Are the Futures Computable? Knightian Uncertainty and Artificial Intelligence

The growing sophistication of artificial intelligence (AI) tools in entrepreneurship is transforming how new ventures identify, gather, analyze, and utilize information from their internal and external operating environments to automate critical choices, decisions, and tasks. For many startups and corporate ventures, prior research suggests that AI provides significant task performance advantages to entrepreneurs in addressing the problem of uncertainty, in part, through enhanced predictive capabilities. What is less clear, however, is whether AI tools enable entrepreneurs to manage the problems of "Knightian uncertainty"—a fundamental type of uncertainty that manifests in entrepreneurship through a cascading set of four interrelated problems: actor ignorance, practical indeterminism, agentic novelty, and competitive recursion. In this study, we argue that the predictive capabilities and task performance advantages of AI are contingent upon the ability of these systems to grapple with the problems of Knightian uncertainty. We investigate the logic of this approach through an in-depth analysis of the limits of foundational and emerging types of AI to address these problems, identifying fundamental areas of computational irreducibility where the manifestation of these problems limits the use of AI in entrepreneurship.