AI Impact Initiative

2 days ago
11 min read

Updated: 9 hours ago

AI Insights and Opportunities from Community Leaders Across the Global South

Background

Artificial intelligence presents one of the greatest opportunities in human history to accelerate progress on education, healthcare, agriculture, and economic empowerment. Yet its benefits are not shared equally. AI development remains concentrated in wealthy nations, built on datasets reflecting high-income environments and deployed primarily in markets where purchasing power already exists. The communities where the return on AI investment, measured in human outcomes, would be greatest remain the least represented in the spaces where AI is being designed, funded, and deployed. Without deliberate intervention, this risks deepening the very inequalities that social entrepreneurs across the Global South work every day to overcome.

Aiforgood Asia and NELIS Global established the AI Impact Initiative to address this gap directly. The partnership combines Aiforgood Asia's applied research expertise and hands-on experience deploying practical AI projects in real-world community settings with NELIS Global's established network of community leaders and social entrepreneurs through the One Million Leaders Fellowship, because identifying real solutions requires both technical rigour and genuine trust on the ground. Together, the initiative is built on a single principle: the people closest to the challenges are best positioned to identify the solutions.

This study asked a direct question: where and how can AI create the greatest positive impact for communities across the Global South, as defined by the people living and working within those communities? A structured survey was designed and distributed through the One Million Leaders Fellowship network, capturing AI awareness, current challenges, barriers to adoption, and what support would make deployment practical and sustainable. The result is 611 responses from social entrepreneurs, educators, health workers, and community leaders across 56 countries, making this one of the most comprehensive studies of its kind. These findings form the empirical foundation of the AI Impact Wheel, a framework that will guide future investment, training, and partnership decisions. The need is documented. The communities have spoken. What follows is the evidence.

The Participants

The 611 respondents to this survey represent a cross-section of social entrepreneurship across the Global South that is, in itself, significant. They are directors and founders, educators and doctors, engineers and youth leaders, researchers and community organisers. Many are operating in conditions of real constraint, working in post-conflict areas, remote regions, and communities with limited infrastructure. What they share is a commitment to practical, ground-level impact.

AI engagement across this group is already underway. 44.4% of respondents report already using AI in some form, with a further 29.6% actively considering adoption. Familiarity is strongest with generative AI, which scored an average of 3.68 out of 5, followed by computer vision at 2.92 and natural language processing at 2.74. This is not a group encountering AI for the first time. It is a group that understands the technology's potential and is ready to move, given the right support.

Geographic Reach

Responses came from 56 countries across South Asia, Sub-Saharan Africa, Southeast Asia, the Middle East, and beyond. Pakistan contributed the largest share at 26.0% of responses, followed by Nigeria at 18.0% and Nepal at 10.5%. Afghanistan, Bangladesh, Kenya, Yemen, India, South Sudan, and Ghana each contributed between 1.5% and 6.7% of responses. This is not a survey of one region or one type of community. It reflects the breadth of what social entrepreneurship in the Global South actually looks like.

Figure 1. Survey Responses by Country

World map highlighting countries representing response numbers in a legend on the right.

This map shows the geographic spread of the survey dataset across 56 countries, illustrating the breadth of participation across the Global South. Larger bubbles indicate higher response counts, highlighting stronger concentrations in countries such as Pakistan, Nigeria, and Nepal while also showing broad representation across Africa and Asia. Detailed results can be accessed in the interactive map in the link below.

Interactive Map of Survey Responses

Sector Representation

Education accounts for 41.7% of respondents, reflecting the central role that learning plays in community development across these regions. Environment follows at 15.7%, Health at 14.1%, Agriculture at 9.7%, and Governance at 4.7%. A further 11.1% of respondents work across sectors that do not fit a single category, including microfinance, engineering, and community services. Together, these sectors represent the frontline of social impact work and align directly with the areas where AI has the greatest demonstrated potential.

