Upgrades of Large Language Model Products and Firm's Strategy: The Role of Quality Improvement

Products based on Large Language Models (LLMs) can leverage consumers' consumption data to upgrade to an advanced version with improved overall performance through fine-tuning technologies. This paper highlights the significant role of consumers' consumption data in enhancing the quality of LLM-powered products, contrasting this to the traditional use of data for learning consumer preferences. Specifically, our study explores a firm's pricing decisions when consumers' consumption data of the basic product version can be used to develop a higher-quality version and meanwhile reveal their preferences. The key tension for the firm is to balance maximizing the profit of the basic version and increasing the consumption quantity of the basic version for a higher quality of the advanced version. Our paper shows that when the diminishing marginal utility from consumption is low (high), the firm offers the basic version of the product for free (at a positive price). Interestingly, our results show that quality improvement (rather than preference learning) through consumers' consumption data can yield an "All-Win" situation in which all stakeholders benefit from the upgrade of the LLM-powered product, highlighting the importance of fine-tuning techniques on consumption data. Managerial implications are discussed.