Can AI Help Islands Keep Their People?

Abstract

Can artificial intelligence help islands retain residents, attract returners and create livelihoods that do not require departure? The question is urgent but easily overstated. Many islands face ageing populations, selective youth out-migration, narrow labour markets, recruitment difficulties and high costs. AI may help local firms reach wider markets, augment scarce workers, support remote employment and convert island knowledge into new products. It may also automate entry-level work, concentrate value in distant platforms, increase pressure on housing and allow employers or governments to remove locally present jobs.

This working paper examines the relationship between AI, work and island demography without treating technological adoption as demographic success. It combines research on island youth migration and return with evidence on occupational exposure to generative AI and cases of island economic agency. Anguilla has captured public revenue from global demand for its .ai country-code domain. Icelandic start-up GreenFish is developing forecasting services from fisheries knowledge and high-performance computing. Te Hiku Media in Aotearoa New Zealand demonstrates how Indigenous data governance can underpin locally controlled language technology. Mauritius is using public programmes to build an AI start-up pipeline. Madeira’s digital-nomad initiative, while not itself an AI programme, supplies a cautionary prehistory of attempts to turn location-independent work into island development.

No verified case presently establishes that AI has caused sustained island repopulation. The paper therefore distinguishes four demographic pathways—retention, return, attraction and circulation—and four economic relationships—augmentation, diversification, capture and displacement. It proposes an Island AI Work and Demography Test. The central claim is that the appropriate unit of success is not an AI product, automated task or visiting remote worker, but a durable island livelihood embedded in local institutions and compatible with housing, services, culture and community authority.

Keywords: artificial intelligence; islands; demography; depopulation; work; youth migration; remote work; entrepreneurship; economic development

Evidence note

The references are limited to sources whose existence and bibliographical details were checked against journal, publisher, government, intergovernmental or institutional records. Complete texts or relevant pages were inspected wherever accessible. DOI, repository and stable institutional links are supplied. The paper distinguishes operational activity, pilots, public programmes and promotional claims. Figures reported by governments, programme operators or businesses are attributed to those institutions and are not treated as independent impact evaluations. Digital remote-work initiatives are included only as evidence about the conditions surrounding mobile work; they are not represented as AI deployments.

1. The demographic promise of AI

Island depopulation is often narrated as a shortage of people. More accurately, it is a shortage of viable ways for particular people to remain, return or arrive. A young person may value island life but leave to study because the desired programme is unavailable. A graduate may wish to return but find no suitable work. A business may have orders but be unable to recruit a specialist. A family may be ready to settle but cannot find housing, childcare or dependable transport. Population change emerges from these interlocking decisions rather than from a single demographic lever.

Cooke and Petersen’s study of rural island youth shows why employment, education and location must be considered together. Their typology does not reduce mobility to either forced exile or personal preference; it shows how aspirations, available work, education and attachment to place combine in different ways (Cooke and Petersen, 2019). Alexander’s research on graduate return to rural island communities likewise finds varied experiences of return rather than a uniform homecoming trajectory (Alexander, 2023). These studies precede the current generative-AI wave, but they establish the problem that AI proposals must address.

AI enters this geography with a powerful promise. It can make expertise available at a distance, lower some costs of producing and translating information, help small firms analyse markets, and enable workers to perform tasks previously requiring a larger team. A person may be able to undertake knowledge work from an island while serving clients elsewhere. A local enterprise may transform fisheries, linguistic or environmental knowledge into a service with a global market. Public bodies may use productivity gains to sustain services under labour shortage.

The same technologies can produce the reverse effect. If an employer can automate customer support, bookkeeping, translation or elementary design, it may remove work from an already thin labour market. If a distant firm captures the value of island data, local activity may rise without local income. If remote workers arrive with metropolitan salaries, housing costs may increase while community participation remains temporary. If AI makes centralised management easier, local posts may disappear even where the service is still needed.

The question is therefore not whether AI creates or destroys jobs in the abstract. It is whether particular arrangements widen the range of viable island lives. That requires demographic, economic and institutional evidence—not a count of tools, start-ups or training places.

