A bogeyman lurks around the world: the fear that intelligent technological beings will substitute human beings in (almost) all their activities, starting from work. In this sense, AI is often perceived as a magical helper, a high technology (maybe the highest one) potentially capable of performing any task. Guided by Gilbert Simondon's takes on technical alienation, this paper argues that automation is actually the lowest form of human-machine interaction. Recalling some milestones in the history of AI, it makes some suggestions on why, regardless of their human design, the so-called 'intelligent' digital machines based on data reduction algorithms, tends to incorporate ideological assumptions, beliefs and values. And by analysing human interactions with search engines, it shows that automated systems implies human-machine mutual conditioning: the more humans address these machines in a mechanically, simple and unambiguous way, the more the latter act 'automatically' and appear 'intelligent'. Drawing on this analysis this paper finally discusses the exploitation of that mutual conditioning mechanics by a myriad of for-profit digital micro-activities. "Trained" machines, in this context, are far away from the ideal of the Simondonian "open machine".