Figure 2. Respondents by Sector

Bar chart showing education as largest sector at 41.7%. Other sectors include environment, health, agriculture. 611 respondents, 56 countries.

This chart shows the sector distribution of survey respondents, highlighting the fields in which participants are primarily working. Education represents the largest share of the sample, followed by environment, health, and agriculture, providing important context for how the findings in this report should be interpreted.

The Community Challenge Landscape

The survey asked respondents to identify the most pressing daily challenges facing their communities. The answers establish why this research matters.

Access to quality education was cited by 68.6% of respondents, making it the most widespread challenge across all regions and sectors. Unemployment followed at 65.0%, pointing to where a significant share of community effort and attention is concentrated. Climate-related issues were identified by 52.2%, affordable healthcare by 49.9%, and reliable agriculture and food security by 43.2%. Financial services and inclusion (41.6%) and public safety (39.6%) complete the picture.

These challenges do not sit in isolation. Poor educational outcomes limit employment prospects. Food insecurity is worsened by climate stress. Limited healthcare access puts additional pressure on households already stretched by poverty. The communities represented in this survey are contending with several of these at once, with limited resources and infrastructure.

The qualitative responses add texture to the statistics. Respondents described building learning programmes for children without school access, developing agricultural tools for smallholder farmers, training women in remote regions for green economy work, supporting communities adapting to climate change, and running health awareness campaigns where preventable diseases remain common. These are active, practical efforts being made right now, often with very little support.

The relevance of AI to these challenges is direct. Education, healthcare, agriculture, climate resilience, and financial inclusion are areas where practical AI applications already exist and are deployable at scale. The gap is not technological. It is one of access, investment, and localisation. This research documents precisely what is needed and where. That is what makes it a credible basis for action.

Figure 3. Most Pressing Community Challenges

Bar chart showing priorities of respondents. "Access to Quality Education" leads at nearly 70%, followed by unemployment and climate-related issues

This chart ranks the daily challenges most frequently identified by respondents. Access to quality education and unemployment emerge as the most widespread concerns, followed by climate-related issues, healthcare, and food security, giving a clear picture of where community needs are most acute.

Where AI Can Create the Greatest Impact

A central goal of this research was to understand where community leaders believe AI could create the greatest positive impact. Respondents ranked five sectors in order of priority.

Healthcare and Education emerged as the joint top priorities, effectively tied. Healthcare was ranked first by 29.8% of respondents with an average rank of 2.56, and Education by 37.6% with an average rank of 2.57. It is worth noting that 41.7% of respondents work in the education sector, which may influence these rankings. The data should be read with that context in mind, and further research across a more evenly distributed sample would strengthen these findings.

In healthcare, respondents identified opportunities ranging from diagnostics and community health worker support to health data analysis and patient outreach. The consistent view across responses was not that AI solves every health challenge, but that the right model trained on the right local dataset could meaningfully extend the reach of already overstretched health systems. Not every problem is an AI problem. The value is in matching the right tool to the right need, and that judgment can only be made by asking the communities themselves.

Agriculture ranked third, with 27.8% of respondents placing it first. Respondents described practical applications including crop monitoring, pest diagnosis, weather prediction, and market access. Community development followed at 24.5%.

Conservation ranked last of the five areas, with 20.6% of respondents placing it first and an average rank of 2.70. While environmental concerns featured strongly in the community challenges section, respondents did not prioritise conservation as the primary domain for AI deployment, likely reflecting the more immediate weight of education, health, and food security in their daily work.

A consistent theme across the qualitative responses is that respondents do not view AI as a replacement for human work. They see it as a tool that can strengthen what they are already doing. This is a notable finding. The debate around AI displacing jobs tends to centre on white-collar and knowledge work. For the community leaders in this survey, that framing does not reflect their reality. Their concern is not replacement. It is access. They are looking for practical tools that extend the reach of work already underway in education, healthcare, agriculture, and environmental resilience, and that is precisely the framing that should guide how investment and deployment decisions are made.

