Reimagining work

Abstract

Organizations are shifting from a job-based to a task-based organizational structure (Chalutz-Ben Gal, 2023). This increased task focus requires assessing who would be fit for a task directly by utilizing person-skill fit instead of more vocational perspectives such as person- job fit (Chalutz-Ben Gal, 2023). Allocating tasks to suitable employees currently takes a significant amount of time for human research professionals (Bouajaja & Dridi, 2017). This paper explores how natural language processing (NLP) models based on neural networks can support people within organizations in efficiently identifying the most suitable individuals for specific tasks. A prototype system was developed and tested on a synthetically generated dataset of resumes based on the O*NET framework (National Center for O*NET Development, 2024b), to automate the process of allocating tasks to candidates. This was tested by utilizing large language models (LLMS), which proved unsuitable to accurately assess large amounts of resumes within a short amount of time. Vector embeddings were also tested to rank resumes based on person-skill fit. A quantitative analysis has shown a strong correlation between the ranking and the ability to perform the tasks (ρ = -0.7505). Domain experts who tested the prototype expressed satisfaction with its ranking and user-friendly design, emphasizing its potential to streamline HR processes and enhance efficiency. However, the reliance on synthetic data, must be addressed to confirm usability in real-world scenarios.