Building the
autonomous company.

Something strange is happening in the American economy. The part building artificial intelligence is already responsible for a disproportionate share of its growth, while most of the economy AI is supposed to transform still becomes more productive the old-fashioned way, if at all: by adding people, accumulating experience and asking those people to make better decisions. We are spending hundreds of billions of dollars building intelligence while trillions of dollars of existing economic activity remain constrained by the fact that the knowledge required to run them still lives in human beings.

Economics has long treated that constraint as fundamental. Hayek gave its canonical account in The Use of Knowledge in Society. The economic problem, he argued, is how to use knowledge that exists only dispersed across many minds. General rules can sit with experts, but knowledge of the particular circumstances of time and place sits with individuals throughout the economy and can only be used by leaving decisions responsive to whoever holds it. Economic problems arise from change; change is constant and local; and the knowledge required to meet it cannot first be turned into statistics and handed to a center without losing precisely what made it useful. By 1945, the socialist-calculation debate had largely conceded that calculation required prices; Hayek’s deeper claim was that no center could ever possess the dispersed knowledge those prices coordinate. As Whitehead said, civilization advances by extending the number of important operations we can perform without thinking about them. A business likewise runs on thousands of such judgments: decisions whose rules their holders may never have articulated and which therefore could not simply be detached from the circumstances in which they were exercised and made available to a central decision-maker.

But this is where a structural fact may have concealed a contingent one. It is structurally true that knowledge originates dispersed, locally and often tacitly. It does not follow that its exercise must remain inaccessible to anyone but its original holder. Hayek wrote in a world in which observing judgment at scale meant asking someone to explain it, reducing it to a rule, or converting its circumstances into statistics. Today, an increasing share of economic activity occurs inside systems that record the circumstances in which a decision was made, the action that followed and the outcome it produced. A model need not be told the rule if it can learn the judgment from the trace of its exercise.

The entire economy may remain too open-ended to make fully knowable. A business is different. It is a bounded world, with defined systems, permissions, tools, histories, objectives and feedback. Within those confines, Hayek’s knowledge problem begins to look less like an impossibility theorem and more like a research problem: how completely can we observe the world in which an institution acts, infer the judgment encoded in its behavior, and recover what it is actually optimizing from the consequences of those decisions?

The stakes are larger than automation. The modern economy is still organized around the assumption that productive knowledge lives in people and cannot become capital. If that assumption is breaking, then the next great productivity regime may come from converting the accumulated judgment of institutions into software that can compound. Advanced economies cannot grow indefinitely by adding more people and more hours; they have to make existing knowledge more productive. The companies and countries that learn to turn human judgment into machine capital will be able to increase output without increasing labor in proportion. Those that do not will become older, more expensive and less productive while those that do pull away. The opportunity, therefore, is not merely to automate businesses. It is to own the institutions in which economically valuable knowledge has already accumulated, make that knowledge learnable, and convert it into the productive capital of the next economic era.

If working on problems like this at scale interests you, come join us (opens in a new tab).