Presumably all these agents would have been trained on different data, with different viewpoints? Otherwise, what makes them different enough from each other that such a "conversation" would matter?
Different skills or plugins, different views and different tools for the analysis of the same object. Then the debate starts.
Different skills or plugins, different views and different tools for the analysis of the same object. Then the debate starts.