Recent advances in machine learning (ML) have triggered many firms to try putting the technology into commercial use. However, the creation of ML-based organizational capabilities remains a major challenge. With the aim of extending our understanding of organizational capabilities, this paper takes a socio-technical system perspective on the microfoundations of capabilities, develops an integrative conceptual framework, and discusses the resulting insights relevant to organizational ML initiatives. In contrast to past IS research, our framework is more general and versatile, since it is not restricted to dynamic capabilities only, as well as incorporates a temporal dimension facilitating the inspection of processes leading to the formation and change of organizational capabilities. This is illustrated with multiple propositions, which we develop by applying the framework to the context of organizational ML initiatives. Conceptual insights are backed with rich anecdotal evidence.
