The accelerating growth of artificial intelligence (AI) and the threat it poses to social and economic life continue to dominate international debate. Much of the attention focuses on large language models – high‑capacity systems trained on extensive datasets and supported by substantial compute, energy and technical expertise. While these frontier models are driving innovation across sectors, their diffusion remains uneven. High barriers to entry, from infrastructure to specialized talent, limit who can meaningfully participate in or benefit from these advances.
This asymmetry is deepening existing digital divides. Compute capacity is overwhelmingly concentrated in advanced economies; Africa, for instance, accounts for less than 1 per cent of global data‑centre infrastructure. Training a frontier‑scale model can require thousands of megawatt‑hours of electricity, far exceeding what many developing‑country grids can reliably or responsibly supply – raising difficult trade‑offs in systems where every additional unit of power diverted to compute is a unit not available for households, clinics, schools or small enterprises.
Small AI offers an alternative pathway: its lightweight, low‑cost task‑specific applications and ability to operate with minimal compute and limited connectivity reflect the realities of resource‑constrained environments.
Small AI for Learning, Earning and Health
Small AI models are inherently contextual. They are designed to function in environments where bandwidth, infrastructure and digital literacy may be limited. By focusing on discrete tasks such as solving a math problem, interpreting a crop image or diagnosing medical symptoms, they are less complex to use.
This positions Small AI as a practical enabler of the Sustainable Development Goals. By reducing information asymmetries, supporting skill acquisition and lowering the cost of innovation, they create new entry points for individuals and communities that have historically been excluded from digital transformation. Take, for example, these applications bringing Goals 3 (health and wellbeing), 4 (quality education) and 8 (decent work and inclusive growth) further within reach.
In Ghana, the “Rori” AI math tutor (delivered via WhatsApp and trained on 500 micro‑lessons) costs approximately $5 per student annually and yields learning gains comparable to an additional year of schooling. Across Latin America and the Caribbean, AI‑enabled tutoring platforms extend personalized instruction to remote and Indigenous communities. And in South Asia, startups such as Bangladesh’s Shikho also integrate AI into mobile applications that operate offline and in multiple languages, ensuring learning extends to low‑connectivity settings.
Agriculture – an important income-generator and livelihood across the Global South – is another domain where Small AI is demonstrating measurable value. Farmers across Africa are adopting localized tools such as those providing hyper-local weather forecasts via SMS that support decision‑making and enhance productivity while mitigating risk. In Kenya, evidence shows the Nuru app cost-effectively enables farmers to photograph a diseased leaf and receive an immediate diagnosis without continuous internet access. In Senegal, digital agriculture platforms combine farmer profiles with crop data to deliver tailored guidance on disease management and water use.
Small AI is also beginning to influence healthcare delivery, particularly in underserved regions and hard-to-reach communities across developing countries. Lightweight diagnostic tools that interpret rapid tests using a basic smartphone camera or simple SMS‑based tools that help patients describe symptoms help frontline workers to triage, make more informed decisions and reduce delays in care. AI‑enabled chat interfaces can guide patients through basic health questions or direct them toward appropriate services. Image‑based tools assist community health workers in identifying conditions such as skin infections or nutritional deficiencies.
These tools often operate alongside the existing human knowledge, protocols and delivery systems of educators, farmers and medical professionals, reinforcing expertise and fostering positive disruption in the face of climate and market volatility, global insecurity and uncertainty.
Optimizing Impact with Effective Policies, Supporting Practices and Robust Guidelines
The promise of Small AI ultimately rests in part on the strength of the governance systems, policies and investment strategies that shape its deployment and use – including as part of a 'digital demographic dividend' approach. Effective oversight and sustained investment determine whether these tools advance equity or inadvertently entrench existing divides.
Take the illustrations above. In the education sector, the task is to ensure that AI contributes to stronger learning environments rather than fragmenting them. That requires attention to how learning outcomes are assessed, how schools and teachers are resourced and how educators are supported as they incorporate new technologies into their practice. Protecting children’s (and all learners’) data and ensuring fairness in algorithmic decision‑making are necessary for maintaining public trust.
Agricultural applications raise a different set of considerations. Here, governance must create space for local, intergenerational and Indigenous knowledge to shape the design and use of AI systems. When communities are involved in defining priorities and providing feedback, the resulting tools are more likely to reinforce existing knowledge networks and strengthen farmers’ autonomy. Clear rules around data use are needed to ensure that the value generated from agricultural information flows back to the people producing it.
Healthcare introduces yet another dimension, where the stakes of accuracy and safety are particularly high. Small AI tools still need to be validated and adapted to the cultural and clinical contexts in which they operate, and where appropriate, their use should be embedded within established care pathways. Strong privacy protections and responsible data practices are indispensable for sustaining confidence in digital health interventions.
A Path Toward More Inclusive and Sustainable Progress
Small AI cannot replace the structural reforms required for long‑term development, but by identifying needs and understanding and mitigating risks, we can widen access to innovation and reduce disparities in meaningful ways. By opening new avenues for learning, supporting livelihoods, strengthening food systems and extending the reach of basic services, these tools can contribute to development pathways (and the financing of them) that are more inclusive and resilient.
In this sense, Small AI offers a practical route towards a future in which technological advances support shared wellbeing and progress and reinforce the conditions needed for peace and prosperity for all.
