AIEL Monitor · monthly brief · 2026-11
AI and the labour market, November 2026
A monthly snapshot of AI in the Swedish labour market from the AI-Econ Lab, from public data. This month: Job quality and the work environment.
Are AI-exposed jobs good jobs?
Most of the debate on AI and work is about the number of jobs, less about what the jobs are like. We look at what the jobs are like, both as employees describe them and as employers offer them in advertisements.
More demanding, more say, ordinary terms
AI-skill ads minus other ads in the same occupation, percentage points.
The most AI-exposed third of occupations looks like what work-environment research calls active jobs: more demanding and more autonomous. In the most exposed occupations, on average 64% of workers find the work mentally strenuous, against 52% in the least exposed. In the most exposed occupations 33% cannot switch off from work after hours, against 25% in the least exposed. At the same time, 74% of employees in the most exposed occupations can influence how their work is done, against 65% in the least exposed. The difference in influence is far larger among women, 73% against 56%, than among men, 75% against 68%. Last month we showed demand for AI skills spreading from ICT into ordinary occupations. That shows in the terms too. In 2018, 71% of AI-skill advertisements offered a permanent contract, in 2025 only 64%, while other advertisements stayed between 67% and 72% throughout. It looks like worse terms for AI jobs, but almost all of the fall comes from AI-skill advertisements moving into occupations where fixed-term contracts are more common. Compared with other advertisements in the same occupation, AI-skill advertisements offer a permanent contract at least as often, in every year. Full-time work is the exception. Up to 2022 it was slightly more common in AI-skill advertisements than in others in the same occupation, but since 2023 it has been 2 to 3 percentage points less common.
Source: SCB, Arbetsmiljöundersökningen 2024, by occupation (SSYK 2-digit) × DAIOE generative-AI composite (g2gen) v2025, least- vs most-exposed third (about 9 occupation groups each), unweighted mean over groups; JobTech historical job ads (Arbetsförmedlingen), CC0, frozen v1.7 term list, distinct advertisements, 2018–2025, 4,366 AI-skill advertisements in 2025. Full method at ai-econlab.com/monitor/#method.
Occupations, not AI
Both comparisons are descriptive. The exposed occupations are also more qualified, so much of the gap in working conditions reflects the kind of work rather than AI, and the 2024 survey's new method rules out comparison with earlier years. Comparing within occupations removes differences between occupations but not differences between employers, and advertisements show the terms offered, not the jobs themselves.