Generative Artificial Intelligence (AI) is a rapidly evolving technology with extensive applications in text and image generation. One of its most promising use cases lies in intelligent assistance systems supporting human decision-making. However, when applied to Computer-Aided Design (CAD) for manufacturing, current Generative AI approaches face substantial limitations regarding feasibility, usability, and manufacturability. The requirements for a generative assistant in 3D CAD differ fundamentally from those in text- or image-based domains, as geometric accuracy, engineering constraints, and production viability must be guaranteed. This paper investigates these essential requirements that must be addressed to enable the effective deployment of Generative AI as an assistance system in manufacturing-oriented design environments. Following the Data Mining Methodology for Engineering (DMME) Applications as a structured life-cycle framework, this study structures these requirements along the phases of business and technical understanding, data preparation, generative model development, evaluation, and deployment, providing a domain-specific requirement framework for future research and development of Generative AI-assisted design systems in manufacturing environments.