RECONFIGURING WORK IN THE AGRIFOOD CHAIN

Authors

Abstract
Shifting towards a circular economy requires a comprehensive and systemic overhaul of job functions and the associated skill sets, ensuring alignment with the core principles of resource efficiency, waste reduction, and sustainable value creation. This shift transcends traditional industrial boundaries, calling for the seamless integration of circular strategies across a wide range of economic sectors. Among these, the agri-food sector stands out as a pivotal arena for change due to its multifaceted role. Realising this transformation demands a workforce equipped with highly specialised and flexible competencies capable of driving innovation, implementing circular economy solutions, and effectively coordinating circular strategies throughout the entire food supply chain. Such capabilities encompass a robust understanding of interdisciplinary methods, advanced technological applications, and the ability to navigate complex sustainability challenges. Despite the recognised importance of the agri-food sector in advancing circularity on a broader scale, previous research has largely treated circular economy-related competencies or agri-food- specific skills as isolated areas of inquiry, often failing to address the full supply chain in a cohesive manner. Moreover, while an expanding body of literature examines the skills needed for circular economy adoption across various industries, to the best of our knowledge, no study has systematically and holistically mapped the competencies essential for accelerating circularity in the agri-food system. Accordingly, this study addresses that gap by applying an advanced topic modelling and semantic analysis framework to a curated dataset of 7,943 job advertisements collected between June and October 2024 from leading recruitment platforms and corporate career pages across the United States. The analysis focuses exclusively on job postings within the agrifood domain, aiming to uncover the most salient employability skills and their latent thematic groupings. The methodological approach is grounded in natural language processing and unsupervised machine learning. Job descriptions were pre-processed through lemmatisation, tokenisation, and stopword removal, followed by the generation of dense sentence embeddings using a transformer- based large language model. These embeddings served as the input to Bidirectional Encoder Representations from Transformers (BERTopic), a neural topic modeling algorithm that integrates semantic representations with dimensionality reduction and density-based clustering. The findings confirm that employability in the agrifood sector is increasingly defined by a hybrid profile of competencies. The methodological contribution of this study lies in the integration of large language models (LLM)-based text mining with interpretable topic modeling and correlation logic, offering a replicable and scalable approach to skills profiling in sectors undergoing technological and environmental transformation.