The steel industry, known for its complexity and the need to reduce CO2 emissions, is adopting advanced digitalization tools to move towards a more sustainable, integrated, and agile operating model. Digital twins with artificial intelligence-based optimization and scheduling models can improve decision-making in logistics, refractory maintenance, and energy efficiency. By incorporating advanced AI algorithms into this decision support system, the hot metal route scenarios can be evaluated, resulting in minimized hot metal temperature losses and increased scrap utilization. This paper integrated digital twins with reinforcement learning algorithms to investigate the logistics of torpedoes and hot metal ladles. It considered important input parameters such as the ladles and torpedoes' thermal state and location, refractory thickness, hot metal volume and temperature, and crane availability. By incorporating advanced AI algorithms into this decision support system, energy-efficient scenarios can be evaluated, increasing scrap utilization and resulting in a possible reduction of 15 °C in hot metal temperature losses.