The Learning Power of Firms: A New Dimension of Competitive Advantage in the Age of AI

The organizational learning literature has long recognized the strategic value of learning, but has conceptualized it primarily as a human-centered, episodic, and reflective capability. The proliferation of artificial intelligence fundamentally alters these assumptions: learning is increasingly embedded in operational systems, continuous in tempo, and distributed across the full surface of organizational activity rather than concentrated in designated learning units. This paper introduces the concept of learning power—a composite organizational capability constituted by four interdependent dimensions: learning density (the proportion of organizational interactions that generate usable insight), learning velocity (the speed with which insight is translated into changed system behavior), learning scale (the number of simultaneous learning processes the organization can sustain), and learning directionality (the alignment of learning objectives with strategic intent and organizational values). The authors propose that competitive advantage in AI-enabled environments is increasingly determined by the multiplicative combination of these four dimensions—Learning Power = Density × Velocity × Scale × Directionality—rather than by any single component. The authors further distinguish the concept of the learning surface—the distributed organizational architecture through which AI-enabled learning operates—from the learning units and episodic learning processes that characterize conventional accounts. The paper situates the learning power framework within the context of companion contributions on AI competitive architecture (Dutta & Doz, INSEAD Working Paper, 2026/38/STR), programmable strategy (Doz & Dutta, INSEAD Working Paper, 2026/39/STR), and strategic envelope governance (Dutta & Doz, INSEAD Working Paper, 2026/40/STR), and draws out implications for organizational learning theory, dynamic capabilities, and the human governance of machine learning systems.