[adrianmonk]

They're efficient in terms of drivers.

Trains require large headway (gaps between trains, like following distance in a car) for safety reasons. This limits the number of trains you can run on a line. Which places a natural limit on the number of drivers you need.

Instead, you scale up passenger throughput by adding cars to trains. When you do this, you scale up capacity independent of number of drivers. Sort of a marginal cost of zero.

Another efficiency question, though, is bang for the buck. Trains are expensive to build. Headway limits the number of trains, and you can't make trains longer than the platform length. So you hit kind of a brick wall of scalability.

So it's worth asking whether automation could let you squeeze more return out of your investment in building the train line. Maybe it would unlock capacity increases through reduced headway. Or maybe some other way. I think it's an open question how much this can help, though.