As AI systems increasingly promise speed, scale, and frictionless creation, artists and researchers are left to ask whether “faster” is always better—and what is lost when our tools minimize effort to the point of erasing agency. This talk reframes AI not as a shortcut but as a site of embodied practice, where effort, gesture, and situated knowledge become sources of meaning. Building on recent work on small-data, interactive machine learning for music-making and a broader agenda that repositions datasets, mappings, and models as cultural materials rather than neutral substrates, I’ll argue for an alternative trajectory for creative AI: one that bends tools toward human objectives instead of asking humans to adapt to opaque, high-speed, effortless systems. Through examples from research-creation, performance systems, and pedagogical practice, the talk outlines practical strategies for re-centring humans in AI-mediated creative work, presenting a vision of digital skills grounded not in acceleration but in attention, intention, and the ability to steer the machine.