Abstract
Generative artificial intelligence is becoming embedded in the conversational environments through which organisations coordinate work. Large language models now draft messages, formulate recommendations, and summarise discussions within collaborative platforms where requests, approvals, delegations, and commitments are negotiated. This development exposes a limitation in the informational paradigm that has historically structured Information Systems research. While digital systems have largely been designed to represent, store, and process information, organisational coordination depends fundamentally on communicative acts through which actors establish accountable commitments. When machines generate organisational language, the boundary between linguistic expression and institutionally valid action becomes more difficult to reconstruct from message traces alone. This paper argues that the design problem of information systems therefore shifts from managing information to sustaining accountable communication. In response, it introduces the concept of accountability architectures: lightweight infrastructural mechanisms that preserve the institutional framing of communicative acts by recording the roles invoked by participants, the types of acts performed, and the acknowledgements through which commitments become recognised. The paper further outlines a minimal accountability wrapper as a proof-of-concept interaction layer for conversational platforms. By reframing digital systems as institutional communication infrastructures, the paper offers a new conceptual foundation for understanding digital coordination in environments where generative AI participates in the production of organisational language.
Recommended Citation
Jacucci, Gianni, "preprint OISI26 t-12 inst_AI4 - When Machines Generate Language: Rethinking Information Systems as Infrastructures for Accountable Communication" (2026). OISI Workshop 2026. 16.
https://aisel.aisnet.org/oisiworkshop2026/16