As global markets become increasingly complex and fast-paced, effective Supply Chain Management (SCM) is essential for synchronizing production, distribution and customer fulfilment. A reliable demand forecasting is essential to SCM in order to effectively utilize inventories and minimize operational inefficiencies. The bullwhip effect, which occurs when minor fluctuations in demand are magnified upstream due to poor forecasting, results in excess inventory and stockouts. Dynamic market patterns cannot be handled by conventional techniques. In this work, a Hybrid Generative AI Framework is presented, utilizing Bi-directional Recurrent Neural Networks (BRNNs) for demand forecasting and Variational Autoencoders (VAE) for latent feature extraction. Through the use of deep learning and generative AI, with synthetic data augmentation, the model stabilizes supply chains, achieving a Mean Absolute Error (MAE) of 1.5486, Root Mean Squared Error (RMSE) of 2.4737 and an R-squared (R2) value of 0.9996 thereby offering an adaptable way to lessen the bullwhip impact for effective inventory management. This methodology captures time-dependent demand patterns and adapts to changing market dynamics, allowing for informed and strategic decisions. The presented framework exhibits strong generalization capabilities by modelling complex, non-linear relationships and simulating plausible demand fluctuations. Ultimately, it contributes to minimization of surplus stock, lowering of energy consumption from excess production and improving product availability for end consumers.