A report on generative AI and employment written by Cheon Kyung-rok, an economic analyst at the National Assembly Budget Office, has reached a conclusion contrary to the common view. Since November 2022, when ChatGPT appeared, the number of workers in occupations with high AI exposure has increased across all age groups, and the same held for young people. Regression analysis also turned up no evidence that AI had cut employment in highly exposed occupations. The higher the exposure, the more the rate of employment change actually rose, at a statistically significant level.
The exposure measure was taken from an index created by the International Labour Organization. It scores how much of a job's tasks generative AI can handle, and on that basis the report sorted Korean occupations into four groups, high exposure, medium exposure, low exposure and no exposure. The high exposure group included accounting and bookkeeping clerks, finance and insurance clerks and professionals, customer service and monitoring agents, and computer and software developers. The analysis period is understood to run from the release of ChatGPT to the first half of last year.
A closer look at the same report shows the findings do not all point in one direction. That is because results split within the high exposure group as well. The number of workers fell among accounting and bookkeeping clerks, customer service and monitoring agents, and writers and media professionals. It rose, in contrast, among computer systems and software specialists, finance and insurance specialists, and human resources and management specialists.
For the decline in the writing and media occupations, the report treated the fact that their work is effectively the same as the text generation large language models perform as a clue for interpretation. It read the professional occupations differently. It raised the possibility that, for junior staff too, AI works as a form of augmentation that assists rather than takes away their work. The employment that fell among young people was in the medium exposure occupations.
Labor market indicators, though, show the opposite face. The National Data Office's January employment trends put the employment rate for those aged 15-64 at 69.2%, up 0.4 percentage point from a year earlier. The employment rate for those aged 15-29 in the same month was 43.6%, down 1.2 percentage points, and it has been falling for 21 months straight. Youth unemployment jumped from 6.0% to 6.8%, the highest for a January since 2021.
Young people who are neither working nor looking for work, counted as "resting," numbered 469,000, up 35,000 (8.1%) in a year. What job seekers describe on the ground fits these figures. Entry-level applicants who already have experience, so-called "used newcomers," are preferred, and there has been a case of a large company cutting its open recruitment for new hires from twice a year to once. Companies are leaning toward rolling and experienced hiring instead of open recruitment.
Placed side by side, the two sets of data illuminate different layers rather than collide. Even if the total number of workers in an occupation holds steady or grows, the passage for new entrants into that occupation can narrow. Accounts that Big Tech has turned conservative on hiring junior developers and that large law firms and accounting firms are adjusting the scale of their hiring point away from totals and to what is happening at the entrance. There is also the point that macro productivity indicators have not clearly improved even while companies have brought in AI and cut hiring.
Earlier studies have pointed to "seniority-biased technological change," in which youth employment falls and employment of people in their 50s rises in industries highly exposed to AI. The story is that AI pushes up the productivity of people with experience, while the opportunity to gain experience itself disappears. Kristalina Georgieva, managing director of the IMF, warned that AI will strike the labor market like a tsunami, but the NABO analysis found no signal in job openings and youth hiring trends that clearly sets these occupations apart from others. The gap between the warning and the measurement is closer to the reality at this point.
On the manufacturing floor the arithmetic shows up more bluntly. Hyundai Motor Group is pushing to put the AI robot "Atlas" into industrial sites, and Lee Woong-jae, senior vice president of the Manufacturing Solution Division, said on the 7th of last month that it would be applied first to dangerous work and work with a heavy physical burden. The expected price is around 200 million won per unit, and on the assumption that it does the work of two employees, the calculation attached is that it pays for itself in two years. The union said on the 16th of last month that robots could replace people and opposed their introduction without an agreement between labor and management.
Where policy should take aim changes accordingly. Measures to protect jobs overall and measures to build a path for people entering society to gain their first practical experience are designed in entirely different ways. How to redesign training for new hires in occupations with a large share of repetitive work, such as accounting and customer service, is the practical task these data pose.
For a job seeker, this report provides neither grounds for relief nor grounds for giving up. In the development, finance and human resources occupations that have shifted toward using AI, more people have been needed, and occupations where producing text was the whole of the work have shrunk. Whether the statistics of the next 1-2 years will draw the same picture is unknown. Yet for young people facing a narrower door to a first job, there is not much time to wait for the statistics to turn.
