Schooling · Skills · AI adoption

Italy runs a 2026 economy on a school designed in 1923

Slovak firms use AI more than Italian ones. Estonian firms one and a half times as much. Today it is AI. Tomorrow it will be another technology.

In 2025, 18.0% of Slovak firms used artificial intelligence. Italian firms: 16.4%. Estonia is at 23.4%.

Slovenia, Lithuania and the Czech Republic are also ahead of Italy. Thirty years ago these countries were on the other side of the Iron Curtain. Twenty years ago they were poorer than Italy.

This is what a country ends up with when its academic high school still teaches Latin instead of computing. A country that defends a school built before the Second World War.

Today it is AI. Tomorrow it will be another technology.

Italy sits 19th of 33. Here is the company it keeps.

So what does Slovakia do that Italy does not?

It is not money. Adjusted for income per head, the ranking barely moves. It is not firm size. It is not broadband. It is not the labour market. I tried them all.

One thing is left.

Skills

The thing that moves in step with AI adoption more than anything else is this: whether the working adults in a country can handle numbers.

Not the marks of fifteen-year-olds. Not the share of graduates. The OECD tests working adults directly, and the tasks are everyday things.

One released item gives you a card with the measurements of a room and asks which figures you need in order to work out how many tins of paint to buy. Another puts a wallpaper calculator in front of you that has returned an obviously wrong result and asks you to find the mistake.

Italy is on the line, not below it. It gets exactly what 253 points buys.

Estonia scores 287. The Czech Republic 278. Italy 253. The East did not overtake Italy with cheaper workers. It overtook Italy with better ones.

It is not the old people's fault

The Italian reflex is to blame demography. Old workforce, old habits, nothing to be done until they retire. It is a comfortable theory. That way the problem is nobody's fault and nobody's job.

The premise is true, and it is worse than most Italians think. Italy has the oldest workforce of every country the OECD examined. 52.5% of employed Italians are 45 or older. Not among the oldest. The oldest. Older than Japan, which is fourth at 49.5%.

If age were the problem, Italy would be the worst case of all. Here is what age actually does.

It tilts the line down, a little. Countries with older workers adopt slightly less. Then you start naming names.

Is Italy behind because its workers are old? Japan's are almost as old: 49.5% are 45 or older, against Italy's 52.5%. Japanese firms adopt almost twice as much.

Is Finland ahead because its workers are young? No. The Finnish workforce is older than Estonia's, Poland's, Ireland's and Canada's. Finnish firms adopt more than all four.

So is Poland behind because its workers are old? Poland has one of the youngest workforces in the group and the lowest adoption of all.

Three questions, three countries, no answer. Age shifts the number a little. Not enough to put Italy 19th.

So what separates Japan from Italy? Just put skills on the horizontal axis instead of age.

An old but skilled workforce adopts more than a young one that is not.

Japan is old, skilled, and adopts. Italy is old, unskilled, and does not adopt. Italy is the oldest and among the least skilled at the same time. In the numbers you only see the second thing.

One in three

Finding the mistake in that wallpaper calculator is a level 2 task. Level 2 is the threshold: the minimum the OECD considers necessary for any job at all. Below it, a person can do a one-step calculation in a context they know, and little else.

29.1% of employed Italians are below that threshold. Almost one in three. One in eighteen in Finland. One in thirteen in Japan. These are people in work. Not the unemployed, not pensioners. This is the workforce Italian firms have.

The second column is worse, and nobody looks at it. Only 7.6% of working Italians reach the top two levels. These are the people who can reframe a problem, read a model's output and notice the answer is wrong before anyone else. Finland has 35.6%.

Ask yourself what an Italian firm can build with 7.6%. Then ask yourself what a Finnish firm can build with 35.6%.

Nobody tested these people at school. They were tested at work, as adults, decades after leaving the classroom. So where does a workforce like this come from?

The system was designed in 1923

The Italian school was designed in 1923 by Giovanni Gentile, Mussolini's education minister. It was built as a hierarchy, on purpose.

At the top was the classical high school. Latin, Greek, philosophy. It was the only school that opened every university faculty. The science high school was second class by design. Its graduates could not enrol in humanities. They could not enrol in law. Technical and vocational institutes were for the workers' children. Nobody pretended otherwise.

Latin was the ticket to running the country. Knowing how things work got you a place inside something that someone else ran.

Italy has reformed that system many times. It has never overturned the hierarchy. Ask an Italian parent today which is the serious school. They give you Gentile's answer, one hundred and three years later.

One hundred and three years later, the science high school still teaches Latin and still does not teach computing. Only the applied-sciences track does, and to make room for it they had to drop Latin.

The adults the OECD tested left those schools ten, twenty, thirty years ago. They joined the firms that today cannot find anyone able to run a model.

Estonia rebuilt its school after independence, in 1991. Its working adults today score 287 on the numeracy test, against Italy's 253. Italy has had one hundred and three years and has never touched its hierarchy.

And afterwards nobody catches them up

A school can fail people at eighteen, and a country can still make up for it later. Some do.

In Sweden 38.2% of adults took a course of education or training last year. Denmark 31.0%. Finland 28.1%. Estonia 23.8%.

Italy: 10.8%. Below the European average of 13.7%. Twenty-third of thirty-five.

So the school hands over a workforce that cannot do the sums, and then nobody catches it up. Finland trains almost three times as many adults as Italy and ends up with almost five times as many at the top of the scale.

Two systems, one after the other, failing in the same direction.

A caveat. This is a comparison between countries, not an experiment. Firms that adopt AI also hire and train better people, so causation runs in both directions at once. That weakens any hard claim about what comes first. It does not make 7.6% look any better.

The cruel part

Here is the part that should be the easiest to fix, and is not.

AI is the best tool ever built for closing a gap of this kind. It explains, drafts, checks, teaches, and never loses its patience. Most Italian firms have no IT department and no training budget. An assistant at twenty euros a month should be the cheapest upgrade in the history of Italian industry.

Except that using it takes a base. Enough comfort with numbers to notice a wrong answer. Enough order to ask the question properly. Enough confidence to keep going after the first useless answer.

The firms that would gain the most are the ones least able to get anything out of it.

One last thing, and it is the one that worries me most. Everything written above is about firms with at least ten employees. The European survey only measures those. In Italy most firms are smaller. For those, nobody collects the data.

I would not bet a euro that it goes better down there.

The Italian debate about AI is a debate about incentives. Tax credits, cloud infrastructure, national champions, a ministry with a plan. None of that is what holds Italy back.

What holds Italy back left school decades ago and is sitting in an office right now, not knowing whether the number on the screen is plausible.

You do not get out of that with a budget law. It has to be taught, and Italy has spent one hundred and three years deciding that teaching means something else.

Today the thing Italy is behind on is AI. Tomorrow it will be something nobody has named yet, and waiting for it will be the same workforce.

The data

Everything written above comes from public sources. The links go to the datasets, not to summaries of the datasets: anyone can check the numbers, or dispute what I did with them.