"Worth automating" isn't a simple one-time binary decision. Whether a given process is worth automating depends on the cost of automating it, and the labor costs that it will save, both of which change over time.

Japanese manufacturers at least have had similar if not higher costs on manufacturing in Japan( and Europe). And these factories still employ a large number of skilled and non-skilled workers.

I don't think a bump in salaries that puts a country in the same ballpark as existing manufacturing locations( not counting the extra costs of manufacturing in the States or in Europe) is going to be such a big push for more automation. After a certain point the versatility of human workers is still useful.

It needs to be understood that any kind of new engineering project has a risk of failure, often quite high. For example, in software, there's the infamous CHAOS report from the Standish Group that comes out every couple of years. A casual Google for the 2018 results has this quote from someone with a copy (https://hennyportman.wordpress.com/2020/01/03/review-chaos-r...)

> "The Standish Group started with a traditional metrics of success by looking at: “on time, on budget, and on target”. This means in this CHAOS Report 2018 for the year 2017: successful: 36%, challenged: 45%, failed: 19%. If they use their “modern” definition of success: “on time, on budget, with a satisfactory result.” You get almost the same results (33%, 48%, 19%)."

So, about 1/3 of IT projects are successful, including being on-time and on budget, and about 1/5 are total losses that fail to deliver any value at all. Not great odds and I'd expect the numbers are likely to be broadly similar in other engineering fields.

The decision on whether or not to embark on an engineering project (and whether to do it in-house or hire a third party) is among the most challenging duties executives have.

Toyota and Tesla found out the hard way that a lot of tasks can't be automated. Robots are still very very very dumb.