Advanced automation tools are promising wide-range solutions for the various problems still affecting High-Temperature Superconducting (HTS) magnets, including Non-Insulated (NI) ones. However, they are not applicable if reliable models that can run in real-time are not available. This article discusses a preliminary solution for this. For this scope, we propose the construction of reduced-order models, derived from 3D physical-based models. A Volume Integral Formulation (VIM) is presented and reduced using a technique called Proper Orthogonal Decomposition (POD). VIM solves for currents in the conducting domains, relying on Biot-Savart for interactions between elements; meshing insulating domains, such as air, is thus not needed. POD is a reduction technique where the most relevant information is retrieved by processing the full-system response through factorizations such as Singular Value Decomposition. The reduced and full models are then compared, showing that the former offers accurate solutions with a fraction of the computation effort of the latter. Finally, some potential applications of this technique are briefly discussed.