An algorithm rarely changes the world by itself. It changes the allocation of scarce resources: attention, credit, recommendations, search results, labour, risk capacity or managerial time.
Prediction is only the beginning
The interesting economic question is not simply whether a model predicts well. It is what happens after the prediction. A risk score changes who gets reviewed. A recommendation system changes what millions of people see. A fraud model changes which transactions receive scrutiny.
The economic power of an algorithm comes from the decision that follows its prediction.
Technical performance cannot therefore be the only measure of success. A model can be statistically impressive while producing outcomes that are strategically, economically or institutionally poor.
Algorithms create institutional leverage
When an organisation embeds a model into a workflow, it effectively gives the model a piece of institutional authority. That authority may be tiny in one transaction and enormous in aggregate.
The challenge for leaders is to understand algorithms not merely as software but as mechanisms of allocation. Who benefits? Who bears the error? What behaviour does the system encourage? Which outcomes become harder to see?