Poland's vacancy rate is 0.8 percent. AI is not taking jobs, employers are not hiring

Thesis: the debate about how many jobs AI will take drowns out a weaker and more urgent signal: Polish employers are not hiring. A low vacancy rate is a problem of demand, not only of automation, and that is how policy should address it.

The Eurostat data cited by "Rzeczpospolita" are unambiguous. The job vacancy rate in the European Union fell in the second half of 2026 to 2 percent, and in the euro area to 2.1 percent, the lowest in three quarters. Poland comes out even lower, at 0.8 percent. That is one of the weakest readings in the whole community. The interpretation is often automatic: automation is taking the jobs. Yet something else is just as visible here, perhaps more so. Employers simply have no reason to hire.

When demand for labour is low, vacancies are few regardless of robots. If companies really were replacing people with models at a fast pace, we would expect skill shortages, rising wages in complementary positions and a wave of layoffs in narrow sectors. Recent years give a sharper picture: stable, flat demand, in which every new technology serves as an excuse for what would have been done anyway if the market had grown more slowly.

The World Bank analysis described by "Forsal" shows the full spectrum. Artificial intelligence could add as much as 12 percent to Polish GDP by 2035, but the outcome for employment depends on worker mobility. In the high-mobility scenario, roughly 3 000 jobs are lost net; in the low-mobility scenario, as many as 341 000. In financial and insurance services, employment is to fall by 24 percent, and in construction to rise by 19 percent. The effects on public finances run both ways: personal income tax revenue lower by 0.2 percentage points of GDP, social contributions by 0.6 percentage points, partly offset by higher corporate income tax of 0.1 and VAT of 0.3 percentage points.

This is the real substance of the dispute: not "will AI take the work", but "will workers manage to move". If mobility is low, the cost falls on public finances and on regions dependent on a single sector. Poland, with low geographic mobility and weak employment services, will receive a bill that does not depend on how fast the models develop.

The conclusions from the Chinese debate sound surprisingly practical here. During the eighth APEC ministerial meeting on human resources, the first ever final document devoted solely to AI was adopted, titled "AI strengthening employment in the Asia-Pacific region". At the accompanying WorldSkills Congress in Shanghai, participants reached a conclusion worth carrying to the Vistula: technology is not the biggest barrier. The barrier is organisations that do not give workers time to learn, education systems mismatched to the market and the absence of credible mechanisms for certifying qualifications. A certified graphic designer from WorldSkills described how a task lasting an hour or a week is done by a model in a few minutes, and how few people then get a chance to retrain.

That points to what really needs to be done: invest in learning time during working hours, certify skills in a way employers recognise and restore demand for labour, not just warn about automation. If we do not, the vacancy rate will not be a measure of AI's triumph, but a measure of how much we have stopped investing in people.