That really is the key point - there are places where total automation makes sense, but not nearly as many as the people invested in/developing total automation would have us believe. In the vast majority of cases, using automation to *augment* humans is better - or at least, more cost-effective and predictable.
In a few years time you probably could design a totally automated sci-fi-future mass transit rail network that would operate as a completely closed system without any human intervention at all...but is that actually needed? Would the gains of such a system outweigh its costs? Or its opportunity costs, given all the cash it would require to develop and build could instead fund huge improvements to the current system?
Full automation does have its place, but I think more often that not it's a big flashy distraction from more tangible improvements we could be making *now* without funnelling billions of public money into the pockets of our techbro overlords. Driverless cars could solve RTCs and congestion - or we could just make our cities more liveable and reduce car use. Driverless trains could address driver shortages and sparse timetables, but it would probably be cheaper to just hire a few more drivers and give them the pay & conditions they want(I note that even as China spent billions constructing brand new high speed rail, they still seemed to think it was more efficient to retrain rural *steam train* drivers to operate the new stock than to set the system up to be fully automated).
And if someone were to think it's unfair that train drivers can use their union to get better pay and conditions while most people are stuck in crappy jobs with no bargaining power, perhaps making efforts to change the latter would be better than playing Crabs In A Bucket over the former.