Labor automation potential of generative artificial intelligence model in Mexican occupations

Abstract

The widespread use of multimodal generative models such as GPT-5 threatens to replace mental tasks instead of physical ones, sparking debate about their impact. The objective is to propose the GPT Multimodal Automation Indicator (GMI) as a metric to measure the automation potential of occupations, using a task-based approach. The GMI was applied in Mexico and constructed using the 2019 SINCO data, focusing on the tasks of unitary groups. The World Economic Forum (2023) methodological criteria were applied to calculate the potential percentage of GPT-5 automation per task, and the results were averaged across unitary groups and divisions. The results show that the divisions of managers and chiefs, administrative assistants, and merchants and sales have low to medium automation potential, with 38%, 32%, and 28%, respectively. The professional and technical, services and surveillance, as well as agricultural activities divisions, have low automation potential, with percentages of 24%, 17%, and 11%, respectively. However, there are tasks with a high automation potential of 88%. Thus, the effect of GPT-5 is to complement occupations. However, there is a high potential risk of automating some tasks. Consequently, the greater the number of highly automated tasks within an occupation, the higher the risk of job displacement. Finally, a limitation of this study is that IGAM is focused on GPT-5, raising the need to expand its scope to more AI models. The originality lies in the fact that IGAM evaluates the potential impact of generative AI and its exposure.