Armenia vs Azerbaijan: AI Job Risk in the South Caucasus

Two Caucasus economies on diverging paths

Armenia and Azerbaijan sit side by side in the South Caucasus, share a post-Soviet economic inheritance, and face broadly similar AI exposure profiles on paper. Both have approximately 30% of employment in agriculture, both score 8.5/10 for clerical workers, and both have weighted average AI scores in the 3.85-4.01 range per ILO ILOSTAT data. The surface resemblance is where the similarities end.

Azerbaijan's economy runs on oil. The Baku oil fields and offshore Caspian production give Azerbaijan a GDP per capita of $7,411 (World Bank 2025) - well below its full productive potential because resource wealth is not evenly distributed through the labour market. But that capital gives the Azerbaijani state capacity to fund infrastructure, digital government initiatives, and AI adoption in ways that a purely labour-income economy cannot. Armenia, by contrast, has almost no natural resource base and has responded by building one of the Caucasus's most capable technology sectors, with a sizeable diaspora - concentrated in Russia, the US, and France - providing both investment capital and knowledge transfer.

The geopolitical context is impossible to ignore. Azerbaijan's military recapture of Nagorno-Karabakh in September 2023, and the subsequent exodus of over 100,000 ethnic Armenians from the region, represents a structural shock that neither country's ILO data captures. Armenia's 2017 survey and Azerbaijan's 2014 survey predate this event by years. The AI scores in this analysis are structural estimates - what the ISCO-08 occupation composition implies about AI susceptibility - not a picture of the workforce as it stands in 2026.

Side-by-side: all occupation groups compared

The table below shows every occupation group for which ILO ILOSTAT data exists for both countries, with AI exposure scores and employment figures. Score methodology follows ISCO-08 occupation groups, informed by Frey-Osborne (2017), OECD, and IMF task-level automation research. Wage data is available for Armenia (2017) but not for Azerbaijan in the ILO dataset.

Sources: ILO ILOSTAT (CC BY 4.0) - Armenia Labour Force Survey 2017 for employment and wage figures. ILO ILOSTAT (CC BY 4.0) - Azerbaijan Labour Force Survey 2014 for employment figures. Wage data was not available for Azerbaijan in the ILO dataset. "n/a" denotes missing data, not zero. Armenia wages are median annual earnings in USD per ILO ILOSTAT EAR_4MTH_SEX_OCU_CUR_NB, 2017.

Armenia: why the tech pivot changes the AI risk picture

Armenia's ILO 2017 data shows a workforce of approximately 1.0 million workers (1,011,720 to be precise), with skilled agricultural workers as the largest single group at 306,600 workers (30.3% of employment). But the headline story about Armenia in 2026 is not its agricultural sector - it is the tech sector that has grown substantially since 2017 and is not captured in the ILO figures.

Armenia's information technology sector has been a deliberate economic priority since the early 2010s, with the state-backed Enterprise Incubator Foundation (EIF) and a well-organised diaspora in Silicon Valley providing capital and talent pipelines. By the mid-2020s, IT services had become one of Armenia's top export categories. Yerevan hosts a cluster of software development companies serving European and US clients - a concentration of exactly the professional roles (scored 6.5/10 on AI exposure) that are most susceptible to AI augmentation. The 2017 ILO figure of 146,600 professionals almost certainly understates what exists in 2026.

This is the core tension in Armenia's AI risk profile. The sector that has driven economic growth since 2017 - technology and professional services - is the one that scores highest on AI exposure after clerical work. Armenia's HDI of 0.811 (rank 69, UNDP Human Development Report 2025, 2023 data year) and GDP per capita of $9,474 (World Bank 2025) reflect a more developed knowledge economy than its small population might suggest. The AI tools that threaten professional work in Western Europe are increasingly economically rational to deploy in Yerevan, where software engineers earn significantly more than the broader labour market average.

Armenia's 12.87% unemployment rate (World Bank 2025) and a Gini coefficient of 27.4 (World Bank 2024) point to a relatively equal but slack labour market. Workers displaced from clerical or professional roles by AI face a limited domestic job market. However, Armenia's strong diaspora networks provide a partial buffer that does not appear in the data - workers who lose positions in Yerevan's formal economy have historically been able to move toward diaspora-connected opportunities or emigrate.

Azerbaijan: oil wealth, low average exposure, and concentrated risk

Azerbaijan's ILO 2014 data shows a workforce of approximately 4.6 million workers (4,602,900), with skilled agricultural workers again the dominant group at 1.37 million (29.8%). The weighted average AI exposure score of 3.85/10 is pulled down by this large agricultural base - agriculture scores 3.0/10 on AI exposure, low enough to significantly anchor the national average.

