The aim of this poster is to engage a conversation about how administrative expertise is enacted and altered when utilizing Machine Learning (ML) technologies. By adapting a theoretical lens inspired by Barley’s work on technology as an occasion for structuring (Barley, 1986), I report from an ethnographic study at the Danish Business Authority (DBA) and their use of ML technologies. Specifically, I present two use cases where ML plays an important role in detecting and preventing financial fraud. In the first use case, ML calculates the probability that a business will commit fraud in the near future. In the second, ML screens financial accounts and flags them if they appear fictitious or flawed. For both cases, ML directs the resources of domain professionals, such as accountants, legal experts, and auditors, towards high-risk cases, but the way ML contributes to the enactment of administrative expertise varies. In the first case, ML pre-empts uncertain futures by predicting crimes before they occur, enabling the DBA to act as a first line of defense by stopping these criminals at the garden gate. In the second case, ML supports professionals in responding to fraudulent behavior to prevent further crime. Consequently, I argue that understanding the domain-specific use of ML technology is essential to determining its impact on expertise.
