Organizations can increasingly learn from unstructured textual data that is available in, for example, social media or internal databases, by using artificial intelligence above and beyond traditional analysis methods that often prove impractical for handling these voluminous datasets. This paper presents a systematic framework for applying automated techniques to identify topics and code or classify textual data in organizational research. We propose a novel integration of structural topic modeling (STM) for unsupervised topic discovery with large language models (LLMs) for supervised topic classification. We discuss the capabilities and limitations of these two methods and show how to apply the techniques. More specifically, we examine the enhancements in topic discovery granularity provided by STM and the accuracy of topic classification afforded by LLMs. This integrated approach significantly improves the efficiency and precision of classification in large datasets. We validate the methods against manual topic identification and coding in a sample. This comparison is essential to establish the reliability of our findings, providing insights into the complementarity and advantages of our novel approach over traditional methods.
