What does the evidence say about AI and jobs?
There is no single global answer to whether artificial intelligence will create or eliminate jobs. Recent research points to a more precise question: which work tasks are exposed to generative AI, and do workers and businesses have the conditions needed to benefit from it?
An International Labour Organization (ILO) and World Bank working paper, published in March 2026, examines generative AI exposure across 135 countries. It finds that exposure is higher overall in advanced economies, while developing economies face a particular gap between the potential to use AI to assist workers and the digital access needed to realize that potential.
These are estimates of exposure - not counts of jobs already lost, nor a prediction that every exposed job will disappear.
Exposure is not the same as job loss
An occupation includes many tasks. Generative AI may be able to assist with or automate some of those tasks without replacing the whole role. Whether a job changes or disappears depends on factors beyond technical capability, including how employers adopt the technology and how work is organized.
The ILO–World Bank paper also cautions against assuming that a job title means the same tasks are performed in every country. Its analysis finds that conventional exposure measures can overstate potential impact in developing economies when they do not account for differences in the tasks workers actually perform.
This distinction matters for employers and policymakers: exposure indicators can help identify where work may change, but they should not be presented as confirmed job losses.
Three findings from the global research
1. Exposure is higher in advanced economies
The ILO–World Bank paper estimates that around 30–32% of employment in high-income countries is exposed to generative AI, compared with about 10–15% in low-income countries. The difference is driven mainly by greater exposure to automation in clerical and some professional work in higher-income economies. The paper finds that the potential for task augmentation is more evenly distributed across income groups.
“Exposed” describes potential interaction with the technology; it does not mean that all those jobs will be automated.
2. Digital access affects whether workers can benefit
The study identifies an imbalance: workers in jobs with automation exposure are often already connected to the internet, while workers whose tasks could be augmented by AI may lack reliable access.
Across countries with detailed data, the paper estimates that 441.8 million jobs fall into augmentation-oriented exposure gradients; around 66.9 million of those jobs are held by people without internet access. These figures describe potential exposure and an access gap - not realized productivity gains or job losses.
3. The employment implications differ across countries
In low- and lower-middle-income countries, the share of employment exposed to automation is smaller overall in the study. However, the paper notes that exposed jobs can include formal service-sector roles and entry-level clerical and administrative work. It also discusses the possibility that limited growth in such office-based work could constrain access to formal employment. That is a risk identified by the researchers, not a certainty about any one country.
What this means for the Arab region
The Arab region includes economies with very different resources, labour markets, and workforce conditions. It should not be treated as one uniform market.
A joint ILO–UN Economic and Social Commission for Western Asia (ESCWA) report, published in May 2026, uses strategic foresight to examine how AI could reshape employment in the region between 2025 and 2035. The report considers possible changes by sector and occupation, distributional effects across age and gender, and risks of exclusion for groups including older workers, persons with disabilities, migrants, and refugees. It presents alternative scenarios to help policymakers and social partners plan; these scenarios are not predictions of a single outcome.
The report’s focus on skills ranges from basic AI literacy to advanced technical and governance capabilities. For employers, the practical question is therefore not only whether to adopt AI, but also how workers will be prepared to use it and how access to opportunity will be addressed.
What can - and cannot - be concluded about Pakistan
The ILO–World Bank findings provide a useful global framework for considering Pakistan, but the income-group estimates do not establish how many Pakistani jobs will be affected or forecast employment outcomes for a particular Pakistani industry.
The World Bank’s World Development Report 2026: The Promise of Artificial Intelligence argues that AI’s development benefits depend on the capabilities it supports, how concentrated AI production is, and the complementary conditions that enable adoption. It highlights reliable electricity and internet, education, strong institutions, and data - including data in local languages - as important foundations. The report also distinguishes adopting existing tools, adapting them to local needs, and advancing frontier AI; each path requires different levels of investment and capability.
For Pakistan, these findings support a grounded conclusion: national or sector-wide claims about AI-driven job losses require local evidence. Global exposure estimates can inform questions for businesses and policymakers, but they should not be used as if they were Pakistan-specific employment forecasts.
Implications for customer-service and business-operations leaders
The following are practical recommendations informed by the research above, not findings that the cited reports tested specifically for NBS or any individual business:
- Assess tasks before drawing conclusions about roles. Map the work people actually do, then distinguish tasks that may be assisted by AI from work that requires human judgement, communication, or accountability.
- Check the conditions for adoption. Consider internet and electricity reliability, suitable systems and data, language needs, employee skills, and the organization’s ability to integrate AI into existing workflows. These align with the foundations emphasized in the World Bank report.
- Train for real work. Match AI training to employees’ responsibilities and provide clear expectations for checking outputs, handling exceptions, and escalating decisions.
- Evaluate service outcomes, not just speed. When assessing a change, track relevant measures such as accuracy, resolution quality, customer experience, workload, and operating cost. Treat these as evaluation criteria, not as assumed benefits of AI.
- Monitor employment effects with local data. Track changes in recruitment, task mix, training, and workforce outcomes rather than inferring job losses from exposure estimates.
For customer operations, this approach keeps AI adoption tied to service quality and workforce capability - not to headcount reduction as an assumed measure of success.
The central question is who can benefit
The 2026 research does not support a simple claim that AI will eliminate a fixed number of jobs in every region. It shows that exposure, task composition, internet access, and institutional readiness differ - and that these differences shape the risks and opportunities researchers identify.
For decision-makers in advanced economies, the evidence points to higher overall exposure, particularly in clerical and some professional work. For the Arab region, the ILO–ESCWA report offers scenarios and identifies workforce inclusion and skills as planning priorities. For Pakistan, the responsible approach is to apply global findings carefully and seek local evidence before making claims about employment in specific industries.
AI may be adopted across countries, but the ability to use it productively depends on skills, infrastructure, institutions, and the choices organizations make. Preparing for that reality starts with evidence about tasks and local conditions - not assumptions about inevitable job losses or guaranteed job creation.
Frequently asked questions
Does AI exposure mean that a job will be eliminated?
No. Exposure estimates identify work that may be affected by generative AI. They do not count confirmed job losses or establish that a whole occupation will disappear.
Are developing economies less affected by generative AI?
The ILO–World Bank paper estimates lower overall automation exposure in low-income countries than in high-income countries. It also identifies digital-access gaps that may limit workers’ ability to benefit from AI-assisted task augmentation. Lower exposure does not mean there are no risks or opportunities.
Does this research predict AI job losses in Pakistan?
No. The global and income-group findings are not a Pakistan-specific forecast. Claims about employment effects in Pakistan or a particular Pakistani industry require relevant local evidence.
- International Labour Organization and World Bank, Disruption without dividend? How the digital divide and task differences split GenAI’s global impact, 17 March 2026.
- International Labour Organization and UN ESCWA, Artificial Intelligence and Employment Futures for the Arab Region, 21 May 2026.
- World Bank, World Development Report 2026: The Promise of Artificial Intelligence.
This article distinguishes research findings from practical recommendations. Exposure estimates and scenario analyses should not be read as counts of jobs already lost or as precise forecasts for a particular country.
Related reading: AI adoption in Pakistan: applying global business lessons to local decisions · Back to all NBS insights