> If job roles are arranged like a pyramid (wide base of lower paid jobs, with lower number of higher paid jobs above) - the largest benefit is always going to be automating away the high number of low paid jobs.

Not quite. If you can automate a CEO you might cut a million dollars out of the budget -- maybe even ten or a hundred million in some companies. But if you can automate 500K people making $15/hour, that's over fifteen billion dollars a year. Plus benefits and employment overhead. So yes, there's more incentive to automate the larger number of people.

But if the same number of people made $25/hour, then it would be over 25 billion a year. So it's both.

This has actually played out in practice with both automation and outsourcing. The auto workers were some of the first to go, specifically because there were both a lot of them and they had high wages, resulting in a very large incentive to reduce those costs once they had real competition from Japan.

It's not unreasonable for people to be willing to accept less money as one of the ways to stay competitive. The problem is when that amount of money isn't enough to make a living -- but it's not automation that makes stuff cost more. Automation makes stuff cost less. The problem is the high cost of the things that haven't been automated, and the barriers to entry preventing all of these people from doing those jobs, which should in turn bring down those costs -- unless we're dealing with artificial scarcity like zoning regulations. In which case we've identified the real source of the problem.