Applications of Large Language Models (LLMs) in Business Analytics – Exemplary Use Cases in Data Preparation Tasks

The application of data analytics in management has become a crucial success factor for the modern enterprise. To apply analytical models, appropriately prepared data must be available. Preparing this data can be cumbersome, time-consuming, and error prone. In the current era of Artificial Intelligence (AI), Large Language Models (LLMs) like OpenAI’s ChatGPT offer a promising pathway to support these tasks. However, their potential in enhancing the efficiency and effectiveness of data preparation remains largely unexplored. In this paper, we apply and evaluate the performance of OpenAI’s ChatGPT for data preparation. Based on four real-life use cases we show, that ChatGPT demonstrates high performance in the context of translating text, assigning products to given categories, classifying sentiments of customer reviews, and extracting information from textual requests. The results of our paper indicate that ChatGPT can be a valuable tool for many companies, helping with daily data preparation tasks. We demonstrated that ChatGPT can handle different languages and formats of data and have shown that LLMs can perform multiple tasks with minimal or no fine-tuning, leveraging their pre-trained knowledge and generalization abilities. However, we have also observed that ChatGPT may sometimes produce incorrect outputs, especially when input data is noisy or ambiguous. We have also noticed that ChatGPT may struggle with tasks that require more complex reasoning or domain-specific knowledge. Future research should focus on improving the robustness and reliability of LLMs for data preparation tasks, as well as on developing more efficient and user-friendly ways to deploy and interact with them.