(Photo by iStock/jeffbergen)
In 2007, Nairobi vendors discovered a newfound freedom: buying inventory with only a phone. No bank account. No credit card. Just M-PESA, a mobile payment that now processes over KES 140 billion annually (roughly $980 million US) or 3,000 transactions per second. Since then, M-PESA has become the lifeblood of East Africa’s economy. But here’s the uncomfortable part: As millions across Africa went mobile, development experts were writing checks for landlines. Development brought wires while Africa went wireless.
Today, global development institutions face unprecedented challenges—from funding cuts to institutional withdrawals—and artificial intelligence offers a unique opportunity to bridge these gaps.
I’ve been here before, having spent decades watching development institutions try and fail to catch up with tech. After missing mobile’s initial wave, development organizations raced to create flashy apps designed in New York boardrooms. Most flopped. Meanwhile, local entrepreneurs created solutions that worked. “Hello Tractor” became “Uber for tractors”: Born in Nigeria, it now connects 5,000 tractors with 1.2 million farmers through its network of over 2,000 booking agents. Similarly, Farmerline has digitized 2.2 million farmers across 50-plus countries, providing them with tools to monitor and manage over 3.2 million acres of farmland.
As the same story unfolds, this time, we have to rewrite the script: AI offers a path to do more with less, if we invest in the infrastructure that will make it work for everyone.
AI Is Already Transforming Lives
We don’t need to imagine what’s possible, though. It’s already saving lives across multiple sectors. Take the Area 25 health center in Lilongwe, Malawi: AI-powered fetal monitoring has slashed stillbirths and neonatal deaths in that clinic by 82 percent. Instead of relying on periodic checks, this AI system gives doctors real-time alerts on a baby’s vital signs, enabling quicker intervention in a country where birth asphyxia is a leading killer.
Implemented through partnerships with organizations like PeriGen, Texas Children’s Hospital, and local health authorities, this system demonstrates the potential of appropriate technology. But what makes this approach different is that it was developed with local clinicians and adapted to work with existing health-care infrastructure.
Similar AI-powered diagnostics are making waves elsewhere. In Rwanda, Zipline’s AI-guided drones deliver blood supplies to remote hospitals in minutes rather than hours, with over one million deliveries completed since 2016 (now one delivery every minute). Funded through a public-private partnership between the Rwandan government and Zipline (with initial support from organizations like the Gates Foundation), the system has since expanded to four continents, eight countries, and serves over 4,500 health care centers. Across these areas, Zipline has delivered over 20 million vaccines, reduced blood product wastage by 67 percent where it operates in Rwanda, and helped reduce maternal deaths in Northern Ghana by 56 percent.
In education, AI is helping bridge critical teacher shortages. M-Shule in Kenya uses AI to create personalized learning plans for students based on their performance data, delivering custom SMS-based lessons to thousands of students. The platform has since been adopted by over 20,000 households across Kenya, Uganda, and Tanzania, and reports show that students score 7 percent to 20 percent higher on classroom exams compared to peers not using the system.
In agriculture, the transformation is equally compelling. IBM’s Liquid Prep, an open-source project now hosted by the Linux Foundation, exemplifies how AI can empower smallholder farmers: this intelligent mobile app collects soil moisture data from sensors and combines it with weather information to guide farmers on when to water crops. Initially tested in Karnataka, India, the solution has since expanded through partnerships with Texas A&M AgriLife and SmartCone Technologies, and demonstrates how locally adapted AI solutions can help farmers in drought-prone regions optimize water usage and increase yields.
AI is also making a tangible difference in climate adaptation. In Bangladesh, flooding displaces millions annually, but the UNDP, in partnership with Google’s Flood Forecasting Initiative and local ministries, has implemented early warning systems that leverage AI to process flood data and predict floods up to three days in advance. Sending real-time alerts to smartphones (tailored to work even in remote areas with limited connectivity) helps communities get vital time to prepare.
AI Can’t Work Without a Signal
AI is already reshaping health care, education, and agriculture in the Global South, but the barrier to more widespread use isn’t application. It’s infrastructure.
AI promises to fill gaps where human resources are scarce and training takes years. Sub-Saharan Africa has just two physicians per 10,000 people, for example (compared to the EU’s 41). This critical gap won’t be bridged quickly enough through training. But AI has the potential to empower community health workers (CHWs) to diagnose patients at home, allowing them to reach more people with fewer resources.
Yet for every success story, there are cautionary tales of AI solutions failing due to poor connectivity. In Tanzania, AI-powered tuberculosis detection using computer-aided detection (CAD) software for chest X-rays has been integrated into TB screening, but low diagnostic coverage remains a major issue and prohibitive costs remain another barrier to access, limiting the reach of AI-driven diagnostics in rural areas. In Zambia, an initial attempt to implement the AI-based CAD4TB system via cloud processing failed due to insufficient internet bandwidth, leading implementers to shift to an offline deployment model instead. Even well-funded digital programs struggle to achieve their full potential when the fundamental connectivity infrastructure is missing.
AI Needs More Than Hype
Development institutions need to focus on building the right foundation. As budgets shrink and traditional aid channels falter, it’s all the more essential that development institutions concentrate dwindling resources on building systems that can endure political turbulence, especially as AI’s massive computing demands continue to outpace the limited bandwidth available across the Global South.
Multilateral development banks (MDBs) and development finance institutions (DFIs) must play a bigger role in investing in the infrastructure that makes AI possible. This means focusing on three key areas:
- Connectivity: We can’t build an AI-powered future on patchy networks. MDBs need to invest in affordable, high-speed internet across the Global South—via satellites or fiber-optic networks—to ensure AI tools work for everyone, not just those with the latest devices.
- Unglamorous basics: This isn’t about just rolling out apps. We need digital ID systems, payment networks, and regional data centers. India’s UPI system shows how the right foundation can fuel a digital revolution.
- Break the AI language barrier: Current AI models are trained on Western data and languages. DFIs need to fund AI research centers focused on local knowledge and African, South Asian, and Indigenous languages. This way, AI can address people’s unique needs and realities.
We missed the mobile revolution. Let’s get AI right.
If development finance institutions are to be a central part of AI’s larger transformation, they need to fund the rails, not chase the rockstars.
At the Area 25 health center, AI is already saving lives. But these aren’t just isolated success stories: they’re showing us that, as development aid shifts, we need to place our bets on the foundation that enables local innovation. If we get it right this time, the future of AI won’t just be written in Silicon Valley. It can be made in Lagos, Nairobi, and Mumbai. But while development institutions don’t need to write the future, they need to build the right rails that let local entrepreneurs lead the way.
Read more stories by Raj Kumar.