When Standard Is Not Enough: A Conceptualization of AI Systems' Customization and its Antecedents

The centrality of information systems (IS) customization to match companies' needs with software systems available in the market has been researched extensively. The distinctive characteristics of Artificial Intelligence (AI) systems compared to other types of IS suggest that customization needs a new conceptualization in this context. We draw on evidence from expert interviews to conceptualize customization of AI systems as composed of four layers: data, models, algorithms, infrastructures. We identify a continuum of levels of customization, from no to complete customization. Since companies customize AI systems in response to business needs, we develop a theoretical model with six antecedents of AI systems' customization choices. In so doing, we contribute to both AI management research, by introducing the IS customization perspective in the field, and IS customization literature, by introducing AI systems as a novel class of systems and enlarging the understanding of customization for a specific class of software systems.