Quantum computing does something completely and specifically different from classical computing. I am still earning my depth in quantum computing - BUT - where quantum would come in handy is solving NP-hard problems that classical computers are bad at. Vehicle routing (traveling salesman, anyone?), cryptography cracking, protein folding, etc etc, are all well-suited. Even, potentially, in setting the weights and biases in a neural network for optimal performance and throughput - rather than running billions of cycles of gradient descent to rule out local maxima and minima to find the system min/max for the most generally suitable solution.
These are problems in which the solution is easily verified, but it's extremely difficult to solve using classical computing methods. Quantum computing promises to find the right parameters to make the math work the best. But you still need something to actually DO that math, and that's where classical computing will remain.
Broadly speaking, "AI" tries to guess or approximate the best answer to a response, and it informs that guess based on the data it was trained on iteratively over time. "Quantum computing" simply collapses all possibility spaces in O(1) time such that only the best response still exists and is not obliterated.
That is a massive oversimplification, and quantum computing is (as far as I know) not anywhere close to general utility to actually set up a quantum system to approximate a complex real system. But as a student of computer science if I were to stick my finger in the air and guess where we are on the timeline we're closer to UNIVAC than we are to x86 in terms of quantum computing.