Two Hundred Thousand Jobs

Luxembourg's statistics office and the General Inspectorate of Social Security have been running a pilot project to model how artificial intelligence might affect the labour market and the pension system. The initial results were presented in July at the EcoMod2026 conference, though the government has not published the conclusions in detail.[1]

Here is the number that stuck with me: employment in Luxembourg is projected to rise from 522,000 people in 2025 to somewhere between 735,000 and 927,000 by 2070. That is a spread of nearly 200,000 jobs between the optimistic and pessimistic scenarios.

Two hundred thousand jobs. In a country with 660,000 residents and roughly 522,000 employed people today, that is not a rounding error. That is the difference between a system that comfortably pays pensions and one that does not.

The government is careful to note that these projections do not specifically isolate AI's impact. They are general scenarios testing different productivity trends. But the exercise exists because someone in the ministry realised that if AI automates enough jobs fast enough, the contributions that fund the pension system could take a hit that nobody planned for.

Statec intends to take the work further. The ministry of labour has commissioned a sector-specific study on AI's impact in Luxembourg, which should help identify which industries and skills are most exposed. A working group called "KI-Dësch" brings together government, unions, civil service, and business representatives to discuss these questions.[2]

The parliamentary question that triggered this reply came from an ADR MP who was thinking about automation and wages. The answer was signed by the labour minister, with copies to the health, finance, economy, and prime minister's offices. In other words, this is not a fringe concern. It is on the agenda at the top.

What I find striking is the honesty of the uncertainty. The government is not pretending it knows what AI will do to employment. It is modelling scenarios, acknowledging a 200,000-job spread, and admitting that the figures highlight "the extent of the uncertainties." That is more refreshing than a confident prediction that turns out to be wrong.

The real question is whether any scenario ends up mattering more than the others. If AI displaces workers faster than new jobs appear, the pension system depends on fewer contributors paying for more retirees. If AI boosts productivity enough to raise wages, the contributions might hold. And if the truth is somewhere in the middle, as it usually is, the system muddles through with adjustments.

But 200,000 jobs of uncertainty is a lot to plan around. Luxembourg is wise to start modelling now rather than waiting to find out the hard way.