The built environment (BE) sector faces growing pressure to improve sustainability performance and artificial intelligence (AI) is increasingly used to support sustainability decisions. Beyond model performance, the decision-support capacity of AI also depends on underlying data characteristics. However, existing research mainly focused on AI techniques and model performance, with limited attention to the data characteristics underpinning AI applications. This study examines how data characteristics shape AI-driven decision-making in the BE sector.
Following PRISMA methodology, this study systematically reviews 125 studies to analyse patterns of AI-driven sustainability decision-making and the underlying data characteristics in the BE sector.
AI-driven sustainability decision-making contexts concentrate in energy management and multi-dimensional sustainability trade-offs. Prediction and optimisation are dominant decision focuses, while monitoring and control remain comparatively limited. Most AI workflows rely on secondary, structured and historical datasets, as unstructured and real-time data require more extensive preprocessing. Hybrid AI approaches and diverse preprocessing techniques are adopted to address fragmented data characteristics. Recurring data limitations highlight the importance of data governance in supporting AI-driven sustainability decision-making.
The review advances current knowledge by synthesising AI-driven sustainability decision-making alongside the underlying data characteristics. Findings show that data characteristics and preprocessing requirements influence the decision-making contexts, decision focuses and AI techniques that can be supported. The study further indicates that AI maturity depends not only on AI technique advancement, but also effective data governance, AI skills and data literacy.