2. Four ways people remain connected to islands

Population policy often treats retention and attraction as interchangeable. They are not. Four pathways should be distinguished.

Retention means enabling existing residents to stay. It may involve better work, more flexible work, stronger services or the ability to combine several occupations. Retention is especially important because a resident already possesses relationships, local knowledge and a practical stake in the community.

Return concerns people who have left, often for education or experience, and later come back. Return depends not only on vacancies but on the perceived quality of careers, family circumstances, housing and whether skills acquired elsewhere can be used without professional isolation. Alexander (2023) shows that graduate return is heterogeneous; a digital job alone cannot explain it.

Attraction brings new residents. Remote work, entrepreneurship and recruitment can diversify a community, but attraction may be temporary or uneven. An initiative that succeeds in marketing an island may still fail to create affordable homes, long-term attachment or reciprocal relationships.

Circulation recognises that island lives are often distributed across places. Seasonal residents, commuters, students and diaspora members may contribute without making a permanent return. Digital and AI-enabled work can strengthen these relationships through mentoring, investment, collaboration and part-time presence. Treating circulation as failure would miss an important form of island agency.

These pathways should not be collapsed into a headline population number. Retaining a care worker, enabling a graduate to establish a business and attracting a short-term digital nomad create different effects. Policy needs to record who benefits, for how long, on which island and with what consequences for services and housing.

Repopulation initiatives also require legitimacy. Craigie and colleagues’ analysis of the proposed Scottish Islands Bond argues that community resilience and islanders’ understandings should shape population interventions; the proposal was withdrawn after consultation revealed limited support and doubts about its capacity to meet its aims (Craigie et al., 2025). The lesson applies to AI. A strategy designed to “bring people back” can fail if it defines the island as an empty setting for imported workers rather than a community entitled to decide what kind of development it wants.

3. What AI changes about work

Generative AI does not map neatly onto whole occupations. It alters tasks. The International Labour Organization’s refined global index combines occupational task data with expert assessment and finds that one in four workers globally is in an occupation with some exposure to generative AI. Clerical occupations remain the most exposed, while the study emphasises that transformation is generally more likely than complete replacement (Gmyrek et al., 2025). Exposure is not a forecast of unemployment. It identifies technical potential whose outcome depends on job design, adoption, bargaining power and institutions.

The OECD’s review similarly finds limited evidence of aggregate employment loss at the time of publication, alongside worker reports of both improved performance and concerns about work intensity, privacy and inequality (OECD, 2023). The IMF estimates higher exposure in advanced economies than in emerging and low-income economies, while warning that preparedness, infrastructure and skills influence whether countries can capture complementary gains (Cazzaniga et al., 2024). These global classifications are useful, but they do not contain an island variable.

That omission matters. A small labour market has little redundancy. The disappearance of three junior administrative posts may be negligible in a national model but remove a whole entry route on an island. Conversely, automating documentation may allow a scarce nurse, planner or engineer to spend more time on work that cannot otherwise be staffed. The same task-saving technology can either sustain a post or justify its removal.

Four economic relationships clarify the possibilities.

Augmentation occurs when AI increases the capability of an island worker or enterprise without removing meaningful human responsibility. It may reduce paperwork, improve forecasting or make specialised knowledge easier to access.

Diversification occurs when islanders create new products, services or markets rather than merely making an existing task cheaper. It is strongest when new activity connects to local capabilities and produces careers with progression.

Capture concerns who receives the value generated by AI-related assets, data and demand. Public revenue, community licensing and local ownership are possible forms of capture. A project can create high revenue with few direct jobs, so capture and employment must be measured separately.

Displacement or extraction occurs when a task, income stream, decision or dataset moves away from the island without an equivalent local gain. It includes conventional job loss, but also vendor dependency, uncompensated data use and the hollowing out of entry-level learning.

A single initiative may contain all four. The policy task is to change their balance.

4. Island agency in the AI economy

4.1 Anguilla: capturing value from a jurisdictional asset

Anguilla’s .ai country-code top-level domain is the clearest example of an island jurisdiction capturing value from the current AI boom without itself deploying an AI system. IMF reporting records that registrations rose from 144,000 in 2022 to 354,000 in 2023. Domain registrations generated EC87million—aboutUS32 million—in 2023, just over one fifth of government revenue (Moreau, 2024). The government expected the share to stabilise at around 15 per cent as the initial surge slowed.

