The advancement in Neural Machine Translation (NMT) has significantly improved the localisation of content across multiple languages, offering fluency and efficiency. However, in complex applications, such as the translation of video games, where there is a need to preserve original player experiences, NMT alone cannot capture subtleties such as humour. As a result, human oversight remains essential to ensure that audio and text translations retain their original intent and appeal. In batch workflows that involve millions of words, the assignment of post-editing/review jobs to the rightful personnel is tedious to complete manually. We present the progress and challenges encountered during the development of a recommender system (RS) designed to enhance video game translation workflows. The main goal of this work is to improve time and cost efficiency in real-time localisation workflows and ensure that unique aspects of game narratives are preserved while meeting the demands of global audiences. The results of the online A/B evaluation in more than 292,000 video game translation workflows demonstrate that the recommender system achieves more than 90% time savings and up to 76% cost reduction compared to manual assignment.