In the museum sector, interest in visitor engagement methods has been growing for several years. Among these, AI and machine learning have proven to be valid engagement and storytelling tools for museums, in particular for the creation of interactive chatbots and better customize visitor experience. This paper aims to reason about the possibilities of these existing tools, discussing their potential and limitations. The collection of visitor flow data within the museum and the interactive use of chatbots allow different levels of personalization of the visit. The real question is: how can an evolving AI tool be transformed into a storytelling tool that is in harmony with the museum's itinerary and allows for personalization but also respects the museum's own identity and peculiarities? These reflections are part of a research proposal for developing a chatbot to orient visitors within the museum, signaling works of potential interest. The application context is the National Museum of Ravenna, characterised by a vast and heterogeneous collection that is difficult to use.