Industrial companies face challenges caused by volatile business environments. In particular, changing customer demands can have large impacts on their performance. Medium-term capacity management addresses this challenge by adapting the production capacity to the customer’s demand. This is usually done by experts or, in the best cases, supported by production planning software using linear programming or genetic algorithms.There is a growing trend for using production simulations or digital twins. These simulations can be coupled with demand forecasts and machine learning to create a planning assistant tool. This tool can support the operations manager to fnd the right set of measures to adapt the production capacity based on changing customer demand.In our work, we design such a planning assistant tool by using different algorithms, including deep reinforcement learning. We apply it in a learning factory for performance comparisons and can show promising results of deep reinforcement learning. Furthermore, we can show that this simple decision support system outperforms humans in the test setting.