Dutch graduate study finds AI changing tasks more often than removing them

A study published on 22 September offers a more detailed view of AI at work than a prediction about jobs gained or lost. Researchers at Maastricht University’s Research Centre for Education and the Labour Market examined the careers of graduates from Dutch higher professional education, known as hbo. Their report follows labour-market outcomes after graduation and asks working graduates about AI in their jobs.

More than six in ten employed graduates surveyed use AI at least weekly. Among AI users, 28.2% reported that existing tasks had changed, 15.0% reported new tasks and 6.7% reported that tasks had disappeared. The report describes changes in drafting, editing and summarising text, processing information and programming. It also identifies new work in applying, implementing and checking AI.

Those figures describe responses from people in work. The researchers explicitly warn that the data cannot show what AI has done to the total number of jobs or to a new graduate’s chance of finding one. A task disappearing from an employed person’s role is different from an entire job disappearing. Equally, a worker reporting a new responsibility does not prove that an employer has created an additional vacancy.

The study also tracks how careers develop beyond the first job. For the group examined, median gross hourly pay rose from €16.40 roughly a year and a half after graduation to €26.80 four to eight years after graduation. More graduates had permanent contracts and leadership responsibilities later in their careers. These are observations about Dutch hbo cohorts, not a forecast of pay for a current student.

Continuing education is another substantial finding. More than four in ten graduates had pursued further formal education, while nearly two-thirds had taken a course or training. The report describes training connected both to broader career goals and to changes in day-to-day work. It suggests that adapting after graduation is already a normal part of many careers in this group.

Individual institutions published their own responses on the report’s release day. Hogeschool Utrecht said nearly two-thirds of its alumni used AI at work weekly or daily, while Saxion reported that 36% of its surveyed alumni used it weekly and 31% daily. Those local figures should not be substituted for the national estimates: they describe graduates of particular institutions. They do, however, show why universities of applied sciences are examining what alumni do after leaving campus.

Students can take a practical question from the findings: when an occupation uses AI, which responsibilities remain with a person? The report points to the need to assess information, supervise processes and check outputs, alongside learning to use tools. That is a more concrete way to examine a degree or internship than asking whether its job title is simply safe from AI.

The findings should travel cautiously beyond the Netherlands. Hbo is a particular higher-education route with its own labour-market links, and the survey covers specific graduates and periods. For students worldwide, the value of the new report is its measured description of changing work and its clear limit: it does not settle how many entry-level jobs AI will create or remove.