My first thought would be: How much unique data is there?  This is to
say, if the PC disk was lost, would most of it actually be rebuilt
from a central reinstall image of some kind, rather than representing
unique data?  If so, a lot of stuff could be excluded as unnecessary.

I would then look into whether all these PCs actually store generated
data, as opposed to having all data worth having actually residing on
central file servers, where a single point of backup/restore would
prevail.

If all these PCs are unique and individually storing data, a better
approach to PC implementation needs to be looked at, as backup/
restore, network loading, and the overhead of fielding possibly
thousands of phone calls could be untenable...  As in, who the heck
came up with such a rag-tag deployment, anyway?

This is all going to make for an "interesting" service agreement...
one which will have to say that unless some kind of real-time backup
is put into place, data will inevitably be subject to
unrecoverability due to standard backup intervals.  Assure that those
served aren't expecting miracles, to avoid political calamities.

  Richard Sims