Predictive maintenance is key in industrial environments, where tool wear directly impacts production efficiency and quality. However, the scarcity of labelled data makes it challenging to implement accurate models. This paper explores Few-Shot Learning (FSL) for tool wear prediction in a data-limited manufacturing environment. A case study is presented of a metal fabrication and precision turning company facing the challenge of optimising the use of its cutting tools. The objective is to develop a predictive system that, based on the analysis of historical data, accurately estimates when a tool is close to wear. The proposed approach is expected to anticipate tool wear, optimise maintenance planning and reduce unplanned downtime. In future work, FSL will be compared with models based on Remaining Useful Life (RUL) estimation to determine the most effective strategy.