The Make-or-Buy Decision

In the AI Story

Before Williamson, the make-or-buy decision was treated as a strategic choice driven by considerations of core competence, competitive advantage, or managerial preference. Williamson reframed it as an economic optimization: the firm should make whatever it can produce at lower total cost (production cost plus governance cost) than the market, and buy whatever the market can supply more cheaply. The reframing had immediate practical implications for corporate strategy, because it meant the analysis of vertical integration—whether to own suppliers, distribution channels, or complementary capabilities—should focus not on strategic narratives but on transaction cost characteristics. High asset specificity, frequent transactions, and substantial uncertainty predict vertical integration. Low specificity, infrequent transactions, and contractual adequacy predict market procurement. The middle ground produces hybrid governance—long-term contracts, partnerships, relational arrangements combining elements of both.

The AI revolution inverts a make-or-buy decision that has governed software firms for four decades. Since the 1980s, the default was to make software internally: assembling teams of specialized engineers because the coordination costs of specifying requirements to external developers, monitoring quality across organizational boundaries, and integrating components built by parties without shared context exceeded the costs of hierarchical coordination through employment relationships. The Orange Pill's Trivandrum demonstration showed that the calculus has reversed. A single AI-augmented individual can now produce what previously required a coordinated team, and can do so through a buy relationship with the AI platform (subscription fee, terms of service, market-mediated access) rather than through the hierarchical employment relationships that previously governed software production. The transaction costs of internal coordination no longer justify the overhead of the team.

But the inversion is partial, and the partiality determines which firms survive. The make-or-buy decision for execution has shifted toward buy (or buy-equivalent relationships with AI tools). The make-or-buy decision for judgment—the specification of what should be built, the evaluation of whether it serves purpose, the strategic direction of organizational attention—has shifted toward make. Why? Because judgment is transaction-specific: it depends on contextual knowledge (customers, competition, institutional history) that cannot be purchased on the market without losing the specificity that makes it valuable. A firm can buy execution services from Claude. It cannot buy the judgment about what Claude should build, because that judgment is embedded in the organizational relationships, accumulated experience, and strategic context that are specific to the firm. The make-or-buy decision has not been eliminated by AI. It has been bifurcated: buy execution, make judgment, and build hybrid governance structures to manage the interface between them.

Origin

The make-or-buy decision is as old as economic organization—every producer has always faced it. But its systematic analysis as a governance choice begins with Coase's 1937 paper and receives full development in Williamson's framework. The concept is central to strategic management (core competence theory), operations management (supply chain design), and corporate finance (vertical integration decisions). Its application to knowledge work is more recent: the 1990s internet boom produced the first wave of make-or-buy reconsideration for information services, leading to the outsourcing movement. The AI wave is producing a second, more dramatic reconsideration whose outcome will restructure the entire knowledge economy.

Key Ideas

The decision is a cost comparison. Make when internal governance costs less than market transaction costs; buy when the reverse holds—the calculus is economic, not strategic storytelling.

Transaction characteristics predict the decision. High asset specificity, substantial uncertainty, and frequent transactions predict 'make' (internal production); low specificity, contractual adequacy, and infrequent transactions predict 'buy' (market procurement).

AI has inverted the execution decision. Activities once made internally because market coordination was expensive can now be bought (or AI-executed) because the coordination costs have collapsed.

Judgment remains a 'make' activity. The specification, evaluation, and strategic direction of AI output are transaction-specific and must be governed hierarchically because market alternatives lack the context.

Bifurcation reorganizes firms. The same organization now makes judgment (coordinating internally through hierarchy) and buys execution (contracting with AI platforms)—a hybrid structure unprecedented in organizational economics.