Artificial intelligence is set to automate certain tasks and alter others, but one thing is clear: it’s changing the landscape of employment. To track which roles are most affected and where these shifts are beginning, researchers at USC’s Information Sciences Institute (ISI) examined LinkedIn job listings and AI-related patent filings.
ISI research assistant Eun Cheol Choi spearheaded the project in collaboration with graduate students from a USC Annenberg data science course taught by USC Viterbi Research Assistant Professor Luca Luceri. The team developed an “AI exposure” score to assess how strongly each occupation connects to existing AI technologies. A higher score signals that automation, new tools, or changes in work processes are more likely to influence the role.
Which Sectors Are Most Vulnerable To AI?
To track how exposure evolved alongside new waves of innovation, the researchers analyzed patent data from before and after a key milestone. “We divided the patent dataset into two segments—pre- and post-ChatGPT—to observe how job exposure scores shifted with recent advancements,” explained Choi. The release of ChatGPT in late 2022 sparked a boom in generative AI development, investment, and patent activity.
Across both timeframes, roles in wholesale trade, transportation and warehousing, information, and manufacturing ranked highest in exposure. Retail also had high exposure early on, while healthcare and social assistance saw a notable increase after ChatGPT’s debut likely driven by emerging AI tools for diagnostics, electronic health records, and clinical decision-making.
Meanwhile, sectors like education and real estate consistently showed low levels of exposure, suggesting that current AI technologies are, for now, less likely to significantly alter them.
AI’s Impact Varies By Job Function
AI exposure isn’t solely determined by industry it also hinges on the nature of the job itself. Positions such as software engineer and data scientist received the highest exposure scores, as they often involve creating or implementing AI technologies. Similarly, roles in manufacturing and repair, like maintenance technicians, showed higher exposure due to AI’s growing role in automation and diagnostics.
On the flip side, jobs like tax accountant, HR coordinator, and paralegal had low exposure scores. These roles typically require complex reasoning, specialized knowledge, or interpersonal interaction tasks that are more challenging for AI to replicate.
AI Exposure and Salary Don’t Always Go Hand In Hand
The study also explored the connection between AI exposure and wages. Overall, roles with greater exposure to current AI technologies tended to come with higher pay, likely due to the growing demand for AI-related skills. This pattern appeared most clearly in the information sector, where software and data-driven jobs faced both high exposure and high pay.
However, the trend reversed in industries like wholesale trade and transportation and warehousing. In these sectors, highly exposed jobs often came with lower wages particularly at the top end of the exposure scale. Researchers believe this could reflect early signs of automation displacing workers rather than enhancing their roles.
“In some fields, AI and human workers may complement each other,” said Choi. “In others, it may indicate competition or even replacement.”
From a Classroom Assignment To a Continuing Research Initiative
The team plans to delve deeper into the contrast between industries where AI enhances human roles and those where it may replace them. They aim to expand their framework by differentiating between types of AI impact such as automation versus augmentation and by monitoring the rise of new job categories shaped by AI. “What’s exciting about this framework,” Choi noted, “is that it allows us to capture these changes in real time.”
Luceri highlighted the importance of experiential learning: “It’s crucial to give students opportunities to tackle real and meaningful challenges, applying the theoretical concepts they’ve studied to actual data and practical questions,” he said.
The study, titled “Mapping Labor Market Vulnerability in the Age of AI: Evidence from Job Postings and Patent Data,” was co-authored by students Qingyu Cao, Qi Guan, Shengzhu Peng, and Po-Yuan Chen, and presented at the International AAAI Conference on Web and Social Media (ICWSM 2025), held June 23–26 in Copenhagen, Denmark.
Read the original article on: Tech Xplore
Read more: Unexplained Radio Signals Under Antarctic Ice Challenge Physics Norms