There are various areas where the ATC function could be "assisted" by computers - whether it's Area, Approach or Aerodrome control. Each area has to be rigorously studied for every eventuality and hence the slow "progress" in these areas.
When everything is "standard" it's quite easy to see the benefit of automation, in practice there are so many variables (some of which will never have been seen before) that even the cleverest level of artificial intelligence would struggle.
That's not to say that humans don't also struggle in new situations, and certainly don't always make the right decisions.
One of the huge issues is, of course, if the computers are determining minimum separations in order to squeeze as much traffic as is possible through the system then what happens when something non-standard happens. Artificial Intelligence is making huge bounds, and will continue to do so, but it relies on the quality of its data. There are huge limitations associated with raw data, the transmission of data, the timeliness of data, the number of sources of data, conflicting data and so on.
How do you quiz a computer about its actions at a subsequent board of inquiry? Who's to blame? The programmers?
Confidence will grow, in time, following many many trials of computer-assisted decision-making but it's still a long way off, in my opinion, before you can remove the human ATCO (for all its failings) from the system. In the short-medium term I can see that ATCO numbers will be reduced, with the functions becoming more and more reliant on computer-assisted decision-making and "tools" but with contingency for human recovery.
It may well take the "fun" out of the job...but will almost certainly be more efficient. In some ways the computerisation may even assist teaching ATCOs - faster and more efficiently that just a human OJTI could.