When Enrico Maria Fenoaltea and his colleagues wanted to know which jobs might be most transformed by artificial intelligence, they did something different: they looked at what AI companies are actually building, not just what the technology could theoretically do. By analyzing products from Y Combinator-backed startups using Meta's Llama3 language model and cross-referencing them against the O*NET occupational database, the researchers created something new—an Occupational AI Startup Exposure index, or AISE.
The study, published in PNAS Nexus, offers a more grounded view of AI's potential impact than previous analyses that focused purely on theoretical capabilities. The logic is straightforward: startups that attract venture capital funding have cleared a high bar for economic viability and social appeal. These are products actually making it to market, not mere possibilities.
What the index reveals is both intuitive and surprising. Occupations facing the highest AI exposure include office clerks, data scientists, computer and information systems managers, and market research analysts and marketing specialists—roles built around data processing, analysis, and routine cognitive tasks. At the other end of the spectrum sit athletes, chefs, and construction workers, whose work remains stubbornly grounded in physical skill and hands-on presence.
But the most compelling insight concerns roles where the technology could theoretically step in, yet probably won't—or at least not anytime soon. High school teachers, judges, and marriage counselors all came up as theoretically vulnerable to large language models, yet the AISE predicts much lower real-world exposure for these fields. The reason is social, not technical: people are reluctant to trust AI with work that requires social skills, judgment, or ethically charged decision-making.
"Rather than hitting the entire economy as an indiscriminate technological wave, AI will gradually spread into the economy, with its path shaped by social factors as much as by the technical feasibility of AI applications," the authors write.
The index also flags that occupations requiring a master's degree or higher, significant experience, or high levels of responsibility may face less disruption than purely theoretical analyses would suggest. The technology may be capable of many tasks, but that doesn't mean the market—or society—will accept it in every role.