Predictive Process Monitoring is a process-mining research direction that aims to predict the future of an uncompleted process execution. The vast majority of research work focuses on techniques that are “data greedy” and require a lot of event data to be sufficiently accurate. However, the recent development of Large Language Models presents significant opportunities and potential benefits across various industrial and research domains. They are capable of leveraging their pre-trained knowledge to understand and complete tasks effectively. This paper reports on the design and implementation of a Predictive Process Monitoring framework based on Large Language Models. Experiments on multiple event logs confirm our hypothesis that Large Language Models are capable of providing very accurate predictions, even with as few as 10 training traces.