Research on artificial intelligence (AI) in organisations has largely focused on ethical risks and decision outcomes. Here, AI denotes organisationally embedded predictive, recommendation and generative systems that supply decision-relevant outputs within workflows. This paper shifts attention to the less visible but more consequential structures that condition organisational decision-making: programmes. The paper examines how AI reconfigures programmes as decision premises in organisations by drawing on social systems theory.
The paper develops a conceptual analysis grounded in Niklas Luhmann's theory of social systems. It revisits the concept of programmes as conditional decision premises and traces how these structures are operationalised through AI-mediated technologies.
The analysis suggests that AI does not replace organisational decision-making. It reorganises the decision premises that make selections defensible in communication. AI can translate conditional programmes into executable code and adaptive models. This translation can intensify programmatic control, and it can also reallocate observability. Outputs and performance indicators may become more visible, and decision premises may become less visible. They often sit in training pipelines, model architectures and threshold settings. These premises can be harder to render communicable and contestable (Burrell, 2016; Pasquale, 2015). Inclusion and exclusion then appear less as design failures and more as systemic effects of how relevance criteria are stabilised in organisational programmes (Introna, 2016; Noble, 2018).
The paper contributes to IS research by reintroducing programmes as a central analytical concept for understanding AI-mediated organisations. It offers a theoretical vocabulary that connects digital infrastructures to classical questions of organisational inclusion and control. It extends current debates on algorithmic governance beyond outcome-focused perspectives. It also extends Luhmannian programme theory, and specifies how AI turns decision premises into versioned and updateable artefacts. These artefacts reshape organisational memory over time and also reshape the communicability of justification.
