I don't know much about this issue and I expect someone like veen or kleinbl00 to come in here and fact check my ass straight to hell, but by gut instinct alone, I call bullshit.

If a 20-person vehicle needs service, that's a big hole in the service stream.

If a 2-person vehicle needs service, the gap in service is much smaller and has far less impact on the flow of people from A to B. (And it is cheaper to dispatch another 2-person vehicle in its place.)

It's cheaper to build and maintain one vehicle that can hold 20 people than it would be to build 10 vehicles that can hold 2. Better for the environment, better for cost of use.

Automated vehicles will also have much smaller following distances, because they will know what is happening up ahead, far in advance, and can maneuver safely in much smaller gaps. That means they have smaller bumper, less need for airbags and monster braking systems, skinny tires, and less of all the other shit associated with protecting soft meat-people from getting t-boned at 60 MPH by a delivery truck.

Closer following distances also allow for less energy exertion to move forward, because the slipstream of the previous vehicle is occupied by the following vehicle.

And finally, a chain with many small links is far more flexible and dynamic than a chain with huge links. It's rail cars vs glorified barco-loungers. Big massive heavy things that start, stop, and turn slowly, versus... well... basically, go-karts running on "rails" of high-speed data from multitudes of sensors.

100 vehicles on the road that can hold 20 people will take up less space than 1,000 vehicles on the road that can hold 2 people. There will be more space in between vehicles for reaction time and safety for the devices and traffic will be less congested. You don't have to worry about "tight" and "flexible" chains which points out to a flaw you brought up yourself.

Honestly? The biggest problem will be the HUMANS inside the autonomous vehicles freaking out because they are following each other so closely. It will be perceived by the meat-person to be dangerously close following, because the meat-person is incapable of perceiving or processing the stream of data the autonomous vehicle is assessing to determine safe following speed and distance.

Once again... people are the problem! :-)

Which will make it a hard sell. Perception of quality, safety, etc. are part of the whole customer experience. In the car world, they have an acronym called NVH, which stands for Noise, Vibration, and Harshness. Car designers literally spend hundreds of development hours to make sure cars are comfortable to their customers from everything from muffling road noise and engine sounds behind the firewalls to the closing of doors and trunks and how fucking seatbelts sound. Car companies are going as far as to pump engine noise in through speakers because modern engines are considered too quiet. People are particular as shit.

I mean, with digital cell phones, we're actually able to have better call quality than we usually do, but our calls are literally degraded on purpose because people have an expectation for their phone calls to sound like they would if they come from a land line. Digital networks have been out for a decade and a half at this point. You'd figure people would get on the ball, but no, they won't, cause it's about consumer perception.