That is a very interesting problem to try and model. The VRP solver does not have a way to provide different service times based on the route that services it.
I would suggest trying the problem twice. One time with the order service times reflecting the time for a route with a single driver and just routes that have the single driver. The other time with the order service times reflecting the time for a route with more than one employee in the vehicle and the routes that match. You might be able to get some insight from this on the cost benefit of the extra employee.
The other thought I have (completely untested) is to use overtime to make a decision if a route would be better with a second person or not. So the thought is you have a general idea of how much time across a workday could be saved if there were two people in the vehicle vs one. Use that information to set up the routes so they can go past the typical work day by that amount and have overtime start at the end of the typical day. The added cost of the additional driver could be reflected by the overtime (probably not needed to fully model the true cost). Then for routes that use overtime assign those to vehicles with an extra person and routes that do not use overtime assign it to vehicles with just the driver. You could then use the preserve route assignment model and adjust the service times to get a view of the expected workday for each route if that final information is needed.
Would love to hear what you end up trying and how it goes.
