Fast, except we have no idea how to measure it.
Selective breeding is powerful.
In the past few thousand years, we have managed to turn a single species -
the wolf - into a range of animals incredibly different from one another. Natural evolution is slow because it has no direction, but if you know what you're aiming for, you can breed a single trait into pretty much whatever you want.
The problem with breeding a species to become intelligent (in the sense of human-like intelligence) is that we don't really understand what intelligence is, let alone how it works.
It's one thing to measure the legs of a puppy, raise and breed the dogs with the shortest legs and the longest bodies together and make a dachshund, but we can't even figure out how to properly measure the intelligence of our own species. How could we measure it in another?
Border collies are considered to be the most intelligent dogs, but they are intelligent in the way we made them to be - they are good at following directions and herding sheep. Ultimately, our ability to breed an animal for a trait is dependent on our ability to measure that trait, and the foundations of true sapience are still too much of a "black box" to measure properly.
It could be that some of the vital steps to creating sapience are qualities we wouldn't be testing for. For example, social behavior is now considered to be an important part of the root of intelligence, but most of our "intelligence tests" do not test for social ability. There is also the issue that as an animal grows smarter, it may become less interested in doing silly tests for humans, and may get lower scores as a result.
How would one measure whether, say, any given rat was closer or further away from sapience? We could breed them based on how quickly they can run a maze, but we don't really know if maze-running has anything to do with the basis of human intelligence. Ultimately we'd wind up with rats who were good at running mazes, but probably no closer to thinking like a human.