The Price of Judgment
The Price of Judgment
Executives as Institutions:
Pesawala Digest Discusses
An Economic Case Against CEO Automation
The claim that CEOs are “hugely expensive” and therefore prime candidates for automation reflects a familiar economic intuition: when the price of an input is high, firms search aggressively for substitutes.
In an era of rapid advances in artificial intelligence, it is tempting to view top executives as an inefficient bottleneck: one individual paid the equivalent of thousands of workers, making judgment calls that might be codified, optimized, and scaled by machines. Yet the economics of automation, especially at the apex of organizations, is subtler than this intuition suggests.
From a cost perspective, the argument is superficially compelling. CEO compensation in large firms routinely reaches tens of millions of dollars annually. In standard production theory, such a factor would invite substitution by capital if capital could deliver equivalent output at lower marginal cost.
AI systems already outperform humans in domains once thought irreducibly managerial: forecasting demand, optimizing supply chains, monitoring performance, and allocating capital under uncertainty. These are core components of what modern CEOs actually do, particularly in complex, data-rich corporations. In this narrow sense, executive work is increasingly legible to machines.
However, CEOs are not merely information processors. Economically, they function as coordination devices and as repositories of authority. Firms exist because markets are costly; CEOs reduce internal transaction costs by making binding decisions where consensus or contracting would be slow or impossible.
While AI can recommend actions, the ability to impose a final choice (and to be held accountable for it) remains socially and legally human. Automation struggles not with optimization, but with legitimacy. Shareholders, regulators, courts, and employees still require a person to blame or praise.
That said, the structure of executive labor is already changing in ways consistent with partial automation. Decision-support systems increasingly compress the informational advantage of top executives. If everyone in the organization can query the same AI-driven dashboards, the CEO’s role as “the smartest person in the room” diminishes.
What remains is meta-decision-making: choosing objectives, resolving conflicts between stakeholders, and interpreting ambiguous social signals. These tasks are less about computation and more about narrative, persuasion, and moral judgment, areas where AI remains brittle.
There is also a political economy dimension. CEO pay is not set in a competitive labor market alone; it is shaped by norms, bargaining power, and imperfect governance.
Even if AI could replicate much of a CEO’s functional output, incumbents may resist displacement, just as other elites have historically done when faced with labor-saving technologies. Automation is rarely a purely technical choice; it is mediated by power.
The most plausible outcome is not robot-induced redundancy at the top, but robot-induced augmentation. AI will hollow out the analytical core of executive work, allowing a smaller, differently skilled leadership layer to oversee larger and more complex organizations.
Paradoxically, this could entrench high CEO pay rather than eliminate it, by making the remaining human functions (symbolic leadership, crisis authority, and accountability) even more concentrated.
My Inference
CEOs are expensive, and parts of their job are eminently automatable. But the economic function of a CEO is not only to decide efficiently; it is to embody responsibility in systems that cannot yet ascribe it to machines. Until that changes, automation will transform the role, not render it redundant.
- Jishnu Chatterjee,
Fri, January 16, 2026.