The breadth and specificity of this dataset position it as a credible evidence base for international bodies. The documented needs and priorities of 611 community leaders across 56 countries make a direct case for targeted investment from development partners, including the UN, EU, and bilateral donors working on digital equity and sustainable development.

Figure 4. Perceived Priority Sectors for AI-Enabled Community Impact

Bar chart ranking perceived priority sectors for AI-enabled community impact. Education (37.6%) is perceived as highest priority, followed by Healthcare (29.8%), Agriculture (27.8%), Community Development (24.5%), and Conservation (20.6%).

This chart shows how respondents ranked the sectors where they believe AI could create the greatest positive impact. Education appears as the leading priority, followed by healthcare and agriculture, reflecting the areas where respondents see the strongest potential for practical, community-level deployment.

Barriers to AI Adoption

Although interest in AI is strong across the respondent community, several clear barriers are preventing wider adoption.

Cost was ranked as the primary barrier by 42.4% of respondents. Many indicated that the financial requirements associated with AI tools, infrastructure, and implementation remain beyond the reach of their organisations. Social enterprises often operate with limited budgets, making investment in new technologies difficult without external support.

Limited technical skills ranked second, cited as the primary barrier by 32.6% of respondents. Many expressed interest in using AI but do not currently have the knowledge required to build, deploy, or manage AI systems. This reinforces the demand for accessible, practical training documented elsewhere in this report.

Technology access and infrastructure followed closely at 31.4%, with internet connectivity cited by 29.3% as their primary constraint. Reliable connectivity, suitable hardware, and access to computing resources are not consistently available across many regions, which makes even simple AI applications difficult to implement. Advances in edge computing and lightweight AI models are beginning to reduce some of these dependencies, and this represents a promising area for follow-up research and practical deployment.

Respondents also raised concerns around trust, ethics, and responsible use. Data privacy (27.3%), regulatory uncertainty (22.1%), and community trust (20.3%) were each cited as primary barriers by a meaningful share of respondents, reflecting a community thinking carefully about how AI is adopted, not just whether it is adopted.

The picture that emerges is not one of disinterest or resistance. The communities represented in this study know why AI matters, they know what problems need solving, and they know where the work needs to happen. What they need is the how. The barriers to AI adoption are structural, not motivational, and what is missing is enablement: the training, tools, and support to move from awareness to action.

Figure 5. Primary Barriers to AI Adoption

Bar chart showing top barriers to technology adoption. "Cost" leads at 42.4%, followed by "Limited Technical Skills" at 32.6%, and Technology Access or Infrastructure at 31.4%.

This chart presents the issues most frequently ranked as the biggest barriers to AI adoption. Cost is the most significant obstacle, followed by limited technical skills and infrastructure constraints, reinforcing that the main barriers are structural rather than motivational.

Capacity Building Needs

Respondents were asked what forms of support would most help them adopt AI in their work. The responses show clear priorities.

Training was the most frequently requested form of support, identified by 86.4% of respondents. Interest in AI is genuine and widespread, but many community leaders do not yet have access to practical learning pathways that translate technology into real-world applications. This is the single most actionable gap the initiative can address, and one where cloud providers such as AWS, Google, and Microsoft are already positioned to contribute through credits, training programmes, and developer support.

Funding followed at 64.0%. For organisations operating with limited budgets, even modest investment in tools or pilot projects remains out of reach without external backing. Technical expertise was close behind at 63.2%, reflecting the need for experienced practitioners who can guide implementation and help translate ideas into working systems.

Peer collaboration (53.0%) and access to AI tools and platforms (51.4%) were each selected by more than half of respondents, pointing to both the social and practical dimensions of adoption. Roughly a third of respondents also identified AI models that work in their community's own local languages (39.9%), computing resources (37.3%), and open datasets (35.8%) as needs, underlining that the infrastructure gap is real and specific.

