When Artificial Intelligence Does Strategy: Learning, Good Times, Lock-in, and Human-Driven Strategic Renewal.

This article investigates the dynamics of industries where firms fully delegate strategic decision making to private artificial intelligence (AI) agents operating under current technological paradigms. It develops a formal model showing that such AI agents, equipped with large catalogs of business models and consequence predictions, converge to a self-confirming equilibrium (SCE) in which their strategy choices are subjectively optimal and well-calibrated along realized paths, though objectively superior but unexplored business models may remain neglected. The study highlights that this equilibrium can sustain high profits but leads to strategic lock-in, with novel business-model innovations becoming rare without human intervention. Humans retain a critical role by intentionally expanding the AI agents' awareness frames—introducing genuinely new business models or consequence categories—which can disrupt the AI-induced equilibrium and trigger strategic renewal; however, when the status quo is sufficiently profitable, managers may rationally choose not to pursue such frame-expanding innovations. The findings emphasize a division of labor where AI excels at within-frame optimization and learning, while humans are pivotal in recognizing and acting upon the incompleteness of strategic frames to foster innovation.