It may be PR, but if so, it's the best kind -- there's a lesson to reflect on here.

There are limits to automation (and, more generally, any reduction of knowledge/craft to process). As you depend more on robots (literal or figurative) to execute process, over time you're likely losing a body of knowledgable/craft-capable practitioners. You can automate worker function, but until we have AI, not their perspective and the potential expertise that comes with engaging with a task over time. And you concentrate the responsibility for improvement in a smaller number of hands.

That doesn't mean there aren't also yields to automation -- sometimes big ones. Just means there are limits too.

By way of comparison... it's great that many software developers are much more productive in Ruby or Python or whatever than they would be in C or assembly. But when the time comes where you have to make a key optimization, if you don't have a "god" around who can be productive at a lower level, you may be bound to a sub-optimal approach (you might not even know about the potential optimizations).

> As you depend more on robots (literal or figurative) to execute process, over time you're likely losing a body of knowledgable/craft-capable practitioners.

We should understand this well from software. There's plenty of airline and financial software people don't want to meddle with too much because the people who remember the business rules behind it are retired or dead.

My father is actually in the process of this right now. He was one of the few people in his company who worked on the rules coded into the current software written in the 70's-80's. He's the only one who can tell the programmers why certain things were done (if they were significant, or just optimizations to make the batches run in time on the VAX hardware of the time). He's heading up the project to port the whole thing to a more modern system, but he's planning to retire soon...