Connect with us

AI

The Bank of Korea Maps a Missing Career Rung

The Bank of Korea finds 94% of 285,000 youth job losses in AI-exposed work, and says the real damage is a broken career ladder.

Published

on

The Bank of Korea counted 268,000 youth jobs gone from AI-exposed industries between June 2022 and June 2026, or 94% of the entire drop among people aged 15 to 29. Workers in their 50s added 230,000 jobs in that window, and 173,000 of those gains sat in the same high-exposure industries.

The central bank still would not treat that split as proof that AI caused the slump. It said the technology is speeding up a career ladder that was already shrinking, as firms lean on staff who already know the work.

Did AI Cause Korea’s Youth Job Losses?

No, not as a standalone shock, according to the Issue Note posted in English on September 3 as “Who Moved the Career Ladder? AI and Youth Employment in Korea.” Youth jobs fell 285,000 from June 2022 to June 2026. High-exposure industries accounted for 268,000 of that decline. Information services were down 31.4%, publishing 27.4%, computer programming, system integration and management 16.6%, and professional services 11.6%.

The researchers used National Pension subscriber files covering about 16 million workers, plus the Economically Active Population Survey. They grouped industries by AI exposure and compared the ChatGPT-era window with earlier years. The pattern matches work Jinsu Han and Samil Oh published in December 2025 on seniority-biased technological change in early AI diffusion, when youth jobs had already fallen 211,000 over three years, 208,000 of them in high-exposure industries.

Oh, who leads the Bank of Korea’s employment research team, put the limit in plain language after the August 18 release.

AI can increase the productivity of young people to a large extent. This also means they can be replaced by AI. However, we cannot say that AI is entirely behind the reduction in youth employment. Instead, AI is accelerating the trend of the diminishing career ladder for young people.

Samil Oh, Bank of Korea employment research team

The note lists other forces in the same breath: the unwind of pandemic-era over-hiring, a stronger taste for experienced recruits, thinner in-house training, and remote work. Population decline also weighs on the youth counts. AI, in this telling, is an accelerant sitting on top of hiring habits Korea already had.

Junior Work Was the Training Year

The jobs that moved first were the ones new hires used to learn on. Drafting, basic analysis, search, translation, and simple coding are the tasks Oh has described as manual work AI handles well. They were also the tasks that used to justify a first salary while a junior watched how a team actually decides.

That is why a programming shop can look busy and still stop hiring graduates. A senior who already knows the client, the codebase, and the unwritten rules can ship more with a model on the desk. The cheap extra pair of hands becomes optional. The optional pair of hands was the apprentice.

People who still write code after hours describe the same swap in the tasks they now hand a model: connecting APIs, repeating CRUD, cleaning documents, spinning up tests. Those used to be the chores a senior passed down the desk. They are not being supervised by a junior. They are disappearing as a teaching load. The first year of a white-collar career was built on exactly that load.

Inflows tell the same story as outflows. Average monthly hiring of young workers into high-exposure industries fell from 32,600 in the 2016 to 2019 window to 29,100 from July 2022 to June 2026, about an 11% drop. Outflows from those industries into unemployment rose from 3,700 to 4,900, or 32.4%. Low-exposure outflows rose only 6%, from 6,600 to 7,000. The Bank of Korea’s own English summary says the problem is no longer only the first offer. Young staff also have a harder time staying in work long enough to pile up skill.

Fifty-Somethings Filled the Same Industries

If AI had simply killed the work, older payrolls in the same industries would have shrunk too. They did the opposite, which is the first cut of the same pension tally that made the seniority pattern obvious.

THE AGE SPLIT IN NATIONAL PENSION RECORDS

Group Job change, June 2022 to June 2026 Share inside high AI-exposure industries
Youth, 15 to 29 down 285,000 94% of the losses (268,000 jobs)
Workers in their 50s up 230,000 75.2% of the gains (173,000 jobs)

Han and Oh’s 2025 note already framed the split as a feature of the tools. AI replaces codified, textbook tasks that juniors do. It helps with tacit knowledge, context, and people skills that sit with staff who have been in the building longer. The 2026 update says that split held. Industries that use AI to automate shed more young staff. Industries that use it to assist people did not show the same youth decline.

