If there is any initial step, my guess is that it would be a single pilot backed up by a dual-rated pilot/dispatcher, or maybe both a pilot and a dispatcher, on the ground. The pilot would have a certain number of flights to monitor, and would use high bandwidth comms only for takeoff, landing, and in the event of an emergency where the pilot flying needs assistance and a pilot monitoring. The comm backup would have to be a full-cockpit equivalent in order to monitor all switch positions and caution/warning annunciators, plus full video. The ground pilot monitoring would have the ability to make autopilot inputs, change switch positions, fly the plane remotely, and set up and fly an auto-land approach and landing, in the event that the pilot flying is incapacitated.
But... It'll only be in a backup role, not as PIC, and the ground pilot monitoring would probably end up assigned to multiple flights, hopping between them to actively shadow and monitor critical phases of flight.
I could see them try to do that... It's more technologically feasable than having every plane flying around without any pilots at all, and it directly addresses the phases of flight where 2 pilots are required or desired. Since each plane is shadowed only during certain flight phases like takeoff, approach/landing, and emergencies, they would need only a fraction of the comm bandwidth and ground pilots.
During non-critical phases of flight, dispatchers will continue to monitor as usual.
I'm hoping they don't but that's the scenario that I see as a possibility. It's based in part on how the military splits apart the launch/recovery and mission elements in RPA operations, and how the military would like mission pilots to control more than one RPA during mission operation and en-route cruise phases of flight.