Do the developers of legal bots have sufficient knowledge and experience of the law? Is the data used to “train” their algorithms timely? Will critical evidence be filtered out?