The objection that the data is too messy, and why it argues for a different method, not against automation

Messy data is real and expensive, and it defeats rules that require clean input. That is exactly why the case for automation rests on a method that reads the data as it is and routes the doubtful cases to a person.