Where the World's Workers Are, and How Exposed They Are to AI

Why look at the biggest workforces

Rankings of AI exposure usually list countries as equals. Luxembourg counts as much as India. That is useful for comparing national economies, but it hides a basic fact: the world's workers are not spread evenly. Two countries hold over a quarter of all workers. This article weights countries by the number of people who work in them. 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, built from every occupation group's score weighted by its workers. The scores are model-estimated and not measurements.

A data note. Our files cover 206 countries with both employment and exposure, and about 2.91 billion workers. Ghana's stored employment figure (48.4 million) is bigger than its population (about 35 million), which cannot be right, so we left Ghana out of the numbers below and have flagged it for checking. That leaves 205 countries and 2.86 billion workers.

The 15 biggest workforces

The picture is two clusters. Four of the 15 biggest workforces score 4.5 or above: Germany (5.30), the United States (5.06), Japan (4.92) and Russia (4.62). China (4.48) sits just below. The other big ones, India, Bangladesh, Vietnam, Pakistan, Nigeria and Indonesia, score between 3.21 and 3.44. These are large farm-heavy and informal workforces. Our informality analysis explains the link.

The world through its workers

Look at the first row. About 43 of every 100 workers in our data live in a country scoring under 3.5, which is 51 countries including India, Indonesia, Pakistan, Nigeria and Bangladesh. At the other end, only about 12 of every 100 live in countries scoring 5.0 or higher, such as Germany, Switzerland and the United States. Counting countries gives a different picture: 25 countries score 5.0 or more, and they are all counted equally with the 51 low-exposure ones.

That is why the two averages differ. Weighted by workers, the world scores 3.94. Counted as 205 equal countries, it scores 4.11. The gap is small, about 0.16 points, but it matters for headlines. A statement such as "the average country is exposed at 4.1" describes many small economies, while "the average worker is exposed at 3.9" describes where people actually work.

What this means

Two things follow. First, the number of people in the most exposed economies is smaller than headlines about "countries" suggest, about 0.33 billion workers. Second, many of the world's workers, in South and Southeast Asia and Africa, are in places where exposure is lowest today and where informal work is most common. For them, the more immediate AI questions are access and tools, not displacement. As incomes rise and economies shift from farms toward offices, the same data says their exposure would rise. That is a direction of travel from the cross-country pattern (see HDI and AI exposure), not a forecast for any one country.

Limits to keep in mind

• Employment data are of different years and sources. Some are modelled ILO estimates. Totals are the latest per country, not one common year.
• Ghana is excluded pending a check of its employment figure. Other countries could have errors we have not found. Ghana is the only one whose employment is larger than its population.
• Country scores hide variation. A country's score is an average. Inside a low-scoring country, there are still many exposed workers.
• Scores are model-estimated. They say how much of a job could be done with AI, not how much will be.

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