“Sooner than expected”

Three million fewer workers: why Italy needs AI

By 2050 Italy will have about 7 million fewer people of working age. With fewer people, AI is what keeps hospitals, offices and businesses open. The brake is skills.

Line drawing: an age pyramid made of horizontal bars, wide at the top and narrow at the bottom; the middle bars are shorter than their dashed outlines, and a black robotic arm standing on the ground to the right holds in its gripper a black bar that fills the gap in the middle of the pyramid.
Older people are increasing, people at work are decreasing. Technology can carry part of the load, if someone knows how to use it.

Listen to the episode

The podcast episode, in Italian, with Matteo's voice recreated by artificial intelligence · 10 min

In short

  • According to ISTAT projections published on 31 August, between 2025 and 2050 the number of people aged 15 to 64 will fall from 37.3 to 30.2 million. Even with many more women and older people at work, the labour force will lose about 2.9 million people.
  • The shortages are already visible: firms struggle to find 43.6% of the people they want to hire, nurses are scarce, registered domestic workers have been falling for four years and productivity has grown by 0.3% a year since 1995.
  • AI can help do the same work with fewer people, but it is used by 53.1% of large firms and 15.7% of small and medium-sized ones. The main brake is skills.

The figures of 31 August

On 31 August ISTAT, Italy's national statistics institute, published its new projections for the population, households and the labour force. For work, the key figure is this: between 2025 and 2050 the number of people aged 15 to 64, the group statistics treat as working age, will go from about 37.3 to 30.2 million. About 7 million fewer in twenty-five years.1

ISTAT also tried to estimate how much of that loss could be recovered if more people worked. It assumed that the activity rate of 15 to 64-year-olds rises from 66.7% to 71%, and that of women from 57.8% to 64.2%, a level Italy has never seen. Even then the labour force falls from 24.8 to 21.4 million, which is 3.4 million fewer people. If 65 to 74-year-olds are counted too, since they will work longer than today, the loss shrinks to 2.9 million, or 11.5%.1

Francesco Seghezzi, president of the labour research association ADAPT, pointed out in the newspaper Domani that this is the favourable scenario. It assumes that the number of economically inactive people drops from 12.4 to 8.8 million, that many more people work past 65 and that women take part in the labour market as never before. If all of that happens, he writes, the labour force will still shrink by almost 3 million.2

Meanwhile people aged 65 and over will go from 24.7% to 34.5% of the population, and those over 85 from 4.1% to more than 7%. They are the people who need the most care.1 According to Eurostat, Italy already has the highest ratio in the European Union between older people and the working-age population: 39 people aged 65 or over for every 100 aged 15 to 64, against an average of 34.5. It also has the highest median age, 49.1 years.3 With the shares ISTAT projects for 2050, the figure reaches about 62 for every 100. That calculation is mine.

30.2 millionpeople aged 15 to 64 projected for 2050, about 7 million fewer than todayISTAT, August 2026
2.9 millionfewer economically active people aged 15 to 74, even with women and older people working far more than todayISTAT, August 2026
34.5%the share of people aged 65 or over projected for 2050. Today it is 24.7%ISTAT, August 2026
355,000babies born in Italy in 2025. In 2008 there were more than 576,000ISTAT, 2025 and 2026

Why the gap will not fill quickly

Up to 2050 these numbers would change little even if more babies were born again. Anyone who will be 25 in 2050 was born in 2025, and in 2025 Italy recorded 355,000 births, 3.9% fewer than the year before, with an average of 1.14 children per woman.4 In 2008 there had been more than 576,000 births. They have fallen every year since.5

Some young people, moreover, go to work abroad. According to CNEL, the National Council for Economics and Labour, 630,000 Italians aged 18 to 34 emigrated between 2011 and 2024; net of arrivals, the balance is 441,000 fewer young people. Among those who left in 2022-2024, 42.1% were graduates.6 In 2025 departures of Italian citizens fell to 109,000, but ISTAT warns that the comparison is distorted by an exceptional number of registrations on the register of Italians abroad in 2024.4

The other lever is immigration, and it already counts for a lot: in 2025 net migration from abroad added 296,000 people.4 ISTAT's projections assume it stays positive throughout, so the loss of about 7 million people of working age already takes future arrivals into account.1 Through its annual migration quotas, the decreto flussi, the government has planned 497,550 work entries between 2026 and 2028.7 The system works badly, though. According to monitoring by the Ero Straniero campaign, reported by Il Post, of the 181,450 quota places for 2025 only 14,349 turned into a residence permit, 7.9%.8

