Research update · 8 October 2026
Danish AI adopters slowed hiring, not layoffs
Fewer hires can narrow an entry route while a firm keeps growing. Bonin, Darougheh and Kuchler report that Danish firms first using AI in 2023 had employment about 9% below their previous growth trend relative to non-adopters by September 2025. This is an association, not a clean estimate of jobs lost because of AI.
What was measured
The authors linked Statistics Denmark’s firm AI-use survey to monthly employer–employee wage records. They studied private, non-financial firms with at least 10 employees. They compared first-time adopters with non-adopters in the same detailed industry, measuring both against each firm’s pre-adoption trend. The main cohort first reported AI use in 2023; a 2024 cohort showed a similar path a year later.
The gap was concentrated in firms with fewer than 100 employees. In the authors’ exposition, these small adopters were about 16% below trend by late 2025. The shortfall showed up mainly through fewer new hires, especially in AI-exposed occupations, rather than more layoffs. The firms still grew in absolute terms. The central bank’s working-paper abstract describes an 11% employment-growth reduction by late 2025; we have not verified why its figure differs from the authors’ roughly 9% trend-relative figure in the 8 October account.
What it cannot prove
Adoption was not randomly assigned. A business facing a different growth path could both start using AI and hire less. The pre-trend and second-cohort checks help, but do not eliminate that possibility. The small 2023-adopter group employed less than 2% of Danish workers, so this is not evidence of a national employment collapse.
Yesterday’s Danish chatbot-encouragement result found no extra early-career-share decline through 2024. It uses a different sample, AI-use indicator, outcome and end date. The two results need a common comparison, not a vote on whether AI “takes jobs.” The next check is whether the hiring gap persists for comparable firms after financing, demand and AI-use intensity are measured together.