While it's true that most people don't drive their car most of the time, when they do drive, most of them are driving within a relatively small time window.
I could imagine a 25-50% reduction in the number of cars, but I just don't see 90% happening unless it's accompanied by a huge increase in people working from home or working non-standard hours.
Additionally you are no longer to blame for your commute since the driverless system would be responsible for timing, in much the same way trains are to blame for timing now. I'd also expect the driverless system to have variable pricing - need to get to work by 9am instead 9.15am and it is $5 more out of your pocket. Ultimately your employer is paying that and they have an incentive to only do so if it is valuable.
Being able to just call a car to drive you to work, while an automatic system figures out the car sharing arrangements in real time, could remove a lot of the barriers.
Driverless cars have many possible solutions. A simple one is pricing - it costs $5 more if you refuse to share. (And another $5 more if you want massage seats.) It is possible to completely partition the seats from each other so each one is effectively its own cocoon where you can play your own music as loud as you want and are completely isolated from the other passengers.
Heck it is even possible like many current electrics to have a base chassis that includes the motors and batteries in the floor, and have pods that mount on top. You could have your own personalized pod you keep at home.
It's going to reshape cities and entire industries. We'll get most of the best of cars (flexible mobility and independence) and get rid of most of the worst of them (cost, insurance, waste of public space, car-centric architecture, deaths, drunk driving, etc.).
I cannot wait.
In economics, the Jevons paradox [..] is the proposition that technological progress that increases the efficiency with which a resource is used tends to increase (rather than decrease) the rate of consumption of that resource.