AI has the potential of revolutionising the services we use in our everyday life, thanks to an increased automation of simple tasks. This could lead to a future in which the interface (or even the service) disappears, replaced by an AI-user accessing specific functions on human behalf. At the same time, AI has the potential to augment services and experiences by shaping them around user needs and anticipating their intentions. Along this continuum from automation to augmentation, how can we design meaningful AI experiences characterized by distributed (or even zero) interfaces? AI experiences could be based on multiple tools or single entities, spot interactions or long-term relationships, generic entities or personalised algorithms: a whole new grammar that designers need to learn in order to envision new service experiences while understanding their potential outcomes.