Imagine trying to predict the success of a complex industrial process, like coating cutting tools, with some ingredients being numerical (temperature, pressure) and others, well, simply names or labels. That's the challenge researchers tackled in a recent paper on industrial process modeling. Traditionally, these categorical variables, such as reactor types or product names, are converted into numerical representations using methods like one-hot encoding. However, this approach often overlooks the meaning and relationships between categories, limiting the potential for in-depth analysis. This is where Large Language Models (LLMs) step in. The researchers found that LLMs excel at creating rich numerical representations (embeddings) of these categorical variables, capturing their true essence. By understanding the relationships between different types of reactors or product categories, LLMs unlock insights previously hidden within the data. These embeddings are then used to train a machine learning model to predict the thickness of the coating on cutting tools. The results are impressive. LLMs significantly improve the accuracy of the model compared to conventional methods. Furthermore, they reveal which process parameters truly matter, information that was previously obscured. This breakthrough allows engineers to understand the process better, leading to optimization strategies and improved quality control. Beyond thickness prediction, this research opens doors to leveraging textual descriptions and expert knowledge commonly found in production logs, which were difficult to integrate into data-driven models before. The implications for industrial processes are vast, from understanding the impact of different machine types on output to assessing the nuances of various production steps. While this research demonstrates the power of LLMs in industrial settings, challenges remain. Exploring advanced sensitivity analysis and uncertainty quantification methods is the next step, along with testing different machine learning frameworks to further refine the approach. This is just the beginning of LLMs’ impact on industrial AI. By transforming text-based insights into meaningful data, LLMs promise to revolutionize how we understand and optimize industrial processes.