Is AI replacing junior jobs? What the 41-country study means for who you hire (October 2026)

AI is shifting hiring toward senior roles: what to do now

Partly, and not the way headlines suggest. The King's College London and Stanford study of 41 countries (October 2026) found firms adopting generative AI grow headcount about 3.3%: senior roles up 6.7%, junior roles down 2.5%. Davenport and Mittal's All-In On AI points to the fix: redesign junior roles around reviewing and integrating AI work instead of cutting them.

If an AI model now writes the first draft, the research memo and the test suite, why hire a junior at all? That question sits under a lot of hiring plans this autumn, usually unspoken. Two pieces of evidence published this week answer it differently, and the gap between them is the useful part.

The first is a study by Bharat Chandar of Stanford University and Bouke Klein Teeselink of King's College London, "How Does AI Change Labor Demand? Evidence from 41 Countries," which King's announced on October 6, 2026. It draws on 1.25 billion job advertisements and 158 million employee positions. A firm counts as an adopter once its job ads say applicants will need to use generative AI. After adoption, total employment rises about 3.3%, senior employment rises 6.7%, and junior employment falls 2.5%. Chandar's summary is careful: generative AI "is changing the mix of jobs within organisations rather than reducing overall employment," and the pattern shows up from the US to Brazil to Saudi Arabia.

The second is a Brookings article published October 6, 2026 by Anders Humlum and Emilie Vestergaard, based on surveys of 25,000 Danish workers in 11 AI-exposed occupations linked to monthly administrative records through December 2024. They find precise null effects of chatbots on earnings and hours, ruling out average effects larger than 2%. And on the junior question specifically: the share of early-career workers evolved no differently at workplaces that encourage chatbot use, with estimates tight enough to rule out differences larger than a third of a percentage point. Early-career jobs did fall in exposed occupations such as software, legal and marketing, but in Denmark the firms adopting AI were not the ones driving that fall.

How can both be true? The studies define adoption differently (a line in a job ad versus workers' reported use and employer policy), cover different countries, and the Danish records stop at the end of 2024, before this year's agent tools. What they agree on matters more for a founder: headcount at adopting firms is not collapsing, and the work inside those firms is being reorganized. Humlum and Vestergaard report that 85% of users spend the time they save on other job tasks, about a third of the new tasks involve overseeing AI (quality review, ethics and compliance), and about a quarter are integration work: tuning assistants, writing usage policies, building chatbots into workflows. That judging work is where the 41-country data says firms are adding senior people.

This is the pattern Thomas Davenport and Nitin Mittal warned about in All-In On AI. They report that in one of their AI surveys only 10% of US respondents stated a clear preference for retraining and keeping current employees, and they call the replace-first instinct shortsighted: you cannot find or afford enough outside AI talent, and the people who already know your business and customers are the ones worth upskilling. A founder who stops hiring juniors in 2026 because the model drafts well will be short of seniors who can judge the model in 2030. The junior role is being redefined, not deleted. Here is the rule I would apply before cutting one:

Name the judgment the role grows into. Before closing a junior requisition, write which senior decision this person would own in three years. If you cannot name one, the role was weak before AI, and AI is not the reason to cut it.
Rebuild the job around the new tasks. Give the junior ownership of the work Brookings found appearing: reviewing AI output against a checklist, and wiring the tools into one team workflow, paired with a senior who signs off.
Measure caught errors and built workflows, not drafts. A junior who produces drafts competes with a $20 subscription. A junior who catches what the model got wrong is building the judgment your future seniors need.
Check your own numbers in two quarters. Track how many hours your seniors spend reviewing AI work. If that number falls, the redesigned role is paying for itself; if it rises, you are buying senior time to cover a gap.

The limits deserve naming. The 41-country study measures postings and positions, not productivity, and firms that announce AI requirements in their job ads may differ in other ways, for example by already growing faster. I have read the King's release and the Phys.org write-up, not the full paper, so treat the exact model behind the numbers as unverified here. The Danish evidence ends in December 2024 and may not describe what agents do to hiring in 2026.

If you are the junior in this story, the Brookings data has a quieter finding for you: chatbot adopters were more likely to switch occupations, and the switchers' earnings growth ran 12 percentage points ahead of other Danish workers, mostly into roles where credentials do not bar entry. The tools are not only a threat to the first rung. Used hard, they are a way onto a better ladder.

Sources: AI adoption linked to increase in employment globally, King's College London, October 6, 2026; AI adoption linked to increase in employment globally, Phys.org, October 7, 2026 (paper DOI 10.2139/ssrn.7498743); Still waters, rapid currents: Early labor market transformation under generative AI, Brookings, October 6, 2026; Thomas H. Davenport and Nitin Mittal, All-In On AI (Harvard Business Review Press), passage quoted in Notes About "All-in on AI" by Thomas Davenport and Nitin Mittal, akos.ma.