HOW THE TWO BOK NOTES LINE UP

Measure 2025 note, three years 2026 note, June 2022 to June 2026
Youth jobs down 211,000 down 285,000
Of which high AI-exposure 208,000 (98.6%) 268,000 (94%)
Jobs for people in their 50s up 209,000 up 230,000
Of which high AI-exposure 146,000 173,000

The later note covers a longer stretch, so the extra losses are not a clean extra year. Even so, the AI share of the youth decline eased from 98.6% to 94% as the total grew. High-exposure work is still where most of the damage sits. A larger slice of the newer losses sits outside that bucket, which fits the bank’s caution that hiring habits, training cuts, and demography are moving too.

The Degree Premium Flipped After ChatGPT

Korea spent a generation pushing students toward four-year degrees and white-collar specialties. Those are the jobs with more AI exposure. From January 2019 to October 2022, the average unemployment rate for college-educated youth was 8.2%, against 8.0% for youth with a junior-college education or less. Since November 2022 the averages are 7.0% and 5.4%. Both rates fell. The gap opened from 0.2 percentage points to 1.6 because the less-credentialed group improved more.

That reversal is confined to the young in the Bank of Korea’s cut. Among middle-aged and older adults, unemployment by education barely moved after ChatGPT’s launch. A four-year diploma still correlates with the kind of desk work a model can draft. A junior-college or secondary track more often leads to work the model does not yet do.

The 2025 annual print for youth unemployment was lower than the mid-2026 monthly spikes. Statistics Korea’s indicator series puts the annual youth unemployment rate of 6.1% for 2025, against an OECD average of 11.2% in 2024. The July 2026 monthly rate for ages 15 to 29 was 6.8%, up 1.3 points from a year earlier. The annual figure and the monthly print are different series. Together they show a country whose youth jobless rate is not high by rich-country standards and is still getting worse at the entry door.

Who Will Be Senior in Ten Years?

The 50-somethings adding jobs this year are not a renewable resource. They became usable to a model because they already had a decade of messy, supervised work. If firms stop buying that first decade, the stock of people who can direct a model will thin out as this cohort ages out.

The 2025 note said as much. A slide in youth hiring, Han and Oh wrote, could weaken the future talent pipeline, which is why they would not treat the early drop as a permanent law. Businesses might yet decide that growing people beats a short-term headcount cut. The 2026 follow-up found the same seniority bias still in the pension files. The pipeline bet has not been placed.

U.S. payroll data now rhyme with Seoul’s administrative files. Stanford’s Digital Economy Lab, using ADP records through June 2026, finds employment for workers aged 22 to 25 in AI-exposed occupations 19% below their less-exposed peers. That gap was 15% on the July 2025 data vintage. In levels, employment for that age group in the two most exposed quintiles fell about 11% from November 2022 to June 2026, while the same age group in less-exposed work grew about 10%. The U.S. adjustment, like Korea’s, is showing up as less hiring, not a wave of layoffs, and it is concentrated where AI substitutes for tasks rather than complements them.

Jiyeun Chang, a senior research fellow at the Korea Labor Institute, has pointed to a local amplifier. Incumbents at large firms enjoy strong employment protection, so when a company adjusts to a new tool the cut falls on the hiring margin, on the people who were trying to get in. Korea did not need AI to invent that reflex. AI made the reflex cheaper.

Chip Exports Climb as Youth Payrolls Fall

The same months that produced the Issue Note also produced an export boom that barely touches graduate hiring. Bin Hyun-jun, director general of the Social Statistics Bureau at the Ministry of Data and Statistics, told the July jobs briefing that semiconductors have a low employment multiplier and need heavy capital kit, so extra output does not translate into extra heads.

