Buildings own diverse datasets, including geometric, product, logistic, real-time monitoring, regulatory, and occupant feedback data. However, challenges such as data scarcity, insufficient labelling, and the complexity of multimodal data limit conventional AI’s ability to provide accurate, scalable and content-aware insights, often confining its application to specific buildings and time. General-Purpose Artificial Intelligence (GPAI) offers the transformative potential to maximise the value of data. Early research explores adaptive AI, meta-knowledge transfer, synthetic data, and foundation models to support generalisation across tasks. This paper examines how these developments position GPAI as a step toward general-purpose intelligence in buildings.