One problem with automation is that the people in upper management themselves rely on having lots of people to manage. Getting rid of those people puts their own position and standing at risk. Also, we don't yet have humanoid robots with full feature-parity to actual humans, so automating some jobs might require a change of process to convert it to one that is amenable to automation - in doing so, it may arise that the entire process is pointless and this would obsolete the jobs of a lot of paper-pushers, raise an uncomfortable question and ultimately be a politically unpopular move.

A second problem is that non-tech companies lacking competent in-house technical expertise will always choose to outsource it to a shitty consultancy which will take advantage of them, waste years and not deliver anything viable or cost-effective in the end.

I think an issue may be that a lot of the easy wins for automation have been done. There are many things that are seen as “low skilled” but very hard to actually automate.

Isn’t it trivially true that the things that are more worthwhile to automate have been automated and the things that are less worthwhile to automate haven’t been automated yet?

I think that’s not actually true because people have incomplete information. I think it’s also not static – for example it’s much easier to automate ordering of your product, arranging payment, etc, now than it was 20 years ago, even though the technology for forms/payments over the Internet still existed back then.

To be clear, when I write automation, I don’t really mean automating whole jobs but rather automating the bits of jobs where people don’t add value so that they may spend more of their time doing the things that add value.