Assume that it's not perfect, and you have about 2 weeks of downtime a year, plus 2 additional days of down time a week, just for service, small issues, etc.
So you'd have to buy a few to make it work, so their downtime didn't too significantly overlap, but that's $8/hr40hrs/wk50wk/yr*3shifts = $48000 per year.
So if it essentially does the work of 3 people (120 hours per week), then you can come out ahead on replacing one station with a robot (assuming you have 3-4, so downtime doesn't totally kill production).
Edit: I was thinking of something like this - http://en.wikipedia.org/wiki/Baxter_(robot)