Dutch AI productivity gains depend on investment and worker mobility
High AI adoption does not automatically raise Dutch productivity, according to De Nederlandsche Bank (DNB). Gains require reliable access, investment in skills and workflows, and movement of labour and capital toward more productive businesses.
Exposure does not mean displacement
More than a third of Dutch employees use generative AI at work.
About a quarter work in occupations where AI can accelerate at least half of tasks; lowering the threshold to 40 percent covers roughly half of workers.
Higher-education graduates represent 62 percent of highly exposed workers versus 45 percent of all workers.
Exposure does not predict job losses: AI can support workers as well as automate tasks.
Annually, 17 percent change employers; 57 percent of job changers also changed industry in 2024–2025.
Separate research finds involuntary transitions cost about 9 percent of annual income over five years, not an AI-specific loss.
DNB recommends training, early transition support and fewer barriers to productive firms expanding.
Investment precedes the productivity dividend
Concentrated model, cloud and chip supply creates access risks; DNB recommends open standards, portable data, fallback options and European infrastructure cooperation.
AI investment can raise demand before productivity expands supply.
Computer prices rose about 10 percent in Europe over the past year and 15 percent in the Netherlands.
A model compares a 1 percent productivity increase with one anticipated four quarters ahead: anticipation initially raises inflation.
Later productivity gains can reduce costs; evidence on AI-driven pricing remains limited.
Adoption is not the payoff
DNB identifies access, diffusion and reallocation as successive conditions for economy-wide gains.
Limited business dynamism and market fragmentation constrain the transmission of firm-level improvements.
For monetary policy, the report distinguishes immediate investment demand from later supply gains.