Current creativity research reveals a fundamental disagreement about the nature of creative thought, specifically, whether it is primarily based on automatic, associative (Type 1) or executive, controlled (Type 2) cognitive processes. We propose that Type 1 and Type 2 processes make differential contributions to creative production depending on domain expertise and situational factors such as task instructions. We tested this hypothesis with jazz pianists who were instructed to improvise to a novel chord sequence and rhythm accompaniment. Afterward, they were asked to perform again under instructions to be especially creative which, via goal activation, is thought to prompt the musicians to engage Type 2 processes. Jazz experts rated all performances. Overall, performances by more experienced pianists were rated as superior. Moreover, creativity instructions resulted in higher ratings. However, there was an interaction between instructions and expertise, revealing that explicit creativity instructions significantly improved improvisation ratings only for the less experienced musicians. We propose that activating or reconfiguring executive Type 2 processes facilitates creativity for less experienced musicians, but does not improve creative performance significantly for more experienced ones because the latter have largely automatized the processes responsible for high-level improvisation or because they have achieved a near-optimal balance between associative Type 1 and executive Type 2 processes. Thus, increasing controlled Type 2 processing is unlikely to help, and may sometimes even diminish, the creativity of experts' performances.