What these findings make clear is that isolated pilots go nowhere without the broader ecosystem to support them. Respondents are not asking for a single tool or a one-off workshop. They are asking for a coordinated strategy: training, funding, technical guidance, local language capability, and the connections to put knowledge into sustained practice. That is the value proposition of the AI Impact Initiative, and it is precisely what distinguishes a programme built on this evidence from one built on assumptions.

This matters enormously for the communities in this study. The rapid advancement of small language models and edge AI is fundamentally changing what is possible without large-scale cloud infrastructure. Models that previously required significant computing resources can now run locally, on modest hardware, without continuous internet connectivity. The total cost of deploying practical AI tools in low-resource environments is falling, and the pace of that change is outrunning most policy and funding cycles.

The infrastructure gap that today prevents adoption will look very different in the near future. The work being done now by Aiforgood Asia and NELIS Global to build awareness, skills, and community readiness is not preparation for a distant future. It is preparation for a transition that is already underway. The communities that are ready when these tools become widely accessible will be the ones that benefit. This research exists to make sure they are.

Figure: Primary Barriers to AI Adoption: Percentage of respondents ranking each issue as the biggest barrier to AI adoption, with cost, limited technical skills, and infrastructure emerging as the most significant constraints.

Figure 6. Support Needed to Enable AI Adoption

Bar chart showing percentages of support needs. "Training or Workshops" leads at 86.4%, followed by "Funding or Grants" at 64%.

This chart highlights the types of support respondents identified as most necessary for effective AI adoption. Training stands out as the strongest need, followed by funding and technical expertise, showing that community demand is not only for technology itself but for the broader ecosystem required to use it well.

Conclusion

This study brings together the perspectives of community leaders, educators, health practitioners, farmers, and social entrepreneurs across 56 countries. Their responses provide a grounded picture of where artificial intelligence stands in relation to the work already underway across the Global South.

The findings reveal both readiness and constraint. Interest in AI is strong across education, health, agriculture, climate, and conservation. Respondents described AI not as a replacement for human effort, but as a tool to extend the reach of work already in progress. What holds most of them back is not ambition but access: to training, technical expertise, and the financial resources needed to move from interest to implementation.

Through the partnership between Aiforgood Asia and NELIS Global, the AI Impact Initiative is designed to close that gap. Working directly with community leaders through the One Million Leaders Fellowship, the initiative deepens understanding of where AI can genuinely serve and builds the capacity of those best placed to lead its adoption. This survey is both a reflection of where things stand and a foundation for what comes next: practical pilots, expanded access to training, and the partnerships needed to support responsible AI deployment in the communities where the return on investment, measured in human outcomes, would be greatest. The leaders represented in this study are already moving. For sponsors and partners, the opportunity is not to create momentum but to join it at a moment when the direction of that momentum can still be shaped.

Disclaimer on Use of AI and Research Support Tools

This report reflects the research design, survey methodology, analytical approach, interpretations, and conclusions developed and overseen by Aiforgood Asia as part of the AI Impact Initiative, conducted in partnership with NELIS Global.

The survey instrument, research framework, and analytical direction were developed by Aiforgood Asia in collaboration with the leadership network of the One Million Leaders Fellowship. All findings are based on responses from 611 participants across 56 countries, gathered through the fellowship network and reviewed, cleaned, and analysed under the supervision of the research team.

AI writing tools were used as assistive aids during the preparation of this report. ChatGPT (model 5.3) and Claude Sonnet (4.6) supported the structuring of draft text, clarity of language, and editorial refinement, based on analysis and direction provided by the research team. These tools were not used to design the research methodology, generate survey questions, independently analyse the dataset, or determine the findings of the study.

All interpretations, findings, and conclusions were reviewed and validated by Aiforgood Asia prior to publication. The organisation takes full responsibility for the accuracy of the information presented and the integrity of the analysis contained within this report.

Jesse Arlen Smith

President & Co-Founder

Aiforgood Asia

David Bergendahl

Co-Founder