This thesis aims to take a pragmatic look at the "algorithmisation" of recommendation and decision-making within organisations. It lies at the intersection of research that explores the use of automation and evaluation technologies in organisations and that which questions the intentions and opacity of algorithms. Starting from the premise that the initial step in algorithmisation is the "datafication" (sometimes massive) of consumption and decision-making activities, we draw upon the concept of "algorithmic governmentality" to examine how designers and users of algorithms harness them to achieve their objectives. To accomplish this, we conducted four studies involving around a hundred actors (initiators, engineers, digital designers, and users) of recommendation algorithms and organisational algorithms. We demonstrate, in particular, that initiators employ algorithms – crafted by engineers and digital designers – not to automate actions but to disengage from any cultural or decision-making prescription. Consequently, it falls upon users to make choices and become engaged. In the case of recommendation, algorithm evaluation is oriented towards consumption, while for decisions support, priority is given to utility and interpretability of results. This conclusion suggests the emergence of a "satisfaction regime," providing an optimal framework for initiators to align objectives related to organisation and use.