Where people are already missing

Firms already see the problem when they hire. In October, according to the Excelsior system run by Unioncamere, the national body of chambers of commerce, and the labour ministry, companies planned more than 539,000 hires and said they had trouble finding people for 43.6% of them.9 A year earlier the share was 46.8%, and the main reason given was a lack of candidates, more than their level of preparation.10

In health care the shortage has a precise age. According to the first report by FNOPI, the federation of Italy's nursing professional bodies, with the Sant'Anna School in Pisa, Italy has about 6.5 nurses per 1,000 inhabitants, against a European average of 8.4. Of the 458,112 nurses on the registers in 2024, almost half were over 50 and only one in eight was 30 or younger.11

In care for older people something seemingly contradictory is happening: the number of people who need help is growing while the number of registered workers is falling. According to INPS, the national social security institute, in 2025 there were 804,464 domestic workers with contributions paid, down for the fourth year in a row. Live-in and home carers for older people, the badanti, make up 51.3% of the total, and almost seven workers in ten are foreign nationals.12

The public administration is ageing too. According to FORUM PA's analysis of the State Accounting Office's annual staff report, in 2024 Italy had 3,388,794 public employees, with an average age of 49 years and 4 months. 54% were over 50 and more than 121,000 had already turned 65. Hiring linked to the national recovery plan increased the number of under-30s by 33% in a year, but they are still only about 210,000 out of almost 3.4 million.13

Doing the same work with fewer people

When the number of people at work falls, the only way not to get poorer is to produce more in every hour worked. That is exactly what Italy has failed to do for thirty years. According to ISTAT, between 1995 and 2024 labour productivity grew by an average of 0.3% a year, against 1.5% in the European Union, 1.2% in Germany and 1% in France.14

Average annual growth in labour productivity, 1995-2024

European Union1.5%
Germany1.2%
France1.0%
Spain0.5%
Italy0.3%
ISTAT, Misure di produttività, December 2025. Value added per hour worked. Striped: the average of the 27 EU countries.

In June the Bank of Italy published a study estimating what AI could do. If it spreads widely, it could add between 0.2 and 1.1 points a year to productivity growth for ten years, depending on how fast and how deep adoption goes. For a country stuck at 0.3% that would be a great deal. The same study, though, says that so far no systematic productivity effects can be seen in the Italian firms that have adopted AI. The gains on individual tasks are real, but to see them in company accounts processes and organisation have to be rethought, as happened with computers.15

Japan, which is ageing ahead of us, offers a useful example. The Japanese government estimates that 2.72 million care workers for older people will be needed in fiscal 2040. In 2023 there were 2.126 million, the first fall since the public long-term care insurance system began in 2000.16 Three economists, Yong Suk Lee, Toshiaki Iizuka and Karen Eggleston, studied Japanese nursing homes that introduced robots and sensors. Where robots arrived, especially monitoring robots, staff numbers rose and fewer people left. The time freed up went to tasks involving direct contact with residents, and the use of restraints and pressure ulcers both fell.17

Seghezzi also points to a paradox worth keeping in mind. Many jobs AI can replace are in offices, where candidates are not in short supply. In care, health and assistance, where people are already scarce, replacing them may not be possible.2 In the Japanese nursing homes technology worked because it took weight off the people who stayed.

The real brake: skills and small firms

In Italy AI is spreading, but at two speeds. According to ISTAT, in 2025 16.4% of firms with at least 10 employees used at least one artificial intelligence technology, twice as many as the year before. Among large firms the share was 53.1%, among small and medium-sized ones 15.7%.18 The European average is 20%.19 The Bank of Italy calculates that Italy is four points below the EU average and almost ten below Germany.15

Firms with at least 10 employees using AI, 2025

Large firms, European Union55.0%
Large firms, Italy53.1%
All firms, European Union20.0%
All firms, Italy16.4%
Small and medium-sized firms, Italy15.7%
ISTAT, Imprese e ICT, December 2025; Eurostat, December 2025. Striped: the EU average.

The firms themselves name the main reason. Among those that considered AI without adopting it, 58.6% cite a lack of skills, ahead of unclear legislation (47.3%), poor data quality (45.2%) and costs (43%).18 The Politecnico di Milano's observatory on digital innovation in small and medium-sized firms reaches the same point by another route: 76% of SMEs have not invested in AI and do not plan to, and only 7% have started structured AI training programmes for their staff.20

For the Bank of Italy the lack of skills is the single biggest obstacle, but not the only one. The study advises against "poorly targeted subsidies" and proposes something else: dedicated bodies that guide firms through adoption, specialist suppliers able to offer solutions to small firms too, public procurement that creates demand, clear rules, shared data and easier access to computing power.15

What I think

I think that in Italy we discuss AI starting from the wrong question. We ask how many jobs it will destroy, while for the next twenty-five years the problem will be finding people to do the work there is. In the piece on layoffs blamed on AI I wrote that the label often covers cuts decided for other reasons. Here the point is a different one. In a country losing three million workers, a town hall office, a hospital ward or a small factory risks closing because nobody answers the job advert.

