Effective domain knowledge acquisition is fundamental in Software Engineering, especially in Requirements Engineering, as it contributes to the development of systems that more accurately represent real-world needs, rules, and contexts, thereby reducing misunderstandings and rework. This is even more critical in AI-enabled systems, which depend on domain knowledge to build models that accurately reflect the intended domain-driven behavior. This work proposes a methodological and semi-automated approach, based on the outcomes of a mapping study and in compliance with ISO/IEC 5338 knowledge acquisition process for AI systems engineering, that combines Large Language Model (LLM)-based agents and Knowledge Graphs (KGs) to structure, formalize, and facilitate the usage of domain knowledge. The approach aims to transform conceptual guidance for knowledge acquisition into a repeatable, scalable, and auditable practice that supports AI Engineering, bridging the technical and management gaps identified in the systematic mapping study.