Israel vs Iran: Which Economy Faces More AI Job Risk?

Two economies, one score ceiling, two very different realities

Israel and Iran share more in the WorldJobsData dataset than most observers would expect. Both countries report through ILO ILOSTAT (CC BY 4.0) with 2024 as their data year. Both have clerical support workers as their highest AI exposure group, both scoring 8.5/10. And both have elementary occupations as their safest group at 2.0/10. On those two data points alone, the pair look identical.

The divergence is in everything else. Israel's workforce of 4.07 million workers is concentrated overwhelmingly in high-exposure white-collar roles. Professionals alone - a group that scores 6.5/10 on AI exposure - make up 32.7% of the entire Israeli workforce (1.33 million workers, ILO ILOSTAT 2024). Technicians and associate professionals add another 15.2% (619,755 workers, 5.5/10). Service and sales workers at 19.6% of the workforce score 3.5/10. The result: an economy where knowledge work dominates, and knowledge work is exactly what large language models are designed to augment and replace.

Iran's 24.07 million workers look radically different. Craft and related trades workers are the single largest group at 19.0% (4.57 million workers, 2.5/10 AI exposure). Plant and machine operators add 14.3% (3.45 million workers, 3.0/10). Elementary occupations account for 14.2% (3.42 million workers, 2.0/10). Agriculture runs at 10.5% (2.53 million workers, 3.0/10). Collectively, the lowest-exposure half of the ISCO-08 occupation hierarchy accounts for more than half of Iran's entire employed population. That structural composition produces a low weighted average, not because Iran's clerical workers are any safer from AI than Israel's - they are not - but because clerical workers are a much smaller share of the total workforce.

Side-by-side comparison: all major occupation groups

The table below compares Israel and Iran across ISCO-08 major groups. Israel wage data is from ILO ILOSTAT EAR_4MTH_SEX_OCU_CUR_NB (wage year 2021, reported in USD). Iran wage data is not available from ILO ILOSTAT in USD terms - this is partly a consequence of international sanctions and currency volatility making reliable USD conversion impractical.

Note: Israel's AI exposure scores are identical to Iran's across all occupation groups because both countries are scored using the same ISCO-08 task-level methodology. The occupation groups themselves are standardised internationally. What differs is how much of each national workforce sits in each group - and that composition difference produces the 1.37-point gap in weighted averages.

Why Israel's average exposure is so much higher

Israel has earned the label "Start-Up Nation" - a reference to its outsized tech sector relative to population. With a GDP per capita of $60,337 (World Bank, 2025) and an HDI of 0.919 ranking 27th globally (UNDP Human Development Report 2025, 2023 data year), Israel has the capital, the infrastructure, and the skilled labour base to build and deploy AI tools at scale. The country produces more Nasdaq-listed companies per capita than almost any other nation - a fact that directly shows up in ILO ILOSTAT 2024 data as an unusually large professional class.

Professionals - software engineers, financial analysts, scientists, researchers, lawyers, architects - make up 32.7% of Israel's 4.07 million workers. This group scores 6.5/10 on AI exposure (ILO ILOSTAT 2024, WorldJobsData scoring). When the single largest occupation group in a country scores 6.5/10, the weighted average for the whole economy will inevitably be pulled high. The US, for comparison, has professionals at roughly 30% of its workforce with the same 6.5/10 score and averages a similar 5.x/10 overall. Israel is in that same structural category: a knowledge-economy where the dominant form of work is precisely the kind of cognitive, information-processing activity that large language models are being deployed to augment and partially replace.

Clerical support workers - the highest-exposure group at 8.5/10 - represent 6.5% of Israel's workforce (265,235 workers, ILO ILOSTAT 2024). Their median annual wage of $27,483 (ILO ILOSTAT 2024, wage year 2021) is the lowest of any major occupation group in Israel except elementary occupations at $19,050. At that wage level, an employer in Israel - where average wages across the economy run above $54,000 per year (OECD Average Annual Wages USD PPP, 2024) - has a strong financial incentive to automate. The cost of a capable AI assistant is well below the cost of a junior clerical hire in a high-wage economy like Israel's.

The most AI-exposed occupations in Israel

Why clerical workers score 8.5/10 and not professionals

The 8.5/10 score for clerical support workers reflects the task profile of the work, not the level of skill or education involved. Clerical tasks - data entry, correspondence, scheduling, record-keeping, filing, basic customer inquiry handling - are overwhelmingly what machine learning researchers call "structured cognitive work": predictable inputs, predictable outputs, easily defined success criteria. These are exactly the task types that large language models handle most reliably as of 2026. A model trained on text can draft a routine email, fill in a spreadsheet, transcribe a meeting note, or route an inquiry faster and more cheaply than a human doing the same thing.

Professionals score 6.5/10 rather than 8.5/10 because professional work, even in knowledge-intensive fields, retains a substantial component of unstructured judgment. A software engineer does not just write code - they make architectural decisions, negotiate requirements, debug systems whose behaviour is not fully documented, and bear accountability for production systems. A lawyer does not just draft contracts - they advise on strategy, assess risk under uncertainty, and exercise judgment in adversarial contexts. AI tools can assist substantially with the structured sub-tasks within each of these roles. They cannot yet reliably handle the full-role judgment layer. That partial exposure is what 6.5/10 captures.

