Make-or-Buy Decision in the AI Age

In the AI Story

Oliver Williamson's extension of Coase formalized asset specificity as the primary driver of vertical integration. When assets required for a transaction — skills, knowledge, relationships, physical equipment — are specific to a particular firm, market transactions become risky because the party investing in firm-specific assets is vulnerable to opportunistic renegotiation. Firms internalize to protect against this vulnerability, accepting coordination costs as insurance against market opportunism. A law firm understanding its client's business is more valuable than outside counsel, no matter how technically skilled, because the firm has accumulated the context making advice applicable. A software company with a proprietary codebase cannot easily outsource maintenance because any external developer learning it acquires bargaining power disproportionate to market rates.

AI reduces asset specificity's economic significance by democratizing context. An AI-equipped contractor given access to documentation and codebase can produce work previously requiring months of onboarding. The knowledge residing in senior engineers' heads — accumulated through years working within the firm — is increasingly externalized into documentation, codebases, and data that AI can process and present to anyone. The firm-specific human capital Williamson identified as vertical integration's driver is being commoditized by the same tool reducing production costs. Activities firms internalized because of knowledge-specificity requirements are migrating to contractors, freelancers, and AI-augmented individuals who produce comparable output without permanent-employment overhead.

The third option — make-yourself-with-AI — eliminates both coordination and transaction costs simultaneously. The architect who previously chose between hiring an in-house development team (make) or contracting with a vendor (buy) can now describe what she wants to an AI and produce it herself. This option works best for well-specified problems, prototypes, and domains where speed matters more than perfection. But the boundary of "well-specified" expands as AI capabilities improve and prompting strategies mature. Problems requiring specialized judgment last year are well-specified this year. The make-yourself option is not static; it is growing into territory the make-or-buy decision previously allocated firmly to make-within-firm.

Origin

The make-or-buy framing originated in operations management and became the canonical application of Coasian logic to practical organizational decisions. Williamson's 1975 Markets and Hierarchies developed the analytical framework identifying when vertical integration is efficient. The AI-era restructuring of the decision emerged through 2024–2026 as practitioners discovered that activities previously requiring teams could be performed by AI-augmented individuals. Edo Segal's Napster Station thirty-day sprint illustrated the shift: a product requiring industrial design, optics, audio routing, conversational AI, and software integration was built with radically reduced headcount because transaction costs between domains had been compressed by AI holding full project context.

Key Ideas

Firm-specific knowledge commoditized. AI's ability to absorb organizational context in hours eliminates the knowledge barrier that made external contractors expensive relative to internal employees.

The third option rises. Make-yourself-with-AI grows from niche to viable for increasing categories of work, eliminating both firm coordination costs and market transaction costs.

Ecosystem layer holds value. Firms whose make-or-buy advantage is in production (code-writing) lose rationale; firms whose advantage is in ecosystems (data, integrations, trust) retain it even as production externalizes.

Continuous reassessment required. The boundary's rapid movement means make-or-buy decisions must be revisited at every cycle — what justified internal organization last quarter may not this quarter.