The Shrinking Firm
26 May 2025 - Hugo O'Connor
Last week, I attended an unconference on "What is Software Engineering in the Age of LLMs?" put on by Web Directions and hosted at the CSIRO Concept Lab. The discussion about organisational change had me thinking; why do companies exist at all?
Ronald Coase's insight was that companies emerge to reduce transaction costs, the friction of finding partners, negotiating terms, and enforcing agreements. But what happens when AI renders these costs not merely negligible, but approaching zero as the answers are only an API call away? Consider the countervailing force: as organisations grow, internal coordination costs rise. Every additional person multiplies the communication pathways; every meeting to create "shared understanding" extracts a cognitive toll. In the worst case, silos form between business units to protect prestige, status and resources.
This tension creates a new inflection point that favors radically smaller organisations. When cloud cognition can coordinate complex activities without the overhead of status meetings, Slack channels, and alignment workshops, the equation that has determined organisational size fundamentally shifts. The optimal firm shrinks. How small is ideal? I don't know, but I would hazard to guess less than ten people. AI provides an asymetric advantage to the small company.
Yet this decentralization trend paradoxically coexists with new concentrations of power. Marketplaces that collect 40% fees. As Yanis Varoufakis observes, we may be witnessing not market perfection but "technofeudalism" digital fiefdoms where platform owners extract rents from every transaction while controlling the algorithmic terrain. "He simply controls the digital system... collecting a ground rent, which I call the cloud rent."
As we transition toward networked economies of smaller entities, what infrastructure might enable truly free exchange without recreating feudal power structures thas seek to erect enclosures to extract rents? How might we distribute the efficiency gains without concentrating control in algorithms we neither govern nor fully comprehend? The answer may lie not in technology alone, but in reimagining digital infrastructure as a commons, building a common wealth.
Go back