Research by computer engineers on the technical and scientific aspects of AI music generation has expanded rapidly in recent years (Cros Vila et al. 2025; Hernandez-Olivan and Beltran 2021; Mycka and Mańdziuk 2024; Rowe 2020; Yang and Lerch 2020; Zhao et al. 2025). However, few studies have examined AI as part of a collaborative creative process (Knees et al. 2020; Newman et al. 2023; Rohrmeier 2022). Existing examples of AI-based artists show that coherent musical output is possible (King and Earl 2025; Kleinman 2025; Lozo 2026; Vaziri 2025), but research rarely documents the artistic processes or reflects critically on human choices in AI music generation (Gioti 2020; Micchi et al. 2021). Studies examining AI as part of a human-led approach, in which composers guide and transform commercial AI-generated material, have yet to be conducted. Addressing this gap, this paper examines AI-assisted music-making through hokutone, a virtual Japanese rock band and AI music project that serves as a platform for practice-based research into human and AI collaboration. It investigates how human composer-songwriters work with AI-generated music as a creative tool, considering its advantages, limitations, and the ethical concerns surrounding intellectual property. Central to this process is the design of prompts or written instructions that commercial AI platforms such as Suno rely on to generate music. This paper explores human interaction with AI-generated music, analyzing how composers experiment with creative choices, express their artistic identity, and the extent to which AI contributes independently, revealing the complex dynamics of creative collaboration with technology.