One of the arguments deployed against human editing or cataloguing of the Internet is that would be too expensive and slow. This is also an argument used in many library contexts (generally not archives since those institutions work with mostly unique materials) against original library cataloguing; it is better to simply digitally import catalogue records from a national library or commercial service.
One of the hats I wear at my university involves doing some cataloguing for an academic research project. Out of curiosity, I started to plot how many steps are involved in cataloging a single item into the database. I stopped counting at step 13, but I would say that there are about 20 steps. That's 20 steps PER ITEM; it takes roughly 3-5 minutes per item catalogued. I would say 80-90% of these steps are highly repeatable, perhaps even amenable to automation with a script. One could respond this by saying "Oh we'll just automate everything and fire the cataloging staff." That's not what I'm advocating here. Rather, I am suggesting that one analyze the process and seek to minimize the ineffective or repeatable steps and let staff focus on the more challenging tasks.
That said, there do not seem to be the right mix of incentives to make this happen. Or perhaps there is just a lack of interest in innovation? I wonder if there is a better way of saying, "A lot of this is boring and highly repeatable; can we work with the IT staff to automate parts of this so that we can all be more effective at work?"