The case is important for two reasons. First, it demonstrates jurisdictional agency. An assigned digital identifier became a public revenue source at a scale that can affect investment capacity. Second, it prevents simplistic equations between digital growth and technology employment. The revenue does not establish that Anguilla created a large AI workforce or reversed depopulation. It is a form of economic capture whose demographic effects depend on how income is governed and spent.

If windfall revenue improves housing, education, connectivity, health or business finance, it may strengthen several conditions of retention. If it becomes structurally necessary to recurrent spending, volatility creates risk. The IMF account notes the need for diversification and prudent fiscal management (Moreau, 2024). Anguilla therefore supplies both an opportunity and a test: can a small jurisdiction convert temporary global attention into durable local capability?

4.2 Iceland: turning sector knowledge into a forecasting service

GreenFish offers a different relationship. Enterprise Europe Network reports that the Icelandic start-up grew from a 2023 bachelor’s project and combines historical fisheries data, satellite imagery and supercomputing to produce ten-day fishing-location forecasts (Enterprise Europe Network, 2026). The company received a €350,000 technology-innovation grant and gained access to European high-performance computing. The source reports anticipated benefits involving fuel, revenue and bycatch, but these are programme and company claims rather than independently evaluated outcomes.

The verified significance lies elsewhere. A locally grounded problem in a foundational island sector has become the basis of a potential exportable service. The founders are not using islandness as scenery for a generic application; they are applying computing capacity to a practice in which Iceland possesses accumulated knowledge, institutions and data. That is diversification through augmentation.

The employment question remains open. A successful forecasting firm might create skilled work, strengthen a marine-technology cluster and make an established sector more competitive. It might also reduce certain tasks or concentrate value among owners and data holders. Evaluation should therefore follow wages, career routes, ownership and effects on fishers, not only prediction accuracy or venture growth.

4.3 Te Hiku Media: language technology under community authority

Te Hiku Media shows how economic capability, culture and governance can be designed together. An International Telecommunication Union case presentation describes a Māori-led team building speech recognition, speech synthesis and pronunciation tools for te reo Māori. Its Kaitiakitanga Licence establishes conditions intended to keep labelled Māori data under Māori guardianship and for Māori benefit (ITU, 2022).

This is not simply a cultural-preservation example. Building and governing language technology creates technical capability, employment, training and negotiating power. It enables a community institution to set terms for an asset that external technology companies might otherwise absorb. The relationship between language data and AI value becomes visible and governable.

The case does not establish a population effect, and the institutional source is not an independent labour-market evaluation. It nevertheless demonstrates a crucial principle: island and Indigenous communities can be producers, licensors and governors of AI rather than end-users. Economic agency includes the ability to refuse uses, require reciprocity and decide which benefits count.

4.4 Mauritius: building an entrepreneurial pipeline

Mauritius has chosen ecosystem-building. The government’s AI StartUps programme, launched in October 2025 through the National Productivity and Competitiveness Council, was described as a pipeline that moved participants through skills development, mentoring and presentation of projects (Government of Mauritius, 2026). The 2025–26 budget also announced an AI innovation start-up programme as part of an economic-transformation agenda (Economic Development Board Mauritius, 2025).

These are verified public programmes, not proof of successful firms or job retention. Their importance lies in the attempt to organise a pathway rather than hold an isolated hackathon. For a small state, training without finance, procurement, compute access or customers may only prepare people to leave. An ecosystem becomes demographically relevant when a participant can progress from learning to paid work, business formation and a durable local career.

Mauritius also illustrates scale and distribution questions. National initiatives may cluster in the principal urban and institutional centres. An island-centred assessment should report participation and benefits across Mauritius, Rodrigues, Agaléga and other communities where practicable. National success cannot be inferred from a capital-city cohort.

5. Remote work is a precursor, not proof

AI may make a wider range of knowledge work location-independent. Yet the social effects of remote work began before the current AI wave, and those experiences provide a necessary control against technological optimism.

