This year, plans to hire entry-level workers were shelved at a number of companies. In Bank of Korea statistics, youth employment fell by 280,000, and on the day the figures came out the reading converged into a single line. AI had swallowed the jobs, the summary went. There is a bill that summary erases.
Over the same period, employment among people in their 50s rose by 170,000. It is hard to see machines pushing people out as a force that operates by picking and choosing generations. That is why the counterargument that there was no mass extinction came out of the same statistics.
The 280,000 drop in youth employment, before it is something technology did, is closer to the result of companies taking the cost of teaching new hires off the books. New hires leave the company almost no profit in their first year. A trade in which the company absorbs the cost incurred while they learn and gets it back as skill several years later, that was what entry-level hiring was.
When the trade is shelved, the training cost moves to individuals, to schools and to society. Who paid it instead?
The claim that AI takes over what new hires used to do is only half right. Generative AI is a tool that produces a document draft or a fragment of code faster than a person. It does not also produce the eye that recognizes whether the draft is wrong.
That eye develops in someone who has written a hundred drafts and been corrected a hundred times. It develops when the senior who corrects and the new hire who is corrected are in the same space.
The logic on the other side is not weak. Companies that cut hiring did in fact bring in AI tools, and it is also true that the repetitive work once assigned to new hires has shrunk. You cannot invent demand that does not exist and demand that people be hired.
Even so, the fact that 280,000 and 170,000 were recorded in the same period has to be explained. The people who went to the side where employment among those in their 50s rose were workers whose training was already finished.
What companies cut was training rather than labor.
The 280,000 who have been pushed out now will not become the people who answer job postings for experienced workers several years from now. When that time comes, companies will say there are no trained workers, and a considerable share of the 170,000 will reach the age of leaving the field. The price does not disappear. Only the moment of payment is postponed.
If the AI-as-culprit account hardens, the remedies narrow as well. Retraining budgets and course listings become the whole of the response, and the question of who shares the training cost drops out of the discussion. Technology cannot talk back, so it is a convenient defendant.
Companies that are not hiring new workers now will not be able to find trained workers several years from now. Retraining budgets alone will not fill that gap. What has to be decided before 280,000 is recorded again in the next set of statistics is who bears the training cost.
