Event Description
Three of the foundations of modern finance are taught everywhere and applied almost nowhere. Bellman’s dynamic programming, sixty-eight years old, is recognised as the correct way to think about sequential decision under uncertainty, and is almost never used to steer a company. Markowitz’s framework, seventy-four years old, is applied to portfolios of securities and not to portfolios of strategic alternatives. The Modigliani-Miller propositions, sixty-three years old, describe an equilibrium and say nothing about what to do on Monday morning.
This talk argues that what stood in the way was never the mathematics. It was the absence of a state space an enterprise can actually populate, of an admissible set its own constraints define, and of a value function its own figures feed. Real option theory attempted that bridge and failed for a structural reason: in a corporate decision the spot and the strike are not of the same economic nature, and treating them as though they were produces prices no practitioner will act upon.
The second half turns to what follows. Once decisions carry their own basis — every parameter declaring where it comes from, every estimate the band within which it may sit, every conclusion testable against its own inputs being wrong — the same architecture that governs a firm governs an autonomous agent. That is precisely the discipline artificial intelligence lacks when it is asked to decide rather than to describe. It is also why real-time management control no longer requires the weight, the implementation cycle or the cost of a conventional enterprise resource-planning system.
The session will close on the question the audience is best placed to judge: not whether a model is elegant, but what an enterprise stops needing once its decisions carry their own basis.