Coined by the economist Ronald Coase in 1937, transaction costs describe the friction involved in conducting any exchange through a market rather than within an organisation. They include the cost of finding a willing trading partner, negotiating the terms, and enforcing the agreement once it is made. Coase's insight was simple and profound: when these costs are high, it becomes cheaper to bring the exchange inside a firm — to hire someone rather than contract them, to own a factory rather than rent time in one.
For nearly a century, this logic shaped how organisations were built. Large companies grew because scale reduced per-unit transaction costs — the bigger the firm, the more activity it could internalise efficiently. The result was hierarchy: layers of management whose job was to coordinate activity cheaply enough to justify the payroll. The org chart, in Coase's terms, is a machine for reducing transaction costs.
AI disrupts this logic at its foundation. When coordination can be automated — when sensing, interpreting, deciding, and communicating can be handled by intelligent agents at near-zero marginal cost — the advantage of bringing everything inside one large firm shrinks rapidly. The transaction costs that once justified enormous organisational structures are collapsing, and with them the structures that were built to avoid them.
The human translation is this: every bureaucratic process in a large company exists, somewhere in its ancestry, as a solution to a transaction cost problem. Understanding that helps explain both why those processes are so hard to change — they were once genuinely efficient — and why AI makes so many of them suddenly optional.
