Abstract
This paper discusses principles for designing instructional content intended for future integration into a virtual AI-assisted learning environment grounded in national cultural codes. The focus is not on building the AI system itself, but on developing a content model ready for digital implementation (culturally coded scenarios, a task bank, an assessment rubric, and dialog specifications). The design process follows a Design-Based Research (DBR) framework through iterative cycles of design–review–refinement. National cultural codes are operationalized into teachable units and described via their communicative functions; the codes are mapped to soft-skills indicators using a matrix approach; a task typology and content specification are developed; and CEFR-based A2–B1 level alignment is applied. A “Health” topic case study demonstrates level-appropriate integration of proverbs, idioms, and lexical polysemy, clarifying the link between cultural codes, pragmatics, and soft skills development. The study contributes a structured content blueprint for a “virtual tutor” with two-layer feedback (linguistic correction plus cultural-pragmatic alternatives). Further work is expected to extend the model to additional topics and validate effectiveness through empirical piloting.