you are already using AI for banning accounts, so it shouldn't cause any more resource, just small adjustments.
You need humans for good CS, and humans are both non-automatable and expensive. Google does have some humans in CS (I think), but they try to automate as much of it as possible, hence the prevalence of boilerplate responses from them.
Because these are the lowest effort and cost actions they can take and there are no consequences to doing it like this.
I am certain that if there are any developers that care about policies, it's the indie ones. And there are still too many rules for them to follow, and invisible lines they're unaware they can't cross.
The developers at Big Tech don't get banned from Google Play because Google isn't going to de-list Netflix, Spotify or Microsoft.