Generative AI Under Uncertainty: Rethinking Managerial Decision Quality in Strategic Business Environments

Abstract

The arsenal of generative artificial intelligence (GenAI) is rapidly advancing to become a decision support tool for executives with high-stakes, complex and irreversible decisions. The dominant scholarly work on the explanations of managerial decision quality under uncertainty concerns are based on a cognitive architecture that is purely human, and the relatively new body of empirical research on GenAI applications in organizations has been largely insulated from the classic theoretical approaches. This paper aims to bridge that gap by creating an integrative conceptual framework that connects the unique characteristics of generative AI (probabilistic and nondeterministic output, scale, speed, and epistemic opacity) with the existing literature on the mechanisms by which managerial decision quality is generated and undermined when confronted with genuine uncertainty. From a Knightian perspective, bounded rationality, upper echelons theory, and the emerging research on algorithm aversion and appreciation, we suggest that GenAI not only alleviates informational uncertainty for strategic decision makers but also creates another kind of uncertainty, one that is opaque and un-reproducible because the reasoning processes of GenAI are essentially black boxes. We advance five theoretical propositions that indicate conditions under which GenAI involvement has the potential to improve or impair the quality of strategic decisions and provide an empirical research agenda to test these propositions. The paper offers a new theoretical lexicon for a rapidly changing practitioner reality which has, until now, outrun management theory.