Artificial intelligence (AI) is reshaping work across Asia and the Pacific, but its effects are neither uni-form nor predetermined. AI can augment workers, transform jobs, create new tasks and employment, or substitute for existing tasks, often within the same occupation. The central policy question is there-fore not simply how many jobs AI may create or destroy, but which workers can benefit from techno-logical change, which face displacement or exclusion, and what conditions enable successful labour-market transitions. Existing evidence also cautions against treating AI capability, adoption, worker use, and occupational exposure as equivalent: these are analytically distinct concepts, and potential expo-sure does not establish that AI has been adopted, used by workers, or caused employment loss (Ghosh et al., 2025; Greenan et al., 2025).
These distinctions are particularly important in Asia and the Pacific because technological change is occurring across highly diverse labour markets. In 2023, around 66% of employment in the region was informal, rising to 87.2% in South Asia and 69.8% in Southeast Asia. Social-protection coverage and occupational structures also vary sharply across subregions. Young people face additional challenges: 20.4% of people aged 15–24 were not in employment, education or training in 2023. At the same time, AI-related demand is expanding in parts of the formal economy. In Southeast Asia, nearly 80 million workers, or 22.9% of employment, are in occupations with more than minimal potential exposure to generative AI, but only 3.3% are in the highest-exposure category. Importantly, employment in exposed occupations has continued to grow, reinforcing that AI exposure should not be interpreted as job loss (ILO, 2024a, 2024b, 2026).
The report therefore treats vulnerability as relational rather than inherent to particular demographic groups. Youth, women, persons with disabilities, migrants and minorities, older workers, rural workers, informal and platform workers, lower-skilled workers, and workers in small and medium-sized enter-prises (SMEs) may face different combinations of risks. Vulnerability emerges through four interacting dimensions: exposure to AI, capacity to adapt, access to training and employment opportunities, and access to social protection. A highly exposed worker may be relatively resilient if they have transferable skills, access to relevant training and opportunities, and adequate protection during transition, while a less-exposed worker may remain vulnerable because of weak infrastructure, limited qualifications, or inadequate institutional support.
For policymakers, six priorities emerge from the evidence:
- Strengthen skills and support school-to-work and labour-market transitions. Combine AI literacy with cognitive, technical, and social skills; strengthen technical and vocational ed-ucation and training, apprenticeships, internships, and career guidance; and support upskilling, reskilling, employment services, and job matching throughout working life (Ra et al., 2021; Loong et al., 2021).
- Ensure equitable access to AI-enabled opportunities and support SME adoption. Ex-pand reliable connectivity, electricity, affordable tools, local-language systems, and accessi-ble technologies. Support SMEs through finance, technical assistance, shared infrastruc-ture, workforce development, and organisational support, rather than technology acquisi-tion alone.
- Strengthen social protection and support informal and non-standard workers. Extend income protection, health coverage, pensions, and portable benefits that can follow work-ers across jobs, employers, and platforms, while linking protection with retraining, employ-ment services, and re-employment assistance (Chigbu, 2026; Loong et al., 2021).
- Govern algorithmic recruitment and management and prevent discrimination. Re-quire transparency, privacy and data protection, assessment of discriminatory effects, meaningful human oversight, and mechanisms for workers and applicants to contest con-sequential automated decisions, particularly in recruitment, performance evaluation, scheduling, and dismissal.
- Promote human-centred job design, job quality, and worker participation. Encourage firms to combine AI adoption with task redesign, worker training, internal mobility, and so-cial dialogue. Monitor wages, working time, autonomy, surveillance, well-being, and career progression alongside productivity and employment.
- Target structural barriers to inclusion. Avoid treating vulnerable workers as homogene-ous groups. Policies should address the mechanisms generating disadvantage, including care responsibilities, accessibility, connectivity, discrimination, weak bargaining power, and limited access to training, employment opportunities, and protection.
The evidence base remains considerably stronger on potential exposure, skills, and preparedness than on actual employment transitions. Future research should distinguish AI capability, adoption, use, and exposure more clearly; follow workers longitudinally through job entry, displacement, re-employment, and occupational mobility; examine intersecting forms of vulnerability; broaden outcomes beyond em-ployment and productivity to include wages and job quality; and evaluate whether training, social pro-tection, and AI-governance interventions actually improve worker outcomes. Stronger linked worker, firm, education, and institutional data are also needed across the region. The overarching lesson is that AI exposure is not destiny: labour-market outcomes depend on how technology is deployed, whether workers can adapt and move into new opportunities, and whether institutions provide effective pro-tection and support during transition.
Read the full report here.
