Abstract:
Purpose: This article examines the theoretical implications of organisational decisions being increasingly prepared, filtered, produced through or attributed to computational programmes rather than individual human actors. It asks under what conditions interactions between such programmes may be analysed as communicatively consequential and what this implies for management and organisation theory.
Design/methodology/approach: The article develops a conceptual analysis grounded in Luhmannian social systems theory. It uses a deliberately simplified, user-moderated exchange between two large language models, ChatGPT and Gemini, as an illustrative vignette. Their interpretations of code poems serve as a heuristic device for examining interpretive divergence, role formation, and double contingency.
Findings: The illustrative exchange suggests that mutual responsiveness between computational programmes need not produce interpretive convergence. Instead, recursive reference to one another’s outputs may stabilise distinct observational positions and patterned forms of mutual irritation. The article does not treat the vignette as an empirical test of programme–programme communication. It uses it to develop the conceptual proposition that large language models may be observed as organisationally instantiated decision programmes whose outputs become communicatively consequential when they are recursively incorporated into further organisational decisions.
Originality: The article contributes to management and organisation theory by reframing artificial intelligence not as a tool or agent, but as a decision programme operating within organisational autopoiesis. By introducing programme–programme communication as an analytical lens, it shows how generative information technologies may create new decision premises, stabilise differentiated roles, and reshape organisational coordination, responsibility, and control in programme-rich environments.
Keywords: Programme-programme communication; organisational autopoiesis; double contingency; artificial intelligence; code poems.
The Author Accepted Manuscript version of this article is available for download here.
Recommended citation: Roth S. and Lien V. (in press), When Programmes Disagree About Programmes: Large Language Models, Code Poetry, and the Problem of Programme–Programme Communication, Information Technology & People, DOI: 10.1108/ITP-02-2026-0241.
