I argue that test automation was more salient at the management level than at the Craft level.
Test automation was about avoiding situations where a bug wasn't found because a test wasn't rerun after a change. However, automation wasn't chosen because of a cost-benefit calculation, made by testers (When Should a Test Be Automated?). Rather it was *primarily* imposed by management because:
* Testing tool vendors claimed the lifetime cost of an automated test would be only trivially greater than of running the test manually once. Managers bought that. (But the tools didn't deliver.)
* Bugs missed because a manual test wasn't rerun after a bug was introduced are more *visible* and *embarrassing* than when there was no test ever that could have caught it. ("I *told* you we were understaffed.")
* Software is *about* automating rote tasks. Why are we, a software vendor, not using it for this expensive part of our process? ("Because bug finding isn't – or at least shouldn't be – a rote task" was not generally an acceptable answer from a low-status part of the organization.)
It was more of a social phenomenon than an economic decision.