I think one major issue is that manual processes often don't really "hurt" managers, they hurt the people doing them. To a developer or ops admin, "This $badthing caused by a typo won't happen again if we automate this" sounds very reasonable but sometimes managers just see it as a mistake that someone will need to be more careful about. Once they've factored the turnaround time into how they estimate they don't seem as interested in speeding things up. All that seems exacerbated by reactionary/feature driven mindsets which makes any kind of improvement that doesn't have an external event/person driving it very hard to get slotted. This seems to be made worse if you're working with a manager that's never really done the work that the people they're managing do or they did it so long ago they don't have an intuitive understanding of how it works today.
Another one is that a lot of people who perform a manual process for a long period of time become invested in it and defensive of it. No matter how much faster or better it is and no matter how involved they are if they end up feeling threatened it's probably dead in the water. Pretty typical human nature thing I think but at a big company with a lot of old established manual processes this can slow automation efforts down to a glacial crawl. Even if there's only one person that fully understands it I don't think managers appreciate the risk of having all of your eggs in one basket.
Eventually things do get automated but it seems like it's driven by a disaster or politics more than anything else. And sometimes not even disaster seems to be enough to drive real change if someone isn't pursuing it.
If anyone has tips on how to deal with any of this I'd love to hear them.