Can we settle for less than full automated grading? How about automated size determination, edge evenness, black light alterations, et all (surface wrinkles?). These would be easy to implement with a simple high res scan and some programming. The intent of the analysis is to
weed out obvious bad/altered cards, not to grade them. Cards would fall into 3 categories: Pass, Fail (altered), more review needed. Fails are rejected and the other 2 go through the same process as before for actual grading.
Heck, based on how many trimmed cards are getting through, if they were to implement this, it would easily cut PSA's return time in half.