This case examines how a non-technical musician used iterative prompt-based “vibe coding” to develop Gary, an AI-driven music tool, illustrating how accessible generative AI enables domain experts to create functional applications without formal programming skills. The analysis utilizes the Balanced Scorecard and ambidextrous organization concepts to illustrate how exploratory, sandbox-based innovation diverges from the structures necessary for scaling and operational reliability. The case identifies factors that supported progress, such as iterative experimentation, creative ownership, and low-cost prototyping, alongside challenges including scalability constraints, role ambiguity, and limited governance. The study concludes with managerial insights on how organizations can support grassroots AI innovation by aligning resources, clarifying roles, and establishing frameworks that allow non-technical innovators to experiment safely while preparing for responsible and sustainable growth.