A company aims to enhance the response accuracy of its generative AI application, which utilizes a foundation model (FM) on Amazon Bedrock. What is the most cost-effective solution to achieve this?