The conditional turn: algorithmic transformation of organizational decision programmes

Purpose This article theorizes how algorithmic decision systems transform organizational decision programmes from purposive (goal-oriented) toward conditional (rule-based) structures. While recent management research on artificial intelligence and strategic decision-making has advanced our understanding of human–AI combination at the task level, the organizational-level structural dynamics through which sustained algorithmic mediation reshapes decision premises remain untheorized. Drawing on Luhmann's social systems theory, we address this gap. Design/methodology/approach This conceptual article develops process theory grounded in Luhmann's concepts of decision programmes, operational closure and structural coupling. We theorize algorithmic implementation drives directional transformation through three stages: algorithmic augmentation introducing specificity bias, conditional programme proliferation through recursive formalization and purposive programme atrophy through operational displacement. The framework engages the complementarity thesis and identifies conditions enabling purposive resilience. Findings Three asymmetries drive directional drift: codifiability, measurability and scalability. Specificity bias operates differentially across economically oriented and multifunctionally oriented purposive programmes, colonizing the former first. The complementarity thesis, that algorithms free human judgment for strategic decisions, describes a possible but unstable equilibrium; attention capture, competence decline and legitimacy erosion systematically undermine its conditions, threatening organizational capacity for multifunctional coupling. Originality/value The article extends Luhmann's programme theory from static typology to dynamic transformation theory and introduces programme drift as an emergent phenomenon distinct from planned change. We distinguish purposive closure from conditional closure, revealing novel heteronomy under algorithmic conditions. By delimiting the complementarity thesis, the framework identifies the structural conditions under which human-AI complementarity can be sustained rather than presumed.