The Coasian Boundary

In the AI Story

The boundary's position reflects the comparison of two kinds of costs. Coordination costs include management overhead, meetings, specification documents, performance reviews, and all the apparatus firms build to direct and monitor internal activities. Transaction costs include search (finding the right provider), bargaining (negotiating terms), and enforcement (ensuring quality). When coordination is cheaper, the activity belongs inside the firm. When transaction is cheaper, it belongs on the market. The Industrial Revolution expanded firms because mechanized production required coordination at scales that markets could not provide. The AI revolution is contracting firms because coordination requirements have collapsed for a wide class of activities.

The boundary's movement is observable in real time through the Software Death Cross — the trillion-dollar repricing of software companies in early 2026. The market recognized that code-production activities, which had justified large engineering organizations, could now be performed by dramatically smaller teams or even by individuals. Firms whose value proposition was "we write the code" discovered their Coasian boundary had contracted beneath them. Only firms whose value resided in ecosystems — accumulated data, customer relationships, regulatory compliance, institutional trust — retained their economic rationale. The repricing was crude but directionally correct: production has migrated across the boundary; what remains inside must be coordination, judgment, or social functions markets cannot replicate.

Three forces determine where the boundary settles in the AI age. First, the magnitude of transaction-cost reduction varies by domain — spectacular for well-specified technical work, modest for ambiguous judgment-intensive work. Second, firm-specific knowledge matters less when AI can absorb codebases and documentation, reducing the asset specificity that Oliver Williamson identified as the primary driver of vertical integration. Third, new coordination costs emerge: evaluating AI output, maintaining architectural coherence across AI-generated components, ensuring quality when production speed outpaces human review capacity. The net effect is a boundary that has moved significantly inward but has not collapsed to zero — individuals can produce more, but complex coordination still requires organizational structure.

Origin

Coase articulated the boundary concept implicitly in his 1937 paper "The Nature of the Firm," though he did not use the term. He observed that "a firm will tend to expand until the costs of organising an extra transaction within the firm become equal to the costs of carrying out the same transaction by means of an exchange on the open market." The equilibrium language made the boundary a calculable threshold rather than a categorical distinction. Oliver Williamson formalized the framework in the 1970s and 1980s, developing transaction cost economics into a rigorous apparatus for predicting organizational boundaries. The "Coasian boundary" as a named concept emerged from this literature as shorthand for the make-or-buy decision understood through transaction-cost logic.

Key Ideas

Equilibrium, not essence. The firm's boundary is determined by cost comparison, not by inherent logic about what activities "belong" inside organizations versus markets.

The boundary moves. Every shift in transaction costs — telephone, email, internet, AI — repositions the equilibrium, and organizational forms must adapt or become inefficient.

Individual as limiting case. The AI-augmented solo builder who needs no coordination represents the Coasian firm's limiting case: a firm of one, where coordination costs are zero.

Production migrates, judgment remains. The boundary is contracting around production activities while holding or expanding around judgment, trust-building, and standard-maintenance functions markets cannot provide.