Abstract
Generative artificial intelligence (GenAI) is rapidly reshaping innovation processes, with organizations increasingly deploying large language models to support idea generation, research, and prototyping. While experimental studies suggest that AI-generated ideas can perform well on common creativity metrics, breakthrough innovations continue to emerge disproportionately from expert ideators. This raises a central question for innovation management: in which stages of the creative process does GenAI meaningfully augment expert ideation, and where does its role remain limited?
To address this question, we conduct a qualitative study of eight award-winning ideators who have repeatedly won competitive crowdsourcing contests both before and after the release of ChatGPT. These contests involve real organizational challenges and selective evaluation processes, providing a relevant context for examining high-level ideation. Drawing on semi-structured interviews and structuring our analysis around Wallas’ four-stage model of creativity (preparation, incubation, illumination, verification), we examine how GenAI is embedded within established expert routines.
Our findings reveal a consistent stage-contingent pattern of augmentation. Expert ideators primarily deploy GenAI in the preparation phase to accelerate research, synthesize information, and map the solution landscape, and in the verification phase to refine, articulate, and visualize ideas. In contrast, incubation and illumination—the stages associated with insight generation—remain predominantly human-driven. While AI occasionally serves as a stimulus or conversational sparring partner, participants consistently attribute breakthrough insights to experiential judgment and contextual understanding.
Interpreting these patterns through a knowledge-based lens, we argue that GenAI predominantly enhances access to and recombination of explicit knowledge, whereas tacit knowledge rooted in experience and contextual sensitivity remains central to outstanding ideation. The study contributes a process-level perspective on AI augmentation in creativity and advances a stage-contingent account of human–AI complementarity. For managers, the findings suggest that effective AI integration requires deliberate process design that positions GenAI as a research and refinement tool while preserving human-led insight generation.