Towards Optimizing the Costs of LLM Usage