Richer Countries Face More AI Exposure. Here Is Why.

What the Human Development Index means

The Human Development Index, or HDI, is a single number between 0 and 1 published by the United Nations Development Programme (UNDP). It combines three things: how long people live, how much schooling they get, and how much money a person earns on average. A higher number means a more developed country. UNDP groups countries into four tiers using fixed cut-offs: low (below 0.550), medium (0.550 to 0.699), high (0.700 to 0.799) and very high (0.800 and above).

We use the latest report, the UNDP Human Development Report 2025, which covers data from 2023. It has HDI for 177 of the 206 countries on this site. AI exposure is our own score from 0 to 10 for how much of a country's work could be done or helped by AI. We calculate it for each country by weighting every occupation group's score by how many people work in it. Higher means more of the country's jobs are the kind AI can already help with, such as writing, calculating, and handling information.

The big picture: more development, more exposure

Read the chart from left to right. On the left are countries with low HDI: 22 of the 26 score between 3 and 4. On the right are the most developed countries: 41 of the 63 score between 4.5 and 5.5. The dashed line shows that every extra 0.1 of HDI goes along with about 0.37 more points of exposure, on average.

The four tiers side by side

Two things stand out. First, the climb is not even. Going from low to medium HDI adds only 0.15 points, but going from high to very high adds 0.78. The big jump happens once a country has a large office-based workforce. Second, the "weighted by workers" column puts more weight on big countries. It shows the same ladder: 3.21 for low, 3.27 for medium, 4.00 for high and 4.88 for very high.

Where do the workers actually sit? Our country files cover about 1.0 billion workers in high HDI countries (the biggest group: it includes China, Brazil, Mexico and Indonesia), about 0.9 billion in medium HDI countries (this tier includes India), 680 million in very high HDI countries and 246 million in low HDI countries. Employment totals come from the ILO and national statistics offices, and years differ by country.

Why development brings exposure: the job mix

AI language tools are good at work done with words and numbers on a screen. They are poor at harvesting a field, laying bricks or caring for a patient. So a country's score mostly depends on how many people do which kind of work. We counted workers in two groups for the 156 countries where we have all nine major occupation groups:

The pattern is large. In low HDI countries farming is the biggest employer, with about 43 of every 100 workers. In very high HDI countries it is about 5. Office-type jobs go the other way, from about 12 in every 100 to about 49. Because our scores put clerical and professional work at the high end and farm work at the low end, development shifts the whole country upwards.

This is also why the tiers are the story of development and not of AI: the technology is the same everywhere, but the work it can touch is not.

What the richest countries look like

The ten highest-HDI countries all score between 4.95 and 5.44. All ten are above 4.9. 6 of the 63 very high HDI countries score below 4.0, which shows that the link is strong but not perfect.

The exceptions: countries off the line

To find the exceptions, we draw the straight-line trend and measure how far each country sits above or below it. A country far above the line is more exposed than its development level suggests. Far below means less.

More exposed than their HDI predicts

Luxembourg is the highest-scoring country of all (5.87) because about 45% of its workers are professionals, such as finance and business specialists, and another 17% are technicians. The Netherlands, Singapore and Belgium also sit above the line for the same reason: a large professional workforce. Djibouti, Somalia, Tuvalu and Nauru are small or low-data countries whose reported job mix has an unusually large share of professional, clerical or service work. A few occupation groups can move the average a lot in such countries, so treat them as flags to check, not firm conclusions.

Less exposed than their HDI predicts

These countries are richer than their exposure suggests because their job mix still leans toward manual work. In Kuwait, about 24% of workers are in crafts and trades and 27% in elementary jobs. In Vietnam, 15% farm and 24% are in elementary jobs. Ecuador (21% farming, 26% elementary) and Morocco (23% crafts, 30% elementary) look similar. The United Arab Emirates is below the line too: despite an HDI of 0.940, about 18% of its workforce is in crafts and trades and 17% in elementary jobs.

We left out the 21 countries that report fewer than all nine major occupation groups from these two lists. In those countries a missing group can distort the average. South Sudan, for example, reports seven groups, and 75% of its workers fall in services and sales.

What this means for you

If you live in a high-income country, a high exposure score does not mean your job will vanish. It means more of the tasks around you can be sped up by AI tools, which can raise pay for some workers and put pressure on others. If you live in a lower-income country, a low score is not a free pass. It mostly reflects that fewer people are in jobs that AI tools can reach yet. As incomes grow and economies move from farms to offices, the score tends to rise, because that is the path every richer country has already taken.

The practical point for workers and planners is timing. Countries in the middle tiers are the ones whose job mix is changing fastest, so their exposure may climb faster than the data in this article shows. We will refresh this analysis whenever UNDP publishes a new report.

Limits to keep in mind

• Correlation is not cause. Development and exposure move together, but we have not shown that one drives the other. Both come from the kind of economy a country has.
• Scores are model-estimated. They describe how much of an occupation could be done with AI, not how much will be.
• Data years differ. HDI is for 2023. Employment data comes from different survey years by country.
• Small data sets. Some countries report only seven or eight occupation groups, which can move their average. We checked that our result holds when we limit the data to the countries with all nine groups: the correlation is 0.84 there.

Spotted an error or know a better source? Report it and we will check it against the source data.