Ethiopia vs Eritrea: AI Job Risk Across a Contested Border

Two of the world's least AI-exposed economies

Ethiopia and Eritrea share a 912-kilometre border, a common history as one country until Eritrea's independence in 1993, a devastating border war from 1998 to 2000, and two of the lowest AI exposure profiles in the WorldJobsData dataset. The comparison is striking not because the two countries are dramatically different from each other - their scores are close - but because both sit at an extreme of the global AI risk spectrum that receives little attention.

The countries that dominate AI workforce commentary are the US (143.1 million workers, average 5.07/10), Germany (42.1 million workers, average 5.30/10, Eurostat), and Japan (70.6 million workers, average 4.92/10, ILO ILOSTAT 2023). Ethiopia and Eritrea sit at 2.94/10 and 3.36/10 respectively. Understanding why requires looking at where the workers actually are.

Ethiopia's ILO ILOSTAT 2021 data covers 36,631,600 workers. Eritrea's ILO ILOSTAT 2025 data covers 691,500 workers. The scale difference is 53x. Ethiopia is the second-most populous country in Africa. Eritrea has a population of approximately 3.5 million, with a large diaspora and significant military mobilisation reducing the civilian workforce.

Ethiopia: a workforce built on agriculture

The single most important fact about Ethiopia's AI exposure profile is that 13,879,100 of its 36,631,600 tracked workers - 37.9% - are skilled agricultural, forestry, and fishery workers, scoring 3.0/10 on AI exposure (ILO ILOSTAT 2021, CC BY 4.0). A further 10,163,800 workers (27.7%) are in elementary occupations scoring 2.0/10. Together, these two groups - agriculture and basic physical labour - account for 65.6% of the Ethiopian workforce. They are the least AI-exposed occupation categories in any economy.

Ethiopia's next largest group is service and sales workers at 4,831,700 (13.2%), scoring 3.5/10. Then plant and machine operators at 2,684,900 (7.3%), scoring 3.0/10. Craft and trades workers add 1,709,700 (4.7%) at 2.5/10. These are all below the global average AI exposure score.

The high-exposure groups are small in Ethiopia's case. Professionals account for 1,434,400 workers (3.9%) scoring 6.5/10 - roughly the same proportion as Eritrea's. Clerical workers are just 428,700 (1.2%) scoring 8.5/10. Ethiopia's managers total 386,100 (1.1%) at 5.5/10, and technicians 1,113,100 (3.0%) at 5.5/10. These groups are real and their task exposure is high, but they represent a small minority of a workforce overwhelmingly shaped by agricultural and subsistence economy needs.

Sources: ILO ILOSTAT (CC BY 4.0) - Ethiopia 2021 data year, Eritrea 2025 data year. AI exposure scores are WorldJobsData research-based estimates per ISCO-08 group, informed by Frey-Osborne (Oxford), OECD, and IMF automation susceptibility studies. Wages are not available at ISCO-1 level for either country in the ILO ILOSTAT series.

Eritrea: a smaller but structurally similar workforce

Eritrea's 691,500 workers (ILO ILOSTAT 2025, CC BY 4.0) break down similarly to Ethiopia's in structure but with some notable differences. Skilled agricultural workers are still the largest group at 244,000 (35.3%), but Eritrea has a proportionally larger craft and trades sector at 111,600 workers (16.1% vs Ethiopia's 4.7%). This craft concentration - construction, manufacturing, repair trades - reflects Eritrea's post-independence economic reconstruction era, when infrastructure rebuilding absorbed a large share of the workforce.

Eritrea's elementary occupations at 82,200 (11.9%) are proportionally much smaller than Ethiopia's 27.7%. Eritrea's service and sales at 97,800 (14.1%) is proportionally slightly larger. Eritrea has a higher share of professionals (6.5% vs Ethiopia's 3.9%), which partly explains the marginally higher average score of 3.36/10.

Eritrea's government operates one of the world's most restrictive economies. State ownership dominates key sectors. An indefinite national service programme, in place since the late 1990s, effectively conscripts a significant portion of the adult population into government-directed work at below-market rates. The ILO ILOSTAT 2025 figure of 691,500 may undercount the full labour force because national service workers' classification in official statistics is unclear.

Why the scores are almost irrelevant at this income level

AI exposure scores measure task susceptibility - whether an occupation's core tasks fall within current AI capability. They do not capture the economic conditions required to actually deploy that capability. In high-income countries, AI deployment is driven by labour cost savings, competitive pressure, and available capital. In Ethiopia and Eritrea, each of those three drivers fails.

Labour costs in both countries are among the world's lowest. Ethiopia's GDP per capita of $933 (World Bank 2025) implies an average worker income well below $1,000 per year. An enterprise software licence, cloud AI subscription, or automation system capable of replacing even a single clerical worker costs multiples of that worker's annual salary to deploy and maintain. The ROI calculation does not work at Ethiopia or Eritrea's wage levels. Employers who might want to automate lack both the capital and the financial incentive.

