Sales forecasting is essential for decision-making and are crucial in many areas of a firm, such as planning and scheduling, resource management, marketing, logistics, and supply chain. Due to the fluctuations in retail sales, prediction with high accuracy is a challenging task. In this context, this study proposes a framework that combines ensemble empirical mode decomposition (EEMD) based on artificial intelligence models to forecast the retail sales of a Rossmann Store, using a multi-step-ahead forecasting strategy, in the task of time series forecasting with one, seven, and fourteen-days-ahead. The forecasting models of the retail sales time series are Bayesian Regularization of Artificial Neural Networks, Cubist Regression, and Support Vector Regression. The performance of the proposed forecasting models were evaluated by using two performance metrics: mean absolute percentage error and root mean squared percentage error. The EEMD models outperform the single models in all forecasting horizons, with a performance improvement that ranges 1.30% - 76.25%. Indeed, EEMD models are efficient and accurate models for retail sales forecasting.