But Azerbaijan's oil economy creates a structural feature that the occupation-level data alone does not capture. The energy sector - which drives a disproportionate share of state revenue and formal employment in Baku - involves a mix of professional roles (6.5/10 AI exposure) and highly specialised technical roles that are less susceptible to current AI capabilities. Azerbaijan's GDP per capita of $7,411 (World Bank 2025) is somewhat misleading as a measure of individual income given resource-revenue concentration, but the state has used oil revenues to fund digital government infrastructure, STEM education, and modernisation initiatives that affect AI adoption rates.

Azerbaijan's 226,000 clerical workers (4.91% of 2014 employment) represent a concentrated risk. These are the workers scoring 8.5/10 - the same peak score as every other country in the WorldJobsData dataset. In an oil-exporting economy with significant state employment, many of these clerical workers sit in government ministries and state-owned enterprises where AI adoption timelines are driven by policy decisions, not pure commercial logic. That can either delay displacement (bureaucratic inertia) or accelerate it (centralised digital transformation mandates) depending on political will.

Azerbaijan's HDI of 0.789 (rank 81, UNDP Human Development Report 2025, 2023 data year) and GNI per capita PPP of $20,668 (UNDP) places it in the high human development category, though below Armenia's HDI of 0.811. The female labour force participation rate of 60.76% (World Bank 2025) is notably higher than Armenia's 51.06% - an indicator of broader labour market inclusion that partially offsets concerns about concentrated AI risk in clerical roles, where female workers are more heavily represented (80.59% female in Azerbaijan's clerical sector per ILO 2014 data).

The Nagorno-Karabakh factor: what the data cannot tell you

No analysis of Armenia versus Azerbaijan is complete without acknowledging the September 2023 military operation in which Azerbaijan retook full control of Nagorno-Karabakh (also known as Artsakh), prompting the displacement of over 100,000 ethnic Armenians from the region. This is one of the largest rapid population movements in recent European or Caucasus history.

The economic implications for both countries are material and unquantified in any ILO data. For Armenia, 100,000 additional people - largely from rural and semi-urban backgrounds - entered a labour market of roughly 1 million workers already operating at 12.87% unemployment (World Bank 2025). The strain on housing, services, and the labour market has been significant. For Azerbaijan, the economic integration of Karabakh - including reconstruction investment estimated in the billions of dollars - represents both a fiscal commitment and a potential source of new employment demand, particularly in construction, infrastructure, and plant operations (groups scoring 2.5-3.0/10 on AI exposure).

From an AI job risk perspective, the Karabakh displacement primarily adds workers to Armenia in occupation categories that score low on AI exposure - agricultural workers, craft workers, elementary occupations. This would, if anything, push Armenia's weighted average AI score slightly downward from the 4.01/10 figure derived from 2017 data. It does not meaningfully change the structural AI risk profile, but it does increase the number of workers in vulnerable economic positions with limited safety nets.

Economy context: Armenia and Azerbaijan side by side

Sources: World Bank Open Data (CC BY 4.0) for GDP, unemployment, labour force participation. UNDP Human Development Report 2025 (2023 data year) for HDI, GNI. ILO ILOSTAT (CC BY 4.0) for employment totals. WorldJobsData AI exposure scores are research-based estimates, not official forecasts.

What this means for workers in both countries

For Armenian workers in clerical and professional roles, the AI risk signal is real and probably underestimated by the 2017 data. Armenia's growing technology sector means that the 146,600 professionals counted in 2017 are likely a larger group today, and they work for companies that are already evaluating AI tools for code generation, documentation, and analysis. The 34,400 clerical workers scoring 8.5/10 are in the most universally vulnerable occupation group globally - the same pattern appears in the US (143M workers), the UK (34M workers), and every other country in WorldJobsData's 206-country dataset. Armenia's 12.87% unemployment rate means there is limited labour market slack to absorb displaced workers.

For Azerbaijani workers, the picture is more complex. The large state-employment sector provides some insulation - public sector AI adoption moves slower than private sector deployment. Azerbaijan's 5.46% unemployment rate (World Bank 2025) is significantly healthier than Armenia's, giving displaced workers a better chance of finding alternative employment. Azerbaijan's oil revenues fund retraining and diversification initiatives that can, in principle, redirect workers away from high-exposure roles. Whether that investment reaches individual workers rather than remaining in infrastructure and headline GDP is the key uncertainty.

Both countries share a structural characteristic: agricultural workers representing roughly 30% of employment provide a natural anchor that keeps the national AI exposure average down. These workers - scoring 3.0/10 on AI exposure - are not near-term AI targets. But the workers above them in the occupation hierarchy, from clerical to professional, face the same global AI pressure as their counterparts in wealthier nations, with less social safety net protection beneath them. For a deeper look at the occupation-level data for each country, explore the Armenia individual analysis and the Azerbaijan individual analysis. For regional comparison, Georgia's workforce data provides the closest third reference point in the South Caucasus.