The growing complexity of supply chains and increased customer expectations require cognitive and responsive approaches to inventory management. Incumbent forecasting and optimization techniques provide a solution to some extent, but they do not entirely reflect entirely dynamic demand fluctuations, campaign-induced sales oscillations, and stock level changes in real time. To address these limitations, this study proposes a hybrid solution by incorporating Artificial Intelligence (AI) techniques in conjunction with mathematical optimization models for inventory control in manufacturing. The approach leverages machine learning models such as Long Short-Term Memory (LSTM) networks and Gradient Boosting Machines (GBM) to perform accurate demand forecasting, yet incorporates promotional and campaign data in order to take into account spontaneous fluctuation in demand. Forecasting results are blended with linear programming (LP)-based optimization for storage and procurement decisions under capacity constraints. Internet of Things (IoT)-induced real-time streams of data are also utilized to reconfigure supply policies dynamically, reducing overstock and stock-out risk. Simulation case study reveals that the hybrid method achieves higher forecasting accuracy, reduces total inventory cost, and maximizes service levels compared to traditional methods. The research outlined contributes towards the development of adaptive, smart, and sustainable supply chains, along with offering a direction towards future integration with deep reinforcement learning to provide completely autonomous decisions