Analysis
Twenty-two percent of U.S. college students have already changed their major or concentration because of job-market concerns, and 69% believe AI will make finding work harder, according to a survey of 2,000 students reported by Axios on August 15.
The individual stories in the data are more specific than the aggregate. Burton Baecker, a rising sophomore at UC Davis, started in accounting and moved on after concluding AI would reshape the field. That is a rational response to what is actually happening in entry-level accounting, legal review and junior software work -- the tasks most exposed are precisely the ones that historically justified hiring a 22-year-old.
The employer side of the survey points somewhere less obvious. Leah Belsky, OpenAI's VP of education, framed the in-demand skills as critical thinking, AI literacy, leadership and collaboration -- transferable capabilities rather than domain knowledge. American University and others are folding AI literacy into curricula alongside communication and teamwork. Whether that is genuine signal or an industry telling schools what it wants to hear is not resolvable from a survey.
“But attributing that entirely to AI conflates two things: post-ZIRP headcount discipline started in 2022, well before agentic coding tools were usable in production.”
The economic backdrop is real. Entry-level tech hiring has been soft since 2023, computer science enrollment growth has flattened after a decade of expansion, and the Bureau of Labor Statistics has repeatedly revised software-developer growth projections. But attributing that entirely to AI conflates two things: post-ZIRP headcount discipline started in 2022, well before agentic coding tools were usable in production. Correlation here is doing a lot of work that causation has not earned.
The counterweight worth holding onto: every prior automation panic produced the same survey result and a different labor market than the one predicted. Students abandoning technical majors en masse would, if it persisted, produce a shortage of exactly the people needed to build and audit these systems -- which is how wage premia get created, not destroyed.
For founders, the near-term consequence is a hiring market where junior candidates arrive with more AI fluency and less domain depth than any cohort before them. That is a training-cost problem, and it is showing up now.
There is a distributional detail the aggregate hides. The students most able to switch majors are the ones with academic flexibility and financial slack; students on scholarship tracks, transfer pathways or accelerated programs often cannot. A statistic that reads as a market adjusting is also a statistic about who gets to adjust.
Institutions are responding at different speeds. American University is integrating AI literacy into curricula alongside communication and teamwork, and several large public systems have added AI requirements to general education. Curriculum change moves on a two-to-four year cycle, which means the students making decisions today are choosing against a job market that will have moved again before they graduate. That mismatch, not the technology itself, is what makes the 22% figure worth taking seriously.
Compare the survey against what hiring data actually shows. Entry-level technical postings have been soft since 2023, but the decline began with post-2021 headcount corrections and rising rates, well before agentic coding tools were production-grade. Attributing the full drop to AI is a causal claim the data does not currently support, and it matters because the policy responses -- curriculum overhauls, retraining programs, degree redesign -- are being justified on that attribution.
