See real-time how AI is restructuring its work

Artificial intelligence takes over some tasks and changes others, but one thing is certain, it is restructuring the job market. Researchers at USC’s Information Sciences Institute (ISI) analyzed LinkedIn job postings and AI-related patent applications to measure which jobs are most exposed and where those changes were first occurring.

The project is led by ISI research assistant Eun Cheol Choi and collaborates with students on the graduate level USC Annenberg Data Science course taught by Luca Luceri, professor of USC Viterbi Research Adcride. The team developed an “AI Exposure” score to measure how closely each role is tied to current AI technology. A high score suggests that jobs may be affected by automation, new tools, or shifts in the completion of work.

Which industries are most exposed to AI?

To understand how exposures have changed in the wave of new innovation, researchers compared patent data before and after major turning points. “We split the patent dataset into two parts, two parts, unsubscribe from the pre- and post-chat releases to see how the job exposure score has changed in relation to fresh innovation,” Choi said. Released in late 2022, ChatGpt caused a surge in the development, investment and patent applications of the generator AI.

Wholesale, transportation, warehouse, information and manufacturing employment were top of the list for both periods. Also, retailers showed early that healthcare and social support had risen sharply after CHATGPT due to new AI tools aimed at diagnosis, medical records and clinical decision-making.

In contrast, education and real estate have consistently exhibited low exposure, suggesting that, at least for now, it is unlikely to be reshaped by current AI technologies.

AI reach depends on role

AI exposure not only varies from industry to industry, but also depends on specific types of work. Jobs such as software engineers and data scientists scored the highest scores as they involved building or deploying AI systems. The role of manufacturing and repairs, such as maintenance technicians, also showed increased exposure due to increased use of AI in automation and diagnosis.

At the other end of the spectrum, jobs such as tax accountants, personnel coordinators, paralegals showed low exposure. They are focused on jobs that AI is difficult to automate. It’s subtle reasoning, domain expertise, or dealing with people.

AI exposure and pay don’t always move together

This study examined how AI exposure relates to pay. In general, employment exposed to current AI technology is associated with higher salaries and may reflect the demand for new AI skills. This trend has been most powerful in the highly exposed and well-compensated information sector for software and data-related roles.

But in sectors such as wholesale trade and transportation and warehouses, the opposite was true. Jobs with higher exposure in these industries tended to offer lower salaries, particularly at the highest exposure levels. Researchers suggest that this could indicate an early impact of automation. Automation is beginning to replace workers rather than AI increasing workers.

“In some industries, there may be a synergistic effect between workers and AI,” Choi said. “Other people might point out competition or exchange.”

From class projects to ongoing research

The contrast between the industries in which AI complements workers and the industries it may replace them is something the team will look into further. They hope to differentiate between different types of impact (automation and augmentation) and build on their framework by tracking the emergence of new occupations driven by AI. “This kind of framework is exciting,” Choi said. “Because we can capture these signals in real time.”

Luceri highlighted the value of practical research in the classroom. “It is important to give students the opportunity to address relevant impact issues that allow students to apply the theoretical tools they have learned to the actual data and questions,” he said.

The paper, “Mapping of labor market vulnerabilities in the age of AI: Evidence from recruitment and patent data,” was co-authored by students Blue Cao, Qi Guan, Shengzhu Peng, and Po-Yuan Chen, and was presented at the International AAAI Conference on Web and Social Media (ICWSM 2025).