BINGO is a project aiming to set the grounds for large-scale bolometric neutrinoless double-beta-decay experiments capable of investigating the effective Majorana neutrino mass at a few meV level. It focuses on developing innovative technologies (a detector assembly, cryogenic photodetectors and active veto) to achieve a very low background index, of the order of $10^{-5}$ counts/(keV kg yr) in the region of interest. The BINGO demonstrator, called MINI-BINGO, is designed to investigate the promising double-beta-decay isotopes $^{100}$Mo and $^{130}$Te and it will be composed of Li$_2$MoO$_4$ and TeO$_2$ crystals coupled to bolometric light detectors and surrounded by a Bi$_4$Ge$_3$O$_{12}$-based veto. This will allow us to reject a significant background in bolometers caused by surface contamination from $\alpha$-active radionuclides by means of light yield selection and to mitigate other sources of background, such as surface contamination from $\beta$-active radionuclides, external $\gamma$ radioactivity, and pile-up due to random coincidence of background events. This paper describes an R\&D program towards the BINGO goals, particularly focusing on the development of an innovative assembly designed to reduce the passive materials within the line of sight of the detectors, which is expected to be a dominant source of background in next-generation bolometric experiments. We present the performance of two prototype modules -- housing four cubic (4.5-cm side) Li$_2$MoO$_4$ crystals in total -- operated in the Canfranc underground laboratory in Spain within a facility developed for the CROSS double-beta-decay experiment.