Modeling AI Automation Potential in Knowledge-Intensive Media Work

Previous studies of artificial intelligence (AI)-driven automation have mostly assessed entire occupations, providing limited evidence about differences among individual tasks. This study addresses this limitation by proposing an NLP-assisted task-level assessment framework for estimating the automation potential of journalistic work. Of the 30 task descriptions listed for the O*NET occupation “News Analysts, Reporters, and Journalists” (27-3023.00), 25 were retained following scope-based screening and grouped into ten composite tasks representing the focal news-production workflow. Each task was assessed across three dimensions—Verb Type, Social Dependence, and Skill Type—using NLP-assisted feature extraction and structured consensus-based evaluation. The resulting component scores were combined into a composite automation potential score ranging from 0 to 1. Editing visual content and filing stories obtained the highest estimated scores (0.97), followed by gathering background information and collecting media content (0.77). Arranging interviews (0.20), conducting interviews (0.23), and receiving assignments to develop story ideas (0.27) received the lowest estimated scores. The study contributes a transparent and interpretable proof-of-concept approach for operationalizing task-level automation potential across procedural, social, and skill-related dimensions. The resulting scores represent relative task-level estimates rather than observed rates or probabilities of workplace automation. With larger labeled task datasets, the three dimensions defined in this framework could provide a basis for supervised classification and subsequent predictive modeling of automation potential across broader task collections.