THE JULY AND AUGUST PRINTS AGAINST THE BOOM

  • Youth payrolls: Employment among ages 15 to 29 fell 191,000 from a year earlier to 3.44 million in July, while total employment rose 108,000 to 29.14 million.
  • Youth rates: The July employment rate was 44.2%, down 1.6 points; unemployment was 6.8%, up 1.3 points. In August the employment rate was 44.1%, down 1 point from a year earlier, the 28th month of decline, and the youth headcount fell year on year for a 46th month.
  • Insured jobs: Employment-insurance subscribers rose 278,000 to 15.91 million in August, and manufacturing subscribers rose 2,000, the first annual gain in 15 months. Subscribers aged 29 or younger still fell 55,700, or 2.4%, the 48th month of decline since September 2022, including 23,000 in manufacturing and 13,000 in information and communications.
  • Export tape: August exports were $98.25 billion, up 68.7%, with semiconductor exports at a record $46.65 billion, up 209%, according to the Ministry of Trade, Industry and Energy.

Bin also said the shift from big annual graduate intakes to rolling, experienced-hire recruitment has made the market harder for the young than the old calendar system did. KB Kookmin, Shinhan, Hana, and Woori hired 485 entry-level staff in the first half, down 18.5% from 595 a year earlier. Hana has a regular experienced-hire track for AI, data, and digital specialists. Banks say they need AI people. They want those people already trained.

A supplementary youth survey in May found that among people aged 15 to 29 who had finished their education, 84.8% had ever been employed, down 1.6 points from a year earlier. The first paid job took 11.2 months on average. Among the young who had finished school and still had no work, 605,000 had been out for at least a year, 48.6% of that group, the highest share since 2009. Another 241,000 had been out for at least three years, 19.4%, the high for a series that starts in 2007.

Seoul Funds Training as Firms Hunt Experience

Prime Minister Han Seong-sook told the National Assembly on September 11 that the drop in youth jobs sits on two facts: the arrival of AI, and companies that would rather hire people who already know the work.

The decline in youth employment is driven by factors such as the changing environment with the advent of AI and a corporate environment that prioritizes hiring experienced professionals. In the current landscape, corporations turn to experienced workers instead of creating job opportunities in general.

Han Seong-sook, Prime Minister, National Assembly interpellation, September 11, 2026

She said the government is working with major firms to create places where young people can build experience, because those places are missing. The Bank of Korea’s policy line is stricter than a hiring subsidy. It wants rebuilding career ladders rather than preserving entry jobs that a model can already do, including work-based apprenticeships, support for firm training and mentoring, and a check on whether public money favors machines over people.

WHAT SEOUL HAS PUT ON THE TABLE

  • Age cap: A Cabinet decision on September 8 raised the youth-policy age from 29 to 34 under the Special Act on the Promotion of Youth Employment.
  • Recovery plan: The August 28 package pledges to train more than 300,000 people for advanced and youth-preferred industries and to create more than 300,000 jobs and startup slots by 2030. The 20-to-24 employment rate had already fallen from 46.0% in 2022 to 43.6% in 2025, and the 25-to-29 rate turned down in 2025.
  • Regional pay: Workers who take small-firm jobs outside the capital region are in line for income support of up to about 1 million won a month for two years, mixing a leap incentive, a savings account, and a culture pass, with youth job spending set to rise toward 5 trillion won.

Training people for advanced industries will not reopen a rung that firms have decided they no longer need. The Bank of Korea’s 2026 note still holds out a longer bet: young workers use the tools, they adapt, and they could take a large share of later productivity gains if the work is set up to complement them. That bet only pays if someone still lets a 24-year-old sit next to the work long enough to learn it. The pension files say that seat is the one Korea has been putting away.

Harry is the editor of Oton Technology, an independent site he owns and edits, covering the part of technology that people actually have to act on. After ten years in journalism, first reporting and then editing, he works from primary material by habit: the advisory rather than the write up of it, the filing rather than the press release, the changelog rather than the launch video. Every figure in an article carries its source and its date, and where a number comes from a vendor or an analyst model rather than a count, he says so plainly instead of letting it stand as established fact. What he leaves out is anything he could not verify himself, which on a beat full of unnamed supply chain claims removes a great deal. That standard applies across all the sections the site publishes for an international audience, from artificial intelligence and security to phones, computers, gaming, crypto and the software businesses depend on. He corrects errors in the open and labels them, because a site that hides its mistakes is asking readers to trust the rest on nothing.

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending