Aggregate Production Planning (APP) is the key operational decision making in manufacturing, that reconciles short- to medium-term demand with actual production capacities and their usage over a time-horizon. As a result orders within supply chains are triggered, machinery and equipment prepared and workers hired. The increasing uncertainty in nowadays economies, requires shorter decision making capabilities, adaptable control methods, intelligent and integrated decision making. In such an environment, traditional, rigid rule-based approaches fail to incorporate the adaptivity, while optimization based approaches that assume stability, are rigid and require intensive computational effort. By using recently emerging Large Language Models (LLMs), quicker and more individualized decision making can be facilitated. Instead of human experts adapting the APP decisions to reality, LLMs offer the capability to generate programming code and adapt existing decision rules or develop new rules. These heuristics can then be benchmarked in a simulation model that runs with the current expectation of future events as a digital twin and optimal decision policies derived.