I have seen many examples that Rasa NLU is very good and efficient for building chatbot solutions. Means it is efficient to find intent and entity out of small chat conversations.
But is it preferable to find the intent from complex paragraphs as well? For example is it capable to use for long emails?
Rasa NLU is most suited for shorter messages since it was built for contextual AI assistants. You can still try running it on larger paragraphs, but you probably need to add a lot of data. Maybe some approaches such as Doc2Vec might be better suited here.