AI systems can generate reliable outputs without human understanding, yet human insight remains valuable for shaping subsequent work, forming preferences, governing collective choices and satisfying curiosity. We introduce understanding-first AI engineering, treating purpose‑specific human understanding as a primary design principle that guides requirements, design and operation.\ \ Understanding purposes are organized into four complementary rationales: reflective agency, effective action, collective governance and epistemic value. A reference architecture links production, verification, use and learning‑exploration through shared records, supporting understanding via explanation and participation.\ \ Five candidate architectural patterns address parallel learning, explanatory models, goal exploration, audience‑specific views and distributed expertise. Qualitative scenario analysis of two hypothetical use cases shows how purpose classification, the reference architecture and chosen patterns steer investigation, participation and decision processes while retaining operational controls. The cases illustrate design consequences and limitations rather than prove empirical effectiveness.\ \ The approach makes human understanding an explicit engineering commitment whose justification need not depend on improving the original AI task.\ \ Review