Demand forecasting plays an important role in effective supply chain management, especially at the manufacturer stage. Despite its significance, data-driven forecasting solutions remain largely unexplored in both academic literature and practical applications. To bridge this gap, we have developed a deep learning-based forecasting system through a design science methodology. Unlike traditional demand forecasting techniques, our approach analyses the customer portfolio rather than solely relying on historical orders. By implementing our system, companies can transition from making decisions based on experience to adopting a more data-driven approach. Furthermore, our AI-powered system assists in understanding demand data and revealing previously unnoticed patterns, thus facilitating the formulation of more informed production plans.