Ronald Coase asked in 1937 why firms exist at all if markets are efficient allocation mechanisms. His answer—that firms internalize transactions when coordination through contracts, negotiations, and enforcement costs more than managing the same work under unified direction—became the foundation of organizational economics. The boundary of any firm sits precisely at the point where the marginal cost of adding one more employee equals the marginal cost of purchasing the same output from the market. This boundary explained General Motors' vertical integration, Hollywood's shift from studio contracts to freelance production, and the internet-era wave of outsourcing. For knowledge work, the boundary remained stable because the transaction cost of creation was high: converting a product vision into working software required specialized roles, sequential handoffs, and coordination mechanisms that only unified management could provide efficiently.
When AI eliminates the translation cascade between intention and implementation, the arithmetic changes. Cowen observes that a traditional six-person software team—product manager, designer, frontend engineer, backend engineer, QA engineer, project manager—exists because the coordination overhead is less than the market alternative. But when a three-person 'vector pod' with AI can produce the same output faster, the Coasian logic demands reorganization. The coordination cost of the six-person team—meetings, handoffs, misunderstandings, spec documents losing fidelity—now exceeds the productive benefit. The optimal team shrinks not because people become unnecessary in the abstract, but because the transaction cost that justified their bundling has collapsed.
The historical precedent is the factory electrification studied by Paul David. When electric motors replaced steam engines, initial productivity gains were modest because factories kept steam-era layouts—machines arranged along drive shafts, close to a central power source. Full gains arrived only when designers realized electric motors freed each machine to have its own power, allowing workflow-driven layouts. The redesign took thirty years because the redesign itself was a human-speed process. AI's unique feature is that it accelerates its own organizational integration: companies can use AI to redesign their workflows, identify leverage points, and restructure teams at machine speed rather than committee speed. The thirty-year lag may compress to five or ten years, still slow enough to generate transition pain but fast enough to outpace institutional adaptation.
The policy implications ripple through labor markets, educational institutions, and national competitive strategy. Nations that restructure their organizations around the new Coasian boundary—cultivating judgment-dense teams, eliminating execution-focused bureaucracy, building educational systems that develop direction rather than implementation—will capture disproportionate returns. Nations that cling to the old boundary, defending the large-team coordination structures that the previous transaction cost justified, will watch productivity potential bleed away quarter after quarter. The boundary has moved. The question is who moves with it and who gets stranded on the wrong side.
The concept originates in Coase's 1937 paper 'The Nature of the Firm,' which asked why autonomous contractors don't handle all production if markets allocate resources efficiently. His answer—that transaction costs (search, contracting, enforcement) often exceed the costs of internal management—won him the 1991 Nobel Prize and became the foundation of organizational economics. Oliver Williamson extended the framework in the 1970s and 1980s with transaction cost economics, showing that asset specificity, uncertainty, and frequency determine whether activities are internalized or outsourced. The AI application is Tyler Cowen's synthesis, developed across his 2023 writings and formalized in his forthcoming work, observing that when AI collapses the creation transaction cost, the Coasian boundary shifts inward with a speed unprecedented in economic history.
Transaction costs determine firm size. Organizations grow when internal coordination is cheaper than market contracting; they shrink when the reverse is true—and AI has inverted the equation for knowledge work.
The coordination tax now exceeds the execution value. Six-person teams whose coordination overhead once justified their existence now lose to three-person teams directing AI, because the translation cascade has collapsed.
AI accelerates its own organizational integration. Unlike electricity or the internet, AI can be used to redesign the workflows it disrupts, compressing the thirty-year institutional lag to less than a decade.
Judgment density becomes the optimization target. The new organizational logic prizes small teams with high evaluative capacity over large teams with broad execution bandwidth, because execution is now the abundant input.