Nordregio’s synthesis of research on remote work and multilocality finds opportunities for Nordic rural and peripheral regions but stresses that outcomes depend on broadband, housing, services, planning and local strategies (Rohrer et al., 2024). Remote work can help an existing resident keep a metropolitan job, allow a returner to carry a career home, or attract a worker who brings external income. These are different pathways with different requirements.

Madeira’s Digital Nomads initiative is a prominent island example. Sardinha et al. (2023) analyse registration data for 14,960 people who expressed interest through the programme between November 2020 and December 2022. The authors report a highly international, technology-oriented group and describe public investment in workspaces, connectivity and community management. Their dataset records registrants rather than a representative sample of all nomads, permanent settlement or causal economic effects. It therefore demonstrates attraction and programme reach, not repopulation.

The distinction is vital for future AI-worker initiatives. A visitor who stays for six weeks and spends locally contributes to tourism and circulation. A resident who can keep a job because AI-assisted work reduces professional isolation contributes to retention. A graduate who returns to build a locally owned firm contributes to return and diversification. Aggregating all three as “talent attracted” hides what happened.

Housing is the most immediate constraint. High external salaries can support local businesses but also increase rents and convert homes to short-term accommodation. Services and social relations matter too. A coworking space may create professional community without integrating workers into civic life. Recruitment campaigns should therefore be accompanied by housing monitoring, long-term residency measures, routes into local organisations and attention to who is displaced.

Remote work also changes risk. An island resident employed by a distant platform can earn beyond the local market but may lack collective representation and be vulnerable to decisions made elsewhere. AI may increase productivity while allowing the employer to monitor, intensify or outsource work. Location independence is not the same as economic security.

6. Thin labour markets magnify both gains and losses

In large labour markets, workers displaced from one firm may find adjacent employment without moving. Islands often have fewer employers, narrower occupational ladders and higher switching costs. A worker may need to change sector or leave the island. This makes task-level analysis more important, not less.

Entry-level work deserves particular attention. Routine drafting, classification, customer service and administration are among the tasks most exposed to generative AI (Gmyrek et al., 2025). These tasks can be tedious, but they also teach systems, clients and professional norms. If they disappear without redesigned apprenticeships, an island may retain senior professionals temporarily while losing the route by which their successors develop.

Augmentation can be especially valuable where vacancies remain persistently unfilled. A small enterprise may use translation, coding assistance or document analysis to serve a larger market. A public office may use AI to reduce repetitive work and make a post more sustainable. The test is whether productivity creates time, quality and learning or simply produces an expectation that fewer people will carry more work.

Distribution within households and communities also matters. The ILO finds that occupational exposure differs by gender because women are overrepresented in clerical work in many labour markets (Gmyrek et al., 2025). Island strategies should therefore avoid presenting average exposure as neutral. Disability, age, language, connectivity and employment status will affect who can use new tools and who bears disruption.

Skills policy must be linked to demand. Short courses can widen awareness, but repeated introductory training does not create a labour market. Island learners need access to applied projects, mentors, recognised credentials, paid placements, procurement opportunities and compute resources. Employers need help redesigning jobs with workers rather than installing tools around them. Where local demand is small, regional networks and diaspora partnerships can create a larger professional community without requiring permanent departure.

7. From productivity to demographic outcomes

AI initiatives commonly measure users, generated outputs, hours saved, model accuracy or start-ups admitted. None is a demographic outcome. A credible theory of change needs at least four links.

First, the technology must improve a productive activity. Second, the gain must be captured by a worker, enterprise, community or public institution on the island. Third, that gain must affect a condition relevant to living there: income, career quality, workload, services, housing or confidence in the future. Fourth, the change must be durable enough to influence staying, returning, arriving or circulating.

Each link can break. A model may improve productivity while the saving flows to an external vendor. A start-up may raise investment and then relocate. Public revenue may rise without improving the services that residents value. Remote work may be available but unsuitable for households lacking space, care or reliable connectivity. A programme may attract people to the capital island while smaller islands continue to lose population.

This explains why causal evidence is presently weak. The current generative-AI cycle is recent, demographic decisions unfold over years, and many island initiatives publish activity rather than outcomes. The absence of proof should not be converted into a claim that AI is irrelevant. It should determine the evaluation design.

