This article establishes the theoretical and methodological foundations for the development of an automated feedback system for mechanical engineering drawings aligned with IRAM standards and the use of generative artificial intelligence (GAI). The study reviews literature on feedback in educational settings and AI-based tools applied to technical drawing, highlighting the need for specific normative verification criteria to guide automated analysis. It proposes an initial classification of these criteria based on key categories of graphical representation in engineering drawing: orthogonal projections, dimensioning, title block, and axonometry, and grounds their selection in official IRAM documentation and previous studies on common errors in educational practices. The article also discusses the current limitations of global AI models when handling local technical regulations. This research is a first step toward creating a computational framework that enables the training of generative AI models to deliver relevant, coherent, and context-appropriate feedback in Argentine technical environments, ensuring alignment with national professional standards in teaching.