Competitive pressure to adopt AI requires operating in markets where competitors are adopting it. Ethiopia's largest employers are in agriculture, government, and basic services - sectors with limited exposure to global competitive pressure of the kind that drives AI investment in US financial services or German manufacturing.

Both countries also face acute infrastructure constraints. Internet penetration in Ethiopia stood at approximately 27% in 2024 (World Bank data). Eritrea's internet access is among the lowest in the world, with the government restricting access as a matter of policy. Cloud AI deployment requires reliable electricity and internet - neither can be assumed outside major urban centres in either country.

Economy context: Ethiopia and Eritrea side by side

The table below uses World Bank Open Data (CC BY 4.0, 2025 reporting year) and UNDP Human Development Report 2025 (2023 data year, licence CC BY 3.0 IGO).

Ethiopia's HDI of 0.497 (UNDP HDR 2025, rank 180 out of 193) and Eritrea's 0.503 (rank 178) place both countries in the "low human development" category. Life expectancy, education, and per capita income all sit near the global bottom. Eritrea's HDI is marginally higher despite lower GDP per capita, reflecting historically stronger government investment in basic health and education outcomes relative to income - a pattern linked to the post-independence mobilisation period under the ruling PFDJ party.

The geopolitical context: from shared country to contested border

Eritrea was a province of Ethiopia from 1952 under a federation that became full annexation in 1962. The Eritrean independence movement fought a 30-year war until independence was achieved in 1993. A border demarcation dispute led to a catastrophic war between 1998 and 2000, killing an estimated 70,000 to 100,000 people across both sides and displacing millions. The boundary was never fully implemented under the 2000 Algiers Agreement, leaving the two countries in a state of frozen conflict for two decades.

A peace agreement was signed in July 2018, ending the formal state of hostility, and Eritrean President Isaias Afwerki and Ethiopian Prime Minister Abiy Ahmed exchanged visits. This opening briefly raised expectations for economic integration. However, the outbreak of the Tigray conflict in November 2020 complicated the relationship - Eritrean forces fought alongside Ethiopian federal forces in Tigray, a fact that generated significant international controversy and human rights scrutiny. The peace dividend from 2018 did not translate into meaningful economic integration by 2025.

From an AI workforce perspective, this geopolitical history matters for several reasons. Both countries have experienced conflict-driven displacement that reshapes workforce distribution in ways official statistics cannot track in real time. Eritrea's national service, extended indefinitely since the late 1990s, functions as a state labour allocation system that distorts market-driven employment patterns. Ethiopia's multiple active internal conflicts as of 2021 (when its workforce data was collected) had already displaced millions.

The safest jobs from AI in both countries

In both Ethiopia and Eritrea, elementary occupations score 2.0/10 - the lowest possible AI exposure in the ISCO-08 framework used by ILO ILOSTAT. Ethiopia's 10,163,800 workers in elementary occupations represent 27.7% of the tracked workforce. Eritrea's 82,200 elementary workers are 11.9%. These are workers in building cleaning, refuse collection, street sales, agricultural labour, and manual freight handling - tasks that require physical presence in variable environments and resist current AI replacement.

Eritrea's craft and trades sector at 16.1% is notably large relative to its size. This concentration in construction, repair, and manufacturing trades means Eritrea has proportionally more workers in the 2.5/10 score range than its larger neighbour. This is not primarily an AI risk buffer - it reflects the specific economic history of post-independence reconstruction and the government's direction of labour into infrastructure projects.

What this means for workers in both countries

For Ethiopian and Eritrean workers, the AI risk question sits far outside any realistic near-term planning horizon. The 428,700 Ethiopian clerical workers at 8.5/10 task exposure and 20,200 Eritrean clerical workers at 8.5/10 are in occupation categories that are highly susceptible to AI automation - but susceptibility and deployment are different things. The clerical worker in Addis Ababa processing government paperwork, or the administrative staff member in Asmara tracking state enterprise records, faces no plausible risk of AI-driven displacement within a 10-year window when their employer lacks the capital, connectivity, and regulatory framework for enterprise AI adoption.

The more pressing risks for workers in both countries are the ones that do not appear in AI exposure scores at all: conflict displacement, food insecurity (Ethiopia experienced severe drought and conflict-driven famine risk in multiple regions during 2021-2024), emigration pressure, and the challenge of building any form of sustainable livelihoods in economies at the low-income threshold. Ethiopia's economy has grown rapidly in aggregate terms under Prime Minister Abiy Ahmed's administration, but from an extremely low base and with significant distributional unevenness.

Ethiopia's 13.9 million agricultural workers and Eritrea's 244,000 agricultural workers score 3.0/10 on AI exposure. They are safer than clerical workers on the task dimension. But they also face the most acute climate change exposure - rainfall variability, drought frequency, and temperature increases that bear on crop yields and livelihoods far more directly than any AI deployment scenario over the same time horizon.