At minimum, governments and funders should record employment location, residency duration, wages, ownership, contract quality, participant origin, return status and island of residence. They should report business survival and local procurement rather than launch counts. Housing costs, service use and displacement should be monitored alongside attraction. Qualitative evidence should ask whether people feel able to build a life, not merely whether they completed training.

8. An Island AI Work and Demography Test

The following test is intended for strategies, grants, procurement and evaluation. It treats technology and population as connected through institutions.

The test should be applied before funding and repeated after implementation. A proposal should state what would count as failure and under what conditions it would be changed or ended. This makes reversibility part of responsible innovation.

9. A policy programme for durable island livelihoods

Ten commitments follow from the analysis.

1. Define the demographic pathway. State whether a programme aims to retain residents, enable return, attract settlers or support circulation; do not claim all four without evidence.
2. Support locally grounded applications. Prioritise problems where island firms, workers and institutions possess knowledge that can become capability and bargaining power.
3. Attach training to work. Connect courses to paid placements, apprenticeships, procurement, finance and continuing professional networks.
4. Protect entry routes. Require employers adopting AI to show how junior workers will gain experience and progress.
5. Use public purchasing strategically. Break contracts into accessible lots, require knowledge transfer and assess local value rather than purchase price alone.
6. Build shared regional capacity. Pool specialist advice, evaluation, compute access and model testing where individual islands cannot sustain them alone.
7. Govern data and intellectual property. Establish consent, licensing, attribution and benefit-sharing arrangements, including collective and Indigenous authority.
8. Join work policy to housing and services. Do not market islands to mobile workers without monitoring availability, affordability and displacement.
9. Publish island-disaggregated outcomes. Report who obtained durable work, who left, who returned and where benefits accumulated.
10. Create worker and community power. Include workers, unions, small firms, youth, diaspora and community organisations in decisions and provide routes to challenge harmful systems.

These commitments recognise that not every island should seek to become an AI hub. Some will gain more by using AI selectively within agriculture, care, fisheries, tourism, public administration or cultural work. The objective is not sectoral branding. It is an economy capable of sustaining varied lives.

10. Conclusion

AI can help islands keep their people, but only indirectly and conditionally. It may allow an existing worker to remain, a graduate to return, an entrepreneur to reach a wider market or a diaspora member to participate from elsewhere. It may help a scarce team do more. It may also remove the junior post that once began a career, transfer island data into external systems, intensify work and amplify housing pressure generated by mobile professionals.

The cases demonstrate several forms of agency. Anguilla captures public income from a jurisdictional digital asset. GreenFish develops a service around Icelandic fisheries knowledge. Te Hiku Media builds language technology under Māori governance. Mauritius is constructing an entrepreneurial pipeline. Madeira shows that attracting digitally mobile workers is possible, while reminding policymakers that registration and arrival are not the same as lasting demographic renewal.

No case examined here proves that AI has reversed island depopulation. That is the paper’s most important evidential finding. The appropriate response is neither rejection nor hype, but better questions and longer evaluation. Population change must be traced through retention, return, attraction and circulation. Economic change must be examined through augmentation, diversification, capture and displacement.

The unit of success is not an AI start-up, an automated task or a remote worker photographed beside the sea. It is a durable island livelihood: work that provides income, learning, security and belonging; retains meaningful authority locally; survives infrastructure and vendor failure; and fits within housing, services, culture and ecological limits.

An island future is not secured merely because work can be done from anywhere. It is secured when people have convincing reasons, practical means and a recognised right to build a life in that particular place.

