hazard), nor what the agent was before the relationship (adverse selection). To align behavior, the principal bears agency costs: monitoring costs, bonding costs, and residual loss, the value destroyed by the irreducible share of divergence. Agentic AI forces a rereading of the theory. Algorithmic agents neutralize the traditional divergence of interests (they have no union and no personal agenda), but they create an information asymmetry of a radically new nature. Deep neural networks are opaque (black box): the human principal cannot retrace the agent's reasoning in high-dimensional vector spaces. Algorithmic hallucination thus becomes the modern equivalent of residual loss. The gap now sits in representation, between reality and the agent's output. This reversal of the information structure changes the conditions for reliable delegation.
Textbook case: an insurance company has automated claims handling, and its claims director (principal) hands file analysis to an LLM agent. The classical divergence of interests disappears: the agent has no personal reason to bias its analysis. Information asymmetry, though, changes form. The director cannot retrace why the agent rejected a given file, and so struggles to give the reasons for the rejection that the law requires.
Follow the research
One letter a week. First, the week's underwriting note, when one is published: a dated, argued decision, signed by Paul Bernardski. Then what the observatory published on cyber risk, political risk and insurance. No account, no tracking.
No spam, one-click unsubscribe, no data ever sold.