Abstract
While Generative AI (GenAI) and in particular large-language models (LLM) holds significant promise for transforming knowledge work, there is limited understanding of how organizations implement and adapt LLM systems in practice. We conducted a case study in a large maritime organization, drawing on semi-structured interviews with leaders, developers, and expert users. Our findings show that the materialization of agency in GenAI use is not determined by the technology alone, but emerges from the interplay of organizational norms, leadership discourses, design choices, and workers’ situated practices. While GenAI carries open-ended potential, its affordances were constrained by prescriptive managerial framings, even as workers re-materialized agency through exploratory and relational engagements. We argue that organizational adoption of GenAI should be approached as an ongoing process of experimentation, attentive not only to technical fine-tuning but also to the relations that shape its use.