Intelligent technologies are increasingly used to evaluate and monitor performance through implementing specific metrics such as Key Performance Indicators (KPIs). Recently, scholars have called to better engage with intelligent technologies' specific features, which otherwise remain black-boxed. We aim at developing greater insights into how practices enacted by different end-user’s shape and reshape the production of KPIs in technological change initiatives. Combined with a ‘performative’ process approach, our research draws on Michel Foucault’s view of power to emphasise both the negative and productive effects of power. We highlight the pursuing (push)-withdrawing (pull) dynamics which lie in iterative and recursive acts of power and resistance involving a range of actors who collectively, albeit inadvertently and differently, change the meanings technological outcomes perform. Finally, we show that technologies that might initially seem to fail can ‘take off’ through persistent and positive power effects.