A socio-technical architecture for culturally adaptive AI: Human-in-the-loop integration for sustainable operations

• Culturally Adaptive Framework: Proposes a "Culturally Adaptive AI" framework to harmonize algorithmic efficiency with social norms in high-power-distance cultures. • Sustainability Mediation: Demonstrates that AI adoption positively mediates the relationship between Green HRM (GHRM) practices and organizational sustainability performance. • The "Kreng Jai" Paradox: Identifies unique cultural moderators, specifically Kreng Jai (deferential consideration) and "face-saving," that necessitate a "human-in-the-loop" approach to AI. • TOE Predictors: Reveals that Technological Readiness and Organizational Support are the primary drivers of AI integration for sustainable futures. • Institutional Theory Contrast: Uncovers a paradox where Environmental Pressure shows a significantly weaker effect on AI adoption than predicted by Western-centric models. • Mixed-Methods Rigor: Utilizes a sequential explanatory design combining PLS-SEM analysis of 420 professionals with in-depth qualitative insights from 25 experts. The integration of Artificial Intelligence (AI) into sustainable operations frequently encounters algorithmic friction due to localized socio-cultural constraints, which Western-centric models often fail to address. This study proposes a novel constraint-aware algorithmic architecture for "Culturally Adaptive AI" that utilizes a structured Human-in-the-Loop (HITL) approach. The proposed system architecture consists of four distinct layers: a Data Integration Layer, an AI Decision Engine Layer, a Cultural Adaptation Middleware, and a User Interface Layer. To empirically validate the system's efficacy in driving sustainable performance, a sequential explanatory mixed-methods design was employed, combining a quantitative evaluation of 420 professionals with qualitative insights from 25 expert interviews. Structural equation modeling (PLS-SEM) confirms that technological readiness and organizational support significantly drive system adoption, effectively mediating the optimization of sustainable operations. Furthermore, qualitative findings reveal that specific cultural variables, such as "Kreng Jai" (deferential consideration) and high power distance, necessitate parameterizing the Cultural Adaptation Middleware to temper the speed of autonomous deployment. By harmonizing algorithmic efficiency with social norms, this engineering framework provides a robust, scalable blueprint for implementing HITL AI systems in emerging economies to achieve Sustainable Development Goals (SDGs).