Well, why does an AI even need to be sentient to be able to replace a human? Self-awareness is not a prerequisite for intelligence, it's just a strange little oddity that mammals happened to develop. There are already AIs that can beat the best human chess players, and the only thing they do is heuristics. Why can't that happen to all fields? So, if we have nonthinking machines that just run through some really complex algorithms to get results, why give them rights?
First - I say they need sentient simply because there's a few intelligence-only fields (unless they somehow manage to figure out the maths for it too - then I would agree). Like media. I highly doubt that a machine that is not intelligent will be able to create the imperfect yet lovable music, video and other sensory experience that we can. Hell - without advanced sensors, certain things are even completely out of reach, like experimental cooking. Chess is entirely mathematics - so are most things, most likely, but the mathematics of imperfection are most likely insanely complex - and by the time a machine could understand them, I'm not certain it will not have intelligence anyway. And even if they are non-thinking, technically - they still behave, and as such they would deserve at least as much rights as the animals we live with.