Inferring which downstream system an action's content implies, rather than requiring explicit per-action-type routing configuration for every case.
Classifying the action's content against known destination-system patterns.
LLM API or a trained classifier, with a rules-based fallback