A small Ritzau article caught my eye this morning. A group of Danish economists are warning that future generations may spend a much smaller share of their adult lives in retirement than today’s pensioners do.
The reason is simple. Denmark’s retirement age rises automatically as life expectancy increases. That sounds fair enough at first glance: if people live longer, they can also work longer. But the mechanism has a nasty little twist. The extra years are mostly added to working life, not to retirement. Work expands; retirement shrinks.
Those retiring now may spend roughly a third of their adult lives as pensioners. Children born today may get closer to a quarter. If the current logic continues, the Danish retirement age could eventually move towards the late seventies. The proposed alternative is to keep retirement at around 30 percent of adult life. The pension age would still rise, but more slowly.
This is already an important debate. But there is an elephant standing in the room, making quiet mechanical noises and politely drafting a policy memo in the corner.
The elephant is AI.
The discussion about pension age is still framed as if the future labour market will look broadly like the present one, merely older. We assume that people will still be needed in roughly the same way, that work will still be organised around human labour, and that the great challenge is simply how to keep enough people working for long enough to finance everybody else.
That may be the wrong question.
If AI and robotics develop even moderately as expected, the next fifty years will not simply be a story of longer lives and strained pension systems. They may be a story of radically altered productivity. Large parts of clerical work, administrative work, translation, basic legal work, programming, customer service, logistics, transport, agriculture and perhaps even parts of care work may be automated or semi-automated. Not overnight, not evenly and not without mess, but enough to change the basic assumptions.
In that world, why exactly should the retirement age keep rising?
The usual answer is demographic: there will be more old people and fewer workers. But this assumes that the number of workers is the key constraint. What if productivity per worker rises dramatically? What if one person with AI tools can do what previously required three or four? What if robots begin taking over more physical work as well? Then the problem is no longer simply how many humans are working. It becomes how the gains from automation are distributed.
This is the part that is often missing from polite economic discussion. If AI merely makes companies more profitable while ordinary people are expected to work longer, the future will feel like a con. We will have invented machines that can do more and more of our work, only to tell human beings that they must remain in the labour market until 73, 75 or 77. That is not progress. That is a spreadsheet wearing a human skin.
There is also a more immediate problem. Raising the retirement age assumes not only that older people are alive, but that they remain employable. That is not the same thing.
We already know from the computer revolution of the 1980s and 1990s that some workers struggled badly when working life was reorganised around new technology. Many were not unintelligent. They were competent people whose habits, identities and skills had been formed in a different system. Younger workers often adapted because they had to, and because they had less to unlearn. Older workers sometimes found themselves stranded.
AI may repeat that pattern, but more brutally. Computers changed how offices worked. AI changes how thinking work itself is done. It does not merely replace paper with screens. It changes drafting, searching, coding, summarising, analysing, planning and communicating. It rewards people who can work experimentally with opaque systems, tolerate imperfect first drafts and reorganise their habits quickly. Those are not evenly distributed skills.
Some people in their sixties will become extraordinarily productive with AI. A domain expert with forty years of experience and good AI habits may be formidable. But it would be reckless to build national policy on the assumption that most people can keep reinventing themselves technologically until nearly seventy. Some can. Many cannot. A society is not made only of exceptional hobbyists who install Linux for relaxation and read documentation for pleasure. Fortunately for civilisation, or unfortunately for the software industry.
If large groups of people in their late fifties and sixties become difficult to employ in an AI-assisted labour market, raising the pension age may save less than it appears to. The state may simply pay in other ways: unemployment benefits, retraining schemes, stress leave, disability pensions, flex jobs, administrative programmes and healthcare costs. A person who is too young for retirement but too old, too tired or too technologically displaced to be fully useful is not a fiscal triumph. They are a policy failure with a course certificate.
This is why the pension debate should not only be about life expectancy. It should be about working capacity, technological change and dignity.
There is a strange contradiction in the way we talk about the future. On one day, we are told that AI and robotics will transform productivity. On the next, we are told that people must work longer and longer because society cannot afford retirement. Both claims cannot remain true in the same simple form forever. If technology makes us vastly more productive, some of that productivity ought to be returned to citizens as time.
Perhaps the radical conclusion is that the AI and robotics revolution should eventually lower the retirement age rather than raise it.
Not necessarily tomorrow. Not as a sudden fiscal experiment conducted with the usual political combination of slogans and panic. But as a serious long-term goal. If machines can carry more of the productive burden, human beings should not automatically be forced to remain in paid employment for longer lives. We could instead imagine a society where full-time labour is concentrated in the healthier, more adaptable decades, and where people move earlier into pension, partial work, mentoring, local projects, care, culture, gardening, repair, teaching, volunteering or simply living.
That would require a different model of distribution. It would mean taxing the returns from automation properly. It would mean treating productivity gains as a common social resource rather than merely as private profit. It would mean admitting that the aim of an advanced economy is not to keep everyone busy until they collapse neatly into retirement.
The economists are right to question whether future generations should get a smaller share of adult life in retirement. But the deeper question is whether the whole idea of ever-rising pension ages belongs to a world that is already passing away.
If AI and robotics fulfil even part of their promise, the central problem of the twenty-first century may not be scarcity of labour. It may be scarcity of meaningful economic roles. In that case, the humane response is not to force people to chase employability into their seventies. It is to redesign the social contract around abundance, dignity and time.
The machines are coming. The least we can do is ensure they do not arrive merely to make us work longer.