Organizational sustainable artificial intelligence capabilities scale development, validation, and implications

As artificial intelligence (AI) adoption accelerates globally, its sustainability implications remain insufficiently integrated into organizational capability frameworks. This study develops and validates the organizational sustainable AI capabilities (OSAIC) construct, extending dynamic capabilities theory by embedding sustainability as a meta-capability in AI governance and innovation processes. OSAIC is conceptualized as a five-dimensional, reflective, higher-order construct, encompassing sustainable AI learning, seizing, sensing, stakeholder integration, and transformation. A multi-phase scale development procedure—including expert Q-sorting, exploratory factor analysis, and confirmatory factor analysis, using partial least squares structural equation modeling—was employed. The scale was assessed and validated using two distinct samples: a pilot study ( n = 188) and a main study ( n = 364), both comprising managers from diverse industries and regions. The findings indicated robust psychometric attributes, characterized by substantial reliability, convergent, discriminant, and predictive validity. A positive and significant relationship between OSAIC and sustainable innovation indicated nomological validity, addressing the AI sustainability paradox by illustrating that sustainability-oriented AI capabilities enhance rather than constrain innovation. By extending the research on dynamic capabilities and paradoxes and presenting a validated measurement tool, this study contributes theoretically and methodologically, respectively, to the literature. Practically, it offers managers a diagnostic framework to align AI implementation with environmental and social accountability while fostering innovation.