For Israel, this matters practically: the 265,235 clerical workers at 8.5/10 are more immediately at risk than the 1.33 million professionals at 6.5/10. In a high-wage, high-tech economy where AI tools are already widely deployed in enterprise settings, clerical automation is not a forecast - it is a current cost-reduction initiative at most large Israeli organisations.

Why Iran's disruption timeline is 8 to 12 years out

Iran's 24.07 million workers average 3.71/10 on AI exposure. That relatively low figure is genuine - it reflects a workforce genuinely structured around lower-exposure activities. But the structural buffer has a second driver that does not appear directly in any occupation score: the economic and political environment in Iran makes AI adoption substantially slower than in comparably-sized economies.

US-led international sanctions have restricted Iran's access to foreign technology since 1979, with the most sweeping restrictions in effect from 2018 onward. Access to US cloud infrastructure - AWS, Azure, Google Cloud - is prohibited for Iranian entities. Access to many AI model APIs, including those from Anthropic, OpenAI, and Google, is restricted. Hardware import restrictions limit the availability of the high-end GPUs required for domestic AI model training. The practical effect is that an Iranian company seeking to automate its clerical workforce faces substantially higher costs and more restricted tooling than an Israeli company doing the same - even setting aside every difference in wages and workforce composition.

Wages reinforce the delay further. Iran's GDP per capita is $3,924 (World Bank, 2025). At that income level, the labour cost of a clerical worker is sufficiently low that the business case for automation is marginal at best. Automation investment requires upfront capital cost that takes longer to recoup when the ongoing labour saving is small in absolute dollar terms. This is a well-documented pattern in development economics: automation tends to arrive later in lower-wage economies because the labour cost saving that justifies the capital investment is smaller. WorldJobsData's risk velocity score for Iran is 4.9/10, corresponding to "Disruption delayed - 7 to 12 years" (as of 2026). For Israel the score is 7.6/10, corresponding to "Disruption arriving - 3 to 7 years" at the dataset baseline, and closer to 1 to 3 years for the highest-exposure clerical roles given current deployment rates in Israeli enterprise.

The safest jobs in both countries

The lowest AI exposure occupation in both Israel and Iran is Elementary occupations at 2.0/10. This group covers cleaning workers, building and grounds maintenance, domestic helpers, manual material handlers, and basic labourers. Israel has 197,451 workers in this group (4.8% of the workforce) with a median wage of $19,050 per year (ILO ILOSTAT 2024). Iran has 3.42 million elementary workers (14.2% of the workforce) with no reliable USD wage figure available from ILO data.

The 2.0/10 score is not an argument that these workers face no labour market pressure. They face significant robotics risk - cleaning robots, automated material handling, autonomous delivery vehicles are all commercially deployed and improving. The 2.0/10 AI exposure score specifically reflects their low exposure to large language models and AI software tools. The tasks are physical, spatially variable, and do not involve generating or processing structured text. A cleaning robot that is commercially viable in a controlled airport corridor is not yet viable in the irregular geometries of a residential apartment or a construction site. That gap keeps the AI exposure score low even as the robotics risk score for elementary occupations sits at 5.5/10.

What this means for workers in each country

For Israeli workers in high-exposure roles - particularly the 265,235 clerical workers and the 1.33 million professionals - the question is not whether AI will affect their work but what specific tasks will be affected first and how quickly employers will restructure around those changes. Israel's labour market is relatively flexible by OECD standards. With a recovery resilience score of 7.3/10 ("High resilience - workers can pivot"), the WorldJobsData assessment is that Israeli workers have better-than-average capacity to adapt: high education levels (mean years of schooling: 13.5 years, UNDP HDR 2025), a strong tech ecosystem with adjacent opportunities, and an economy that has historically absorbed large workforce shifts. HDI of 0.919 (UNDP HDR 2025, rank 27 globally) reflects well-functioning health and education systems that support labour market transitions. The disruption will be real - particularly for younger clerical workers entering the labour market - but the structural conditions for managing it are comparatively strong.

For Iranian workers, the practical picture is more complicated. The 3.71/10 average suggests relatively low AI exposure, which is accurate for now. But Iran has 8.3% unemployment (World Bank, 2025) and a female labour force participation rate of just 14.0% (World Bank, 2025) - one of the lowest in the world. When AI tools do become more accessible in Iran, whether through sanctions relief, domestic model development, or informal channels, the clerical and professional workers who are already employed in high-exposure roles will face the same displacement pressures as their counterparts elsewhere. Recovery resilience is 6.4/10 ("Medium resilience - partial safety net"), reflecting a partial but incomplete capacity for workers to absorb and pivot from disruption. The low-wage, high-physical-work majority of the workforce has time on its side. The professional and clerical minority does not have that buffer - and Iran's HDI of 0.799 (rank 75 globally, UNDP HDR 2025) signals that supporting systems are less robust than in Israel.

Economy context: Israel and Iran side by side

Iran's Gini of 35.9 (World Bank, 2023) is actually lower than Israel's 38.3 (World Bank, 2022) - meaning income is distributed somewhat more equally in Iran than in Israel. But that relatively equal distribution is across a much lower income base. When AI disruption does arrive in Iran - most likely through the professional and clerical sectors first - the workers displaced will have less financial cushion, weaker retraining infrastructure, and a partial social safety net compared to what their Israeli counterparts can access. The disruption arrives later in Iran; it will likely land harder when it does.