In recent advances in generative AI complex parametric CAD models can be automatically created from natural language. Current research focuses on geometric forms like cuboids or cylinders, but these models often lack an understanding of engineering design intent and functional requirements. In particular, they fail to capture machining features such as holes, pockets, slots and parallel keyways, which are essential for real-world manufacturing. Without recognizing the role and purpose of these features, AI-generated models remain difficult to integrate into practical engineering workflows. This disconnect reveals a critical gap between current generative CAD tools based on Large Language Models (LLMs) and the functional expectations of engineering design. This work addresses this gap by proposing a new method for annotating existing parametric CAD models with machining features, enabling downstream generative AI tools to learn a more nuanced understanding of functionality and design intent. The approach leverages the structured nature of parametric CAD, where geometry is created through a sequence of feature-based operations. A pipeline is introduced that parses these design histories to identify machining-relevant features and classify them by function. To support this, a taxonomy of machining features grounded in manufacturing practice is defined, which forms the basis for consistent annotation and model evaluation. The method is validated on a real-world dataset of shaft components, demonstrating its effectiveness in capturing functional design elements critical for manufacturability. The results show that the proposed approach achieves high accuracy in identifying functional features across a diverse set of parts. This demonstrates that generative AI can be extended not only to produce valid geometry, but also to use context-sensitive functional features. This work provides a critical step towards bridging the gap between generative CAD modeling based on LLMs and real-world engineering applications, enabling AI models that are both creative and grounded in manufacturing reality.