Did ATMs kill the bank teller's job?

A thought on Alain de Botton's Status Anxiety (my current reading) that struck me. Writing in Fast Company, Charles Fishman once made the point that employment has grown in some service businesses that have been automated:

At the dawn of the self-service banking age in 1985, for example, the United States had 60,000 automated teller machines and 485,000 bank tellers. In 2002, the United States had 352,000 ATMs - and 527,000 bank tellers. (Link to a KioskCom mirror of the article)

Yet de Botton makes the complete opposite point:

ATMs offered bank tellers few grounds for celebration... In the United States, 500,000 people, around half of the workforce in retail banking, lost their jobs between 1980 and 1995, in part because of the invention of these silkily efficient machines. (Status Anxiety p. 101)

So: which is it? Did ATMs kill retail banking or not? (You might argue that there was a bounce back in the number of retail banking staff between 1995 and 2002, so both are right, but I'm inclined to think that isn't the case: if retail banking was bouncing back so significantly, it would be so visible that de Botton would be unlikely to make the argument he did.) Personally, I'm inclined to go for Fishman's view - that the ATM changed tellers' duties from counting cash for deposits to a greater range of financial assistance. Plus, Fishman uses the actual number of tellers (which the ATM directly competed with, presumably), whereas de Botton uses the retail banking industry workforce as a proxy. There were probably a lot of forces intervening in between 1980 and 1995 that probably shrank that workforce - the Savings and Loan crisis and the increased consolidation of banks come to mind - that didn't have anything to do with the ATM. But I suppose it's always easy to blame a soulless machine rather than mismanagement and inexorable forces.