In the rigorous landscape of product development, sensory panel testing plays a critical role in ensuring that products align with consumer expectations before market launch. Traditionally, this process involves qualitative evaluations where panelists provide free-form comments reflecting their sensory experiences alongside more formal quantitative testing. Additionally, there are many other sources of such qualitative free-from comments, for example surveys and product reviews. However, the subjective nature of these comments poses challenges in quantifying and systematically analyzing the feedback, which is crucial for identifying product pain points and guiding reformulations. To address this gap, we developed a model utilizing the RoBERTa language processing model to predict quantitative sensory scores from free-form panelist comments. This approach leverages Generative Pre-trained Transformer AI (GenAI) technology, enhancing the traditional sensory evaluation by providing a scalable method to interpret qualitative data objectively. The implementation of this model allows for the nuanced understanding of consumer sentiment, facilitating more informed decision-making in product formulation and optimization. This study outlines the model's architecture, its integration into the sensory evaluation workflow, and discusses the implications of automating sentiment analysis in reducing biases and increasing the efficiency of product development cycles.