This talk aims to unpack AI's multifaceted conceptualisations. Drawing on emerging empirical evidence from recent studies aligned with each of these conceptualisations, the talk explores AI's potential to enhance productivity for both teachers and learners. At the same time, it raises critical concerns about the challenges of designing AI systems that genuinely augment human competence and foster synergy in tightly coupled human-AI interactions. The presentation underscores the unintended consequences of focusing narrowly on a single conceptualisation in research or practice and argues for moving beyond reductive applications of AI aimed solely at productivity gains.