As automation accelerates and generative AI reshapes the digital landscape, creativity is emerging as one of the most valuable and uniquely human skills. A recent report identified creativity as the number 1 soft skill in global demand across industries. Yet creativity does not occur in a vacuum; it is intimately shaped by our patterns of attention, emotions, characteristics, way of thinking, cultural backgrounds, etc., which govern how we explore, combine, and evaluate ideas. Recent studies in cognitive neuroscience reveal that flexible attention control—the ability to shift between focused and diffuse thinking is a key predictor of creative insight, linking attention dynamics directly to the stages of the creative process. As a result, modelling attention has become a critical frontier in affective computing and cognitive AI, offering new pathways to understand and augment human creativity. In this work, we propose a computational framework to detect creativity stages by analyzing indicators of attention. We first examine the interplay between cognitive attention types (Focused, Defocused, Flexible, and Sustained) and creative ideation. Then, we explore computational models that infer attentional cues. Building upon this foundation, we propose a multimodal approach that combines AI with user-level interaction signals (e.g., voice cues or behavioral metadata) to predict creativity stages. We conclude with a discussion on the applicability of our method in real-world scenarios and its integration in intelligent systems.