Selecting an appropriate method for demand forecasting plays a critical role in successful supply chain management (SCM). However, manufacturing companies face markets defined by uncertainty and constant change. In this situation, traditional statistical techniques are not always adequate to provide reliable and accurate forecasts. AI algorithms therefore appear as a promising solution to enhance these methods. However, the existing literature is often limited to a general description of artificial intelligence methods used for demand forecasting. This article presents a comparison of three main methods: artificial neural networks (ANN), random forests, and the ARIMA model.