This thesis explores the integration of embodied interactions within Human-to-Artificial Intelligence (AI) collaborative activity to support creative engagement and reduce the perception of AI as an uncontrollable, autonomous entity, also known as the AI “black box” rhetoric. Specifically, it investigates the utilization of kinetic sand as a sensory material and physiological data collector, facilitating the translation of users' hand motions and imprints in the sand into inputs for generative AI image creation. Informed by sociocultural frameworks of creativity, theories of embodied cognition and the positioning of AI as a statistical model, while grounded in iterative design methodologies and phenomenological analysis, the research aims to identify emergent guidelines from this collaborative creative process between humans and AI. The findings hope to contribute to the development of guidelines that inform the future design and implementation of generative AI systems for creative work. These guidelines account for embodied cognition as an essential facet of human creativity, promoting more intuitive and meaningful interactions between humans and generative AI. Ultimately, this research seeks to advance the discourse on human-AI collaboration, emphasizing the importance of embodied techniques in fostering creative synergy and mitigating the black box effect.