When working from an existing FAQ data, the ordering of when questions and answers are entered doesn't really matter as the questions have already been defined.
However, when working with data that does not yet have existing questions, it may feel more intuitive to start with the data being targetted, then build questions around it.
Knowledge owners can easily expose their useful data (answers) but might not know how the AI is best tuned to get to it (questions). I wonder whether knowledge population might eventually be undertaken by different users - one for answers another for questions. That could work both ways: Question user raises questions that are of interest, but answers not known. Or useful answers exposed by Answer User which the Tenjin expert can then tune.
This is pure speculation on my part but when I've been populating data, it feels like I'm starting with the answer.
