Supply chain management (SCM) is considered as one of the key elements to leveraging a company's success. Enterprises must adapt their supply chain to cutting-edge technologies and techniques in order to improve their performance, reduce costs and provide better service. Several of these recent breakthroughs have been allowed by Machine Learning (ML), providing solutions to complex problems that were until now difficult to solve, or enhancing the results of former methods. Because of the importance of the field, this paper aims to investigate representative applications of ML in SCM, with an emphasis on the specific area of Demand and Sales Forecasting (D&SF). The purpose of this study is twofold: firstly, to point out the most recent ML trends used in D&SF; and secondly, to explain when companies should invest in novel forecasting techniques over traditional methods.