This talk presents two complementary routes toward autonomous beam control in particle accelerators. One explores quantum reinforcement learning as a longer-term research frontier. The other demonstrates a concrete engineering application of Bayesian optimisation to Multi-Turn Extraction in the CERN PS.