Concepts and theories to foster automation in organisations can be regarded as an essential research goal of the information systems discipline. Based on a few assumptions about essential inhibitors of successful automation, this paper describes and analyzes three prototypical approaches to accomplish a higher level of automation in organisations: a reductionist, a inductivist, and a constructivist approach. While there is evidence that the constructivist approach is most effective, it will be shown that it implies severe epistemological problems. It will be argued that there is hardly any straightforward solution to these problems. However, for the information systems discipline to establish a profile of its own, it has to accept them as a pivotal challenge.