Advanced Process Control (APC) has historically provided substantial benefits in iron ore operations by stabilising circuits and reducing process variability. Despite these advantages, many sites continue to encounter persistent challenges related to extreme ore variability, rapidly shifting operational constraints, and competing production targets. These dynamic variables severely limit the ability of static control logic to consistently achieve optimal economic performance. To bridge this gap, this paper outlines a practical method for enhancing existing APC infrastructure through AI-driven optimisation, emphasising Reinforcement Learning (RL) as a supervisory layer. Rather than replacing established controls, this approach enables operators to maintain baseline stability while AI continuously adjusts operational targets in response to actual plant conditions.