RESPONSIBLE ARTIFICIAL INTELLIGENCE AND ORGANIZATIONAL PERFORMANCE: THE MEDIATING ROLE OF AI GOVERNANCE CAPABILITY AND THE MODERATING ROLE OF ETHICAL LEADERSHIP

Abstract

The rapid diffusion of artificial intelligence (AI) into core organizational processes has outpaced firms' capacity to govern it responsibly, generating a widening gap between AI adoption and AI accountability. While an emerging stream of research links responsible AI (RAI) practices to firm-level outcomes, the mechanisms through which ethical AI principles translate into measurable organizational performance remain theoretically underspecified and empirically fragmented. Drawing on dynamic capabilities theory and upper echelons theory, this paper develops an integrative model in which AI governance capability — a firm's socio-technical capacity to sense AI-related risk, embed fairness, transparency, and accountability standards into AI systems, and reconfigure decision rights accordingly — mediates the relationship between responsible AI orientation and organizational performance. We further theorize that ethical leadership moderates this indirect effect, strengthening the translation of responsible AI principles into governance capability and organizational outcomes by legitimizing ethical discourse, reducing employee resistance, and aligning managerial incentives with responsible conduct. The paper synthesizes and critically extends recent empirical work (2021–2026) on responsible AI governance, AI-enabled dynamic capabilities, and ethical leadership in technology-intensive contexts, identifying a persistent gap: existing studies examine RAI's performance effects largely as direct or singly-mediated relationships, without accounting for the boundary condition that leadership ethicality imposes on governance institutionalization. We propose a moderated-mediation model, a multi-source, time-lagged survey design (targeting mid-to-large firms undergoing AI-enabled digital transformation), and a partial least squares structural equation modeling (PLS-SEM) analytical strategy suited to complex, formative-reflective constructs. We articulate directional hypotheses and discuss anticipated patterns of results, their theoretical interpretation against dynamic capabilities and stakeholder theory, and practical implications for AI governance design. The paper contributes a testable, boundary-condition-sensitive model that reframes responsible AI from a compliance cost to a performance-relevant dynamic capability, contingent on leadership ethicality.