Yes especially if you are willing to broaden your definition to include multi-armed or contextual bandits (of which Markov decision processes that RL algorithms solve are a generalization). Lot of internet ad companies etc. use those successfully.