References

Alexander, R. (2023) ‘Who returns? Understanding experiences of graduate return to rural island communities’, Journal of Rural Studies, 103, 103112. https://doi.org/10.1016/j.jrurstud.2023.103112

Rohrer, L., Granath Hansson, A., Bogason, Á. and Stjernberg, M. (2024) Planning around Remote Work: Latest Research and Implications for Planners and Policymakers. Nordregio Report 2024:24. Stockholm: Nordregio. https://doi.org/10.6027/R2024:24.1403-2503

Cazzaniga, M., Jaumotte, F., Li, L., Melina, G., Panton, A.J., Pizzinelli, C., Rockall, E.J. and Tavares, M.M. (2024) Gen-AI: Artificial Intelligence and the Future of Work. IMF Staff Discussion Note SDN/2024/001. Washington, DC: International Monetary Fund. https://www.imf.org/en/Publications/Staff-Discussion-Notes/Issues/2024/01/14/Gen-AI-Artificial-Intelligence-and-the-Future-of-Work-542379

Cooke, G.B. and Petersen, B.K. (2019) ‘A typology of the employment-education-location challenges facing rural island youth’, Island Studies Journal, 14(1), pp. 101–124. https://doi.org/10.24043/isj.81

Craigie, M.W., Currie, M., Duffy, P., Philip, L. and Wilson, R. (2025) ‘Principles of repopulation initiatives for community resilience: understandings gleaned from the proposed Scottish Government’s Islands Bond’, Island Studies Journal, 20(2), pp. 199–224. https://doi.org/10.24043/001c.144047

Economic Development Board Mauritius (2025) ‘Budget 2025–2026: Artificial Intelligence’, 9 June. https://edbmauritius.org/newsroom/budget-2025-2026-artificial-intelligence

Enterprise Europe Network (2026) ‘Finding fish is no longer a guessing game with AI-driven location forecasting’, 16 July. https://een.ec.europa.eu/success-stories/finding-fish-no-longer-guessing-game-ai-driven-location-forecasting

Government of Mauritius (2026) ‘AI StartUps Initiative showcases Mauritius’ drive toward digital economy and innovation’. https://govmu.org/EN/newsgov/SitePages/AI-StartUps-Initiative-showcases-Mauritius%E2%80%99-drive-toward-digital-economy-and-innovation.aspx

Gmyrek, P., Berg, J., Kamiński, K., Konopczyński, F., Ładna, A., Nafradi, B. and Pignatti, C. (2025) Generative AI and Jobs: A Refined Global Index of Occupational Exposure. ILO Working Paper 140. Geneva: International Labour Organization. https://doi.org/10.54394/HETP0387

International Telecommunication Union (ITU) (2022) ‘Accelerating the revitalisation of te reo Māori with AI’, AI for Good, 26 July. https://aiforgood.itu.int/event/accelerating-the-revitalisation-of-te-reo-maori-with-ai/

Jennings, A. (2026a) AI-slands: Towards an Island-Centred Approach to Artificial Intelligence. AI-slands Working Paper No. 1. Institute for Northern Studies, University of the Highlands and Islands. https://ai.insularworlds.co.uk/working-paper-1.html

Jennings, A. (2026b) AI and Island Cultures: Preservation, Extraction and the Right to Refuse. AI-slands Working Paper No. 2. Institute for Northern Studies, University of the Highlands and Islands. https://ai.insularworlds.co.uk/working-paper-2.html

Jennings, A. (2026c) Small Islands, Big Decisions: AI Governance, Dependency and the Right to Shape the Digital Future. AI-slands Working Paper No. 3. Institute for Northern Studies, University of the Highlands and Islands. https://ai.insularworlds.co.uk/working-paper-3.html

Jennings, A. (2026d) Bridging the Water: AI, Distance and the Future of Island Services. AI-slands Working Paper No. 4. Institute for Northern Studies, University of the Highlands and Islands. https://ai.insularworlds.co.uk/working-paper-4.html

Moreau, F. (2024) ‘An AI-powered boost to Anguilla’s revenues’, IMF News, 15 May. https://www.imf.org/en/news/articles/2024/05/15/cf-an-ai-powered-boost-to-anguillas-revenues

OECD (2023) OECD Employment Outlook 2023: Artificial Intelligence and the Labour Market. Paris: OECD Publishing. https://doi.org/10.1787/08785bba-en

Sardinha, L., Vieira, M., Sousa de Sousa, Á. and Leite, E. (2023) ‘Digital nomads—the case of the Autonomous Region of Madeira’, Journal of Entrepreneurial Researchers, 1(2), pp. 55–69. https://doi.org/10.29073/jer.v1i2.14