This paper explores the intersection of traditional Model Predictive Control (MPC) and Artificial Intelligence (AI) in control problems. AI-based control is a compelling alternative to classical MPC models, as it effectively manages linear and non-linear systems, even when system equations are unknown. This paper presents a novel framework employing Feed-Forward Neural Networks (FNNs) for BallBot balancing. The proposed AI-based control showcases efficacy, efficiency, and resilience to sensor noise. Furthermore, the framework enables optimization of the neural architecture corresponding to the BallBot’s initial state, offering a robust solution for practical applications.