Artificial intelligence and organizational learning in universities: decision premises, paradoxes, and institutional stability

Purpose – The purpose of this paper is to identify and explain the organizational conditions under which artificial intelligence adoption in universities leads to structural change rather than incremental adaptation. By integrating Luhmann’s theory of decision premises with Argyris and Schön’s concept of organizational learning loops, the study conceptualizes artificial intelligence (AI) adoption as a process mediated by institutional structures and mechanisms of invisibilization and proposes strategies to foster double-loop learning that enable universities to surface and address organizational paradoxes, thereby creating the conditions for meaningful transformation in teaching, research and governance. Design/methodology/approach – This conceptual study develops an analytical framework combining Luhmann’s theory of decision premises (programs, communication channels and personnel) with Argyris and Schön’s distinction between single-loop and double-loop learning to examine how universities process AI adoption. The approach synthesizes literature from organizational sociology, higher education studies and paradox theory to explain how contradictions are mediated by institutional structures and managed through mechanisms of invisibilization. The framework is applied analytically to the context of AI in teaching, research and governance, identifying conditions under which contradictions escalate into paradoxes that destabilize decision premises and create opportunities for structural change. Findings – The study shows that universities often integrate AI within existing decision premises, containing contradictions through mechanisms of invisibilization, reframing contradictions as technical adjustments, recasting innovation as continuity and suspending role redefinitions, sustaining single-loop learning and organizational stability. Structural change through double-loop learning occurs when external pressures, such as regulatory mandates and funding constraints, converge with internal tensions in academic culture, governance and faculty roles, escalating contradictions into paradoxes that destabilize decision premises. The analysis posits that transformation depends on reconfiguring program premises toward reflexivity, redesigning communication channels for deliberative governance and redefining personnel premises to integrate AI-related expertise into formal authority structures. Research limitations/implications – As a conceptual analysis, the study does not include empirical testing, which limits the ability to generalize findings across institutional contexts. Future research should apply and refine the proposed framework through comparative and longitudinal studies of AI adoption in universities, examining variations across governance models, regulatory environments and disciplinary cultures. The framework offers a basis for analyzing how decision premises mediate technological change, highlighting the need for research that investigates the interaction between external pressures, internal tensions and invisibilization mechanisms. Such work can inform both theory development in organizational change and the design of policies that foster reflexive, transformative AI integration. Practical implications – The framework offers university leaders and policymakers strategies to foster transformative AI adoption by making organizational contradictions visible and actionable. Institutions can reconfigure program premises to align AI initiatives with mission and values, redesign communication channels to integrate AI within participatory governance and redefine personnel premises to incorporate AI-related expertise into formal authority structures. These interventions can help balance efficiency gains with academic autonomy, transparency and epistemic diversity. Policymakers can use the framework to design regulatory and funding mechanisms that incentivize reflexive adaptation rather than superficial compliance, thereby creating conditions for sustainable organizational change in teaching, research and governance. Social implications – By framing AI adoption in universities as an organizational learning challenge, the study highlights its potential societal impact beyond technical efficiency. Universities play a central role in shaping knowledge production, professional formation, and public trust in expertise. AI integration that prioritizes reflexivity, inclusivity and participatory governance can strengthen these societal functions, fostering equitable access to high-quality education and preserving epistemic diversity. Conversely, uncritical adoption risks reinforcing managerial logics that marginalize academic voices and narrow the social purposes of higher education. The framework encourages institutions to engage with AI in ways that support democratic accountability and socially responsive knowledge systems. Originality/value – This paper offers a novel conceptual framework linking Luhmann’s theory of decision premises with Argyris and Schön’s organizational learning loops to explain how AI adoption in universities is mediated by institutional structures. By introducing the concept of invisibilization mechanisms, reframing contradictions as technical adjustments, recasting innovation as continuity and suspending role redefinitions, the study advances understanding of why AI often reinforces stability rather than triggering structural change. It also extends organizational change theory in higher education by specifying conditions under which contradictions escalate into paradoxes and by proposing targeted strategies to foster double-loop learning that enable transformative, reflexive integration of AI technologies.