I do not think AI will solve the problem on its own, and I am wary of anyone who presents it that way. It does not change a bed or keep a person with Alzheimer's company. It can, though, give a nurse back part of the time she spends writing and recording, or take repetitive paperwork off a council clerk. In a thirty-person firm it can help prepare a quote or check the quality of a part without having to hire someone who cannot be found. In a country with more and more older people and fewer and fewer people at work, that is what keeps services open.

The risk I see concerns those left behind. Large Italian companies use AI almost as much as large European ones. The delay lies in small firms, which employ millions of people and often have nobody with the time to learn a new technology. If AI only reaches big companies, the distance grows between those who manage to work with fewer people and those who close. That is why it matters who knows how to use it: innovation is still about people, even when there are few of them.

What firms and institutions can do

  1. TrainTraining during working hours58.6% of the firms that gave up on AI cite skills. It takes paid hours tied to a real task in the firm, with someone supporting the people who are learning.
  2. GuidePractical help for small firmsVouchers tied to a specific project, with a competence centre or a supplier that stays with the firm until the tool is in daily use, as the Bank of Italy suggests.
  3. Free upOffice work firstIn public bodies and in health care, AI belongs where things are written, filed and checked, to give hours back to front desks and wards.
  4. WelcomeMake immigration workIf fewer than 8 quota places in 100 become a residence permit, the problem is the procedure. Technology is not enough without new people.

The next figure to watch arrives in December. Eurostat plans to update its data on AI use in European firms in December 2026, and that will show whether Italy's small firms have started to catch up.19

Notes and sources

  1. ISTAT, Previsioni della popolazione residente, delle famiglie e delle forze di lavoro, base 1/1/2025 ↗, 31 August 2026. ↩
  2. F. Seghezzi, Domani, L'Italia avrà meno lavoratori, l'Ia da sola non ci salverà ↗, 2 September 2026. ↩
  3. Eurostat, Population structure and ageing ↗, data extracted on 2 February 2026. ↩
  4. ISTAT, Indicatori demografici, anno 2025 ↗, 31 March 2026. ↩
  5. ISTAT, Natalità e fecondità della popolazione residente, anno 2024 ↗, 21 October 2025. ↩
  6. CNEL, Tra il 2011 e il 2024 emigrati dall'Italia 630mila giovani ↗, 30 December 2025. ↩
  7. Ministero dell'Interno, Decreto flussi 2026-2028, previste 497.550 quote in tre anni ↗, 20 October 2025. ↩
  8. Il Post, Il decreto flussi funziona sempre peggio ↗, 21 February 2026. ↩
  9. Unioncamere, Lavoro: 539mila entrate programmate dalle imprese a ottobre (+3,7%) ↗, 19 October 2026. ↩
  10. Ministero del Lavoro e delle Politiche Sociali, Bollettino Excelsior: 520mila assunzioni previste a ottobre ↗, 10 October 2025. ↩
  11. FNOPI and Scuola Superiore Sant'Anna, Rapporto sulle professioni infermieristiche, prima edizione ↗, May 2025. ↩
  12. INPS, Osservatorio sui lavoratori domestici: i nuovi dati ↗, 18 June 2026. ↩
  13. L. Fanoli, M. Stentella, FORUM PA, La PA al bivio generazionale: in un anno +33% di dipendenti pubblici under 30 ↗, 3 July 2026. ↩
  14. ISTAT, Misure di produttività, anni 1995-2024 ↗, 12 December 2025. ↩
  15. L. Bellomarini and others, Bank of Italy, L'adozione dell'intelligenza artificiale: effetti su produttività e politiche a sostegno ↗, Questioni di Economia e Finanza no. 1009, June 2026. Translation mine. ↩
  16. The Japan News (Yomiuri Shimbun), Nursing care workers in Japan fall to 2.13 million; demand growing higher, but pay remains low ↗, republished by Asia News Network, 19 March 2025. ↩
  17. Y. S. Lee, T. Iizuka, K. Eggleston, Robots and Labor in Nursing Homes ↗, NBER Working Paper 33116, November 2024. ↩
  18. ISTAT, Imprese e ICT, anno 2025 ↗, 15 December 2025. ↩
  19. Eurostat, Use of artificial intelligence in enterprises ↗, data extracted in December 2025. ↩
  20. Osservatorio Innovazione Digitale nelle PMI, Politecnico di Milano, PMI tra crisi, energia e AI: il rischio è sottovalutare la trasformazione digitale ↗, 21 May 2026. ↩