This position paper questions the need for systematically embedding AI technologies in interactive computing systems when other programming-based alternatives are available. Indeed, while AI technologies bring critical benefits in some contexts such as patterns recognition or computer vision, they also bring issues that prevent them from more generic use. For instance, their blackbox functioning calls for specific research on its explainability which, instead of solving the problems try to address issues brought both to developers (how does my Machine Learning system works) and users (why do I get this result). In addition, we demonstrate that these technologies are not mature enough yet to address reliability and dependability concerns in safety critical domains where probability of failures must be demonstrated to be in the order of magnitude of 10-E9. The paper thus argues for exploiting validated methods in the area of interactive systems engineering to build interactive systems even though their exhibited behavior seems similar to the ones of AI-engineered systems.