I've read a couple of books on archiving and they all say that first step
is selecting, i.e. a manual selection like Tom suggests. Just saying
because Tom referred to what people are learning. From my own experience
working in an audio archive that did happen only in, say, half of the
cases. But I do think that archives already have become a lot more cautious
in acquiring new collections because they know their limitations quite
well, e.g. not being able to catalogue it within the next 20 years etc.