If you do look at case studies of major transformation or work getting worse you don’t see IT being central to it. For instance, you don’t see longshoreman slinging bags of sugar with hooks but instead you see a longshore(wo)man lifting a whole shipping container with a big crane. In cases like medicine, teaching, etc. that some would say are becoming awful jobs one could say they are awful jobs because there is not enough innovation, it is not possible to invest in capital to make workers more productive but rather the one knob you have to turn is tightening the screws on workers to lower costs by nickel and dimming them.
Not to say the future is like the past. Since Adam Smith it has usually been the case that trade ‘floats most boats’ but after 2000 the rapid industrialization of China happened quickly enough that manufacturing workers in the US wound up much worse. You might make the case that this will finally happen w/ automation after 60 years of handwringing.
Back in the 70s there was discussion of specifically taxing computer components so as to provide funds for handling the reduction in work hours/positions which such would facilitate --- perhaps we need to do something along these lines for AI?
At some point, automation and efficiency should reduce the number of hours of labor which the world as a whole needs to be done --- what is to be done for the folks who are rendered unemployable? At a job call at my previous employer (a large, privately-owned printer) we hired 100 entry-level folks --- six months later, only 1 of them was still employed at the company, the balance either having moved on to other jobs, or having been terminated for performance and attendance issues, mostly the latter. Pay and benefits are quite competitive.
The documentary American Factory (2019) is a relatively recent human interest case study. Viewed in this light, automation generally acts as a fell-swoop solution to older approaches in businesses capable of making the required capital investment.
In terms of a process engineering business case, you can rarely just fully automate one unit operation, because the benefits will be quashed by adjacent, older process. It makes far more business sense to go the whole way and automate everything.