---
id: "lib-2026-09-20-how-does-ai-change-labor-demand-evidence-from-41"
type: "library"
title: "How Does AI Change Labor Demand? Evidence from 41 Countries"
lang: "EN"
org: "Stanford Digital Economy Lab"
source_type: "Paper"
date: "2026-09-20"
week: "2026-W38"
url: "https://nosov.ai/library/lib-2026-09-20-how-does-ai-change-labor-demand-evidence-from-41/"
original_url: "https://digitaleconomy.stanford.edu/app/uploads/2026/09/HowDoesAIChangeLaborDemand.pdf"
tier: "A"
tier_name: "Strong evidence"
tier_reason: "Stanford Digital Economy Lab working paper using 1.25 billion job postings and 154 million employment records across 41 countries with matching, a cross-border instrument and a general-equilibrium model; not yet peer reviewed."
summary_basis: "full-text"
original_minutes: "90"
tags: ["labor demand", "seniority"]
---

# How Does AI Change Labor Demand? Evidence from 41 Countries

Tier A (Strong evidence): Stanford Digital Economy Lab working paper using 1.25 billion job postings and 154 million employment records across 41 countries with matching, a cross-border instrument and a general-equilibrium model; not yet peer reviewed.

Summary based on the full text.

## In short

Bharat Chandar (Stanford) and Bouke Klein Teeselink (King's College London) estimate how generative AI adoption changes who companies employ, using multinational affiliates in 41 countries (paper dated 20 September 2026). Adopters reduce the junior share of their workforce, but mainly by adding senior staff rather than cutting juniors. Total employment rises slightly. The pattern appears in most countries, so the shift away from junior workers is not only a US phenomenon.

## Key numbers

- 1.9 pp: lower junior share of employment at AI-adopting affiliates than at controls by March 2026
- +6.7% / -2.5%: senior employment change versus (not significant) junior employment change at adopters
- 23 of 31: countries where the junior share fell, significant in seven including the US, Brazil and Saudi Arabia
- +3.3%: marginally significant rise in total employment at adopting affiliates

## What it reports

- By March 2026, the junior share at adopting affiliates was 1.9 percentage points lower than at matched controls, with the gap opening after ChatGPT's release.
- Senior employment rose 6.7%; junior employment was 2.5% lower, not statistically significant; total employment rose a marginally significant 3.3%.
- Within seniors, demand moved 5% toward the most AI-exposed roles; within juniors it moved 2% away from the most exposed roles toward the least exposed.
- In 21 of 22 occupation groups the share changed by less than half a percentage point. Computer and mathematical occupations gained 0.8 point of employment share while their junior share fell 3.3 points.
- Across the 31 countries with a strong enough instrument, the junior share fell in 23, significantly in seven (including the United States, Brazil and Saudi Arabia); the average decline was 1.1 points, senior employment rose a marginally significant 3.4%, and no country showed a significant fall in total employment.
- Junior losses run somewhat deeper in richer, more digitised economies and shallower where labour regulation is more rigid, though most of these correlations are not significant.
- The model implies AI so far gives modest, seniority-biased productivity gains: labour saving for juniors and labour expanding for seniors in exposed occupations. It also shows firm-level growth at adopters can coexist with falling aggregate employment.

## How they got it

Revelio Labs data: 1.25 billion job postings and 154 million employment records, built into a monthly affiliate panel from January 2021 to March 2026. Adoption is inferred from job ads involving generative AI use. Each adopting affiliate is matched to similar non-adopters in the same country and sector, and adoption is instrumented with the adoption rate among other multinational parents headquartered in the same metro area as the foreign parent. A task-based general-equilibrium model separates the firm-wide scale effect from worker-specific effects.

## Limits

- Results apply to digitally visible multinational affiliates in Revelio data, not to all firms or to domestic companies.
- Adoption is inferred from job-posting text, which may miss internal use.
- The instrument assumes peer adoption near the parent's headquarters affects the affiliate only through the parent's own adoption; the authors test this but it cannot be proven.

## Why it matters for companies

The finding for employers is about composition: firms adopting AI are not shrinking so far, but they are hiring more experienced people and fewer entry-level ones, especially in technical roles. That erodes the pipeline through which companies grow future senior staff, and the effect shows up in most countries studied, not only in the US. HR and business leaders should track the junior share and entry-level hiring by role alongside total headcount. The result also suggests that the people gaining most in exposed roles are seniors who can direct and check AI output.

Original: https://digitaleconomy.stanford.edu/app/uploads/2026/09/HowDoesAIChangeLaborDemand.pdf