This is done by a lot of customer support and helpdesk systems that one would almost consider “legacy” that are certainly not related to AI in any way.
So I would assume that the uptick is caused by you moving somewhat up in you career so you deal with such BS systems more often.
On the other hand, the approach with explicit markers in the email is reliable. Alternative is some bunch of ad-hoc rules that will extract the actual reply from the reply, which has a lot of edge cases (which for some systems even extend to edge cases that involve the MIME envelope, not the message text itself).
So I would assume that the uptick is caused by you moving somewhat up in you career so you deal with such BS systems more often.
On the other hand, the approach with explicit markers in the email is reliable. Alternative is some bunch of ad-hoc rules that will extract the actual reply from the reply, which has a lot of edge cases (which for some systems even extend to edge cases that involve the MIME envelope, not the message text itself).