Over 20 years ago when robotics were first being implamented at Dana/Parrish for the production of auto frames, We had problems of (2) different natures.
1) The stamped components varried a fair ammount giving inconsistant fit up.
2) The production operators were of the mind that PEOPLE should be doing the work, and that if this ever caught on, the plant would be full of robots, and the former employees would be SOL.
In order to deal with problen 1, the fixturing was set up to give as tight a joint on (1) side of the part, and THAT side was welded roboticly. The other side was welded by a line welder [person] who could deal with the inconsistant joint gap.
Problem 2 was never completely resolved, but the operators did, to a degree accept the inevitable.
These welding lines produced the same parts all the time for years on end with only minor incramental changes. The robots were Hobart Moto Man units from the late '80s, I would expect modern equipment to have improved substantially.
Consider Joe Kane's points [all of them] but particularly that about lot size and frequency of production. If You run the same jobs repeatedly but in small orders You may stand a chance, but if every job is different and runs are short You have a rough road ahead of You.
Joe had some excellent points about robotics. Another one would be that if your going robotic you go total robotic or you will be maintaining two entirely different systems. Your tolerances have to be much tighter, material lead times will be increased, and if you have a great diversity of parts forget it. If your working to ASTM parts forget it.