This chapter explores effects of the implementation of artificial intelligence (AI) on the organisation of work in the Danish Customs and Tax Administration in a situation where a machine learning algorithm was expected to make the work of customs officers more efficient. The chapter highlights how experiments with machine learning algorithms may redraw intra-organisational boundaries, responsibilities, as well as divisions and meanings of work. The implementation of the machine learning algorithm, as a particular form of AI, in this case, led attention towards data and IT systems, at the expense of the very work the technology was intended to support. One concern at play here, we argue, is that AI developers seem to have generally abandoned the idea of representing work, making it ever more important that ethnographers do so.