I'm sure it would be a decade-long task. But I'm not sure it's monumentally harder than either the process for piloted aircraft, or the task of automating the aircraft in the first place.

And you can much more easily demonstrate the limits of an automated aircraft. Simulated bird-strikes, thunderstorms, power-outages, emergency landing in fields, blinding by laser-pointers, stray bullets or whatever scenario the FAA throws at you can be done without the risk to the (non-existent) human pilots, just the cost to VCs.

Whether that argument would work on them is a different matter, but I find it quite convincing.