PurposeThe purpose of this paper is to better understand management accounting automation by exploring the programmability of management accounting work.Design/methodology/approachWe build upon the literature on digitalization in management accounting and draw upon the pragmatic constructivist methodology to understand how digitalization takes place at the individual actors' level in accounting practice. The paper uses a data set from an interventionist case study of a machinery manufacturer.FindingsWe examine an actual process of automating management accounting tasks. During this development process, surprisingly, calculation tasks remained more fit for humans than machines though, initially, they were thought to be programmable.Research limitations/implicationsAccording to our findings, practitioners may interpret experts' nonprogrammable work tasks as programmable and seek to automate them. Only identifying the factual possibilities for automating accounting-related work can lead to automation-improved efficiency. Our findings can be increasingly relevant for advanced analytics initiatives and applications within management accounting (e.g. robotic process automation, big data, machine learning and artificial intelligence).Practical implicationsPractitioners need to carefully analyze the entity they wish to automate and understand the factual possibilities of using and maintaining the planned automatic system throughout its life cycle.Originality/valueThe paper shows that when processes are assessed from a distance, the nonprogrammable management accounting tasks and expertise can become misinterpreted as programmable, and the goal of automating them has little chance of success. It also shows possibilities for human accountants to remain relevant in comparison to machines and paves the way for further studies on advanced decision technologies in management accounting.