OECD Finds Schooling Reshaped Work More Than AI Exposure Did
A working paper the OECD published with Eurofound on 17 November 2026 tracks how European work was organised between roughly 2011 and 2024 and finds that rising educational attainment explains more of the change than occupational exposure to artificial intelligence does.
Occupations more exposed to artificial intelligence have seen a smaller increase in monotonous tasks than less exposed ones since roughly 2011, according to a working paper the OECD published jointly with Eurofound on 17 November 2026. The same paper concludes that rising educational attainment appears to have done more than AI to change how European work is organised.
The paper runs to 47 pages as number 68 in the OECD Artificial Intelligence Papers series. It draws on the OECD's Survey of Adult Skills, known as PIAAC, and Eurofound's European Working Conditions Survey. Its window covers the early diffusion of generative AI rather than its mature use.
What moved over those years is not in dispute. Manual tasks declined. Problem-solving and information-processing tasks became more prevalent. Teamwork became more common. The question the authors set is how much of that AI can account for, and their answer is cautious: AI may have played some role, with the monotony finding and faster growth in some information-processing tasks as the evidence, while education carried more of the weight. The policy line they draw from it is that investment in education and training still matters.
Twelve per cent, spread very unevenly
The adoption base is thin enough to make that modest finding unsurprising. Eurofound's 2024 survey recorded 12 per cent of workers in the EU reporting use of generative AI tools at work. The spread across member states is wide: at least one in five workers in Luxembourg, Sweden, Belgium and Denmark, against 5 per cent or lower in Greece, Italy, Portugal and Romania. A preprint analysing the same survey data puts its sample at about 36,600 workers across 35 countries and the country range at under 3 per cent to roughly a quarter.
Eurofound's own overview report of the survey reached a conclusion pointing the same way as the new paper. A preliminary descriptive analysis found generative AI use associated with a lower level of monotonous tasks and a greater likelihood that workers assess the quality of their own work, a pattern the report attributes to where the tools have landed, namely among professionals and managers who already hold more discretion. Its summary of the first findings put the point bluntly: technology, Eurofound wrote, 'creates more tasks than it removes'.
The benign reading does not cover everything
A different Eurofound report drawn from the same 2024 survey complicates the picture. Measuring algorithmic management as whether computer programs allocate employees' tasks, set their hours or monitor their performance, it found that employees subject to it report higher job demands, particularly quantitative demands indicating work intensity, and are more likely to experience job strain. Correlations between that managerial control and generative AI use were weak, between 0.065 and 0.124, which suggests the two are separate phenomena rather than one story. Whatever generative AI is doing to task content, digital supervision is a distinct question with worse answers.
The survey instrument has a limit too. The 2024 question asks whether a worker uses artificial intelligence that simplifies complex mental tasks or makes recommendations, naming tools including ChatGPT and Midjourney. It records only whether someone uses such a tool, not how often, and it misses AI embedded in systems a worker never sees.
What this desk could not read
The OECD's catalogue entry and Eurofound's listing both carry the same abstract, which reports the direction of the findings rather than their size. The full text was not retrievable at the time of writing, so no point estimates, confidence intervals or exposure measures from the paper are reported here. The claim that education outweighed AI is the authors' summary of their own results, and it has not been checked against their tables.
The honest reading is that this is a useful negative result arriving at a convenient moment. Two institutions with no stake in deflating the subject have looked at a decade of European task data and found that the single largest driver of how work is organised was the gradual upgrading of the workforce's qualifications, not the technology that has dominated the discussion of work since 2023. That is a finding about a period in which generative AI was new; it says nothing about 2027.
What happens next?
- EU member states must transpose the Platform Work Directive, which contains the first EU rules on algorithmic management, by 2 December 2026.
- The European Commission's proposal for a Quality Jobs Act, which names algorithmic management and AI at work as priority areas, is expected by the end of 2026.
- Eurofound's 2025 to 2028 research programme continues work on AI-enabled work management systems and their effects on employment and social dialogue.
Related topics
Sources & references
- 01Artificial intelligence and work organisation — OECD PublishingprimaryOECD/Eurofound (2026), OECD Artificial Intelligence Papers No. 68, 47 pages, released 17 November 2026; abstract is the source for all findings attributed to the paper
- 02Artificial intelligence and work organisation — EurofoundprimaryCo-publisher's listing of the same working paper
- 03A new world of work: European Working Conditions Survey 2024 first findings — EurofounddataSource for the 12 per cent generative AI figure and the quoted line on technology creating more tasks than it removes
- 04European Working Conditions Survey 2024: Overview report — EurofoundreportMirrored copy of Eurofound's overview report, read for the country breakdown of generative AI use and the descriptive finding on monotonous tasks
- 05Algorithmic management in European workplaces — EurofoundreportSource for the job strain findings and the correlation range between algorithmic management and generative AI use
- 06Generative AI at Work: From Exposure to Adoption across 35 European Countries — arXivresearchPreprint, not peer reviewed; cited only for the EWCS 2024 sample size and the wording of the survey question
- 07Artificial intelligence (topic page) — EurofoundprimarySource for the 2 December 2026 transposition deadline of the Platform Work Directive and the expected timing of the Quality Jobs Act proposal
More from Future
Workers' Share of US Output Falls to a Series Low of 53%
US nonfarm business productivity rose at a 3.6 per cent annualised rate in the third quarter of 2026, the Bureau of Labor Statistics reported on 5 November, as output grew 3.7 per cent and hours worked grew 0.1 per cent. The labour share fell to 53.0 per cent, a series low.
US October Job Cuts Fall 71% as AI Drops Down the List
Announced American job cuts fell 71 per cent against October 2025 while announced hiring plans of 444,692 were the highest for any October since 2009. Initial jobless claims of 189,000 matched a 57-year low, though the composition of the data makes it weaker evidence than the headline suggests.
Britain's Wage Floor Undershoots Its Own Target by 23p
The UK government accepted the Low Pay Commission's recommendations in full on 28 October 2026, taking the National Living Wage to £13.21 from April 2027. The Commission's own estimate of the rate needed to hit its statutory reference point was £13.44.
Italy Bars AI From the Final Call on Hiring and Dismissals
Italy's second decree implementing its national AI law took effect on 23 October 2026. It requires a natural person with effective and independent authority to make final employment decisions, and folds AI systems that reshape how work is organised into the risk assessment every Italian employer already has to carry out.


