BLOG POST

A New Accelerator to Design and Study AI Applications in Global Development

It’s tough to ignore AI. There’s constant press about it, and everyone from the World Bank to the UN Secretary General has been putting forth positions on the topic. Beyond the high-level attention, a growing number of studies have highlighted AI's potential to impact key development outcomes. For example, randomized evaluations of AI tutors in Ghana and Nigeria  have shown impressive results, with the Nigerian study reporting learning gains equivalent to nearly two years of typical education within just six weeks. However, the effects of AI are not universally positive. Beyond safety concerns, research is uncovering disparities in its impacts. In a randomized evaluation from Kenya, for instance, already successful entrepreneurs that used an AI-powered business coach experienced a 15 percent increase in profits or revenue, while already struggling entrepreneurs that used it saw a 10 percent decline.

Given that AI’s impact on development is going to grow, it’s important to develop a better understanding of these risks and opportunities. That’s why CGD is excited to be a collaborator on an exciting new project, the AI for Global Development Accelerator, headed up by the Agency Fund, with OpenAI as another collaborator. The Accelerator’s mission is to fast-track the safe and effective development of AI applications that enhance human agency and drive meaningful development outcomes.

The Accelerator’s first cohort (which Han was on the selection committee for) includes seven nonprofits, most of which either have successful existing programs that have been tested by randomized controlled trials or are building on existing evidence bases about what works. With the integration of AI, these organizations aim to broaden the impact and reach of their programs. As part of the cohort, each nonprofit will receive at least $500,000 in funding, OpenAI credits, and access to engineers, behavioral scientists, and development experts—taking advantage of the Agency Fund’s expertise in these areas and OpenAI’s technical support.

The role that experts at the intersection of research and policy can play in this debate is crucial. There is a natural space for these experts—including those at CGD—that can support these nonprofits in designing and assessing their interventions, helping ensure that technology adoption is tied to improvements in key development outcomes. This is in addition to sharing learnings and fostering critical public dialogue about the role of AI in development interventions with policymakers, researchers, and funders. 

For example, here are some of the questions we’re hoping to generate insights on that can inform both participating organizations and the wider sector:

  • Can interventions that have already been demonstrated to be cost-effective prior to the introduction of AI be delivered even more inexpensively and at much larger scales with AI?

  • Even if costs fall from the introduction of AI, will effectiveness hold?

  • Are there common mechanisms that drive effectiveness across interventions that are particularly well suited for generative AI?

  • How do we design AI-enabled interventions that mitigate potential harms from heterogeneous effects—especially if these effects exacerbate inequality?

  • What are some of the funding, skills, and development expertise that NGOs need to deliver cost-effective, AI-powered applications?

  • What AI applications should the development sector be investing in?

  • What do policymakers and funders need to see in order to adopt AI-powered interventions at scale?

AI is here, whether the development world is ready for it or not. Whether AI-powered interventions prove cost-effective at improving development outcomes will depend on effective collaborations between the developers of AI technologies, governmental and non-governmental public institutions, the constituents they serve, and development experts. This year, we’re hoping to facilitate those conversations. Stay tuned for more.

If you want to learn more about the Accelerator itself or the selected non-profits, you can check out this wiki, or read on for a summary of the interventions:

  • Rocket Learning is an organization that directly serves daycare workers in government-run daycare centers, as well as parents of children between the ages of 3 and 6 who attend these centers. They will integrate AI to create a Personalized Development Coach (PDC) for parents and a Smart Teaching Assistant (STA) for daycare workers, delivering adaptive learning content via WhatsApp to enhance ECE for underserved children in India.

  • Jacaranda Health deploys affordable and scalable solutions through government hospitals to improve the quality of maternal care and newborn outcomes in Kenya. Through two-way SMS exchange, their AI-enabled digital health service called PROMPTs empowers women to seek care at the right time and place, giving them greater agency in the health system. They will be enhancing their PROMPTS platform with a Swahili-speaking voice Q&A functionality, leveraging OpenAI tools to bridge gaps in maternal care for low-literacy and sight-impaired mothers in Kenya, through accurate speech-to-text, fluent text-to-speech, and integrated AI responses.

  • Noora Health equips family caregivers with life-saving skills through its Care Companion Program (CCP), delivering in-facility training, culturally relevant educational content, and simplified health guidance.  They will be enhancing their Remote Engagement Service (RES) serving India with OpenAI’s tools to deliver personalized health support, automate query responses, and prioritize high-risk cases, transforming caregiver training into a dynamic, holistic health companion.

  • Reach Digital Health serves vulnerable populations, primarily focusing on women, children, youth, and marginalized communities across various health areas.  They will integrate AI into programs like MomConnect to enhance personalized healthcare delivery and user engagement for vulnerable populations in South Africa.

  • Digital Green collaborates with farmers, public and private organizations, leveraging its Farmer.Chat platform to deliver localized, actionable farming solutions. They will enhance Farmer.Chat with advanced AI capabilities, including improved automatic speech recognition (ASR) for Hausa, offline content access, and smarter conversational interfaces, leveraging partnerships to provide accurate, inclusive, and scalable agricultural advisory services in Nigeria.

  • Precision Development serves smallholder farmers in low- and middle-income countries by providing timely, accurate, and actionable agricultural information to improve their livelihoods. They will integrate generative AI to deliver highly personalized, real-time agricultural advisories, leveraging multilingual text-to-speech, hyper-local data processing, and scalable automation to enhance farmer productivity and resilience in India.

  • Youth Impact is an evidence-based, youth-led organization that identifies, adapts and scales up health and education programs designed by youth for the youth worldwide. They will integrate generative AI into their ConnectEd program to scale voice-to-voice AI tutoring and automated scheduling, enabling personalized, cost-effective learning for 700,000 children in Karnataka, India by 2025.

Disclaimer

CGD blog posts reflect the views of the authors, drawing on prior research and experience in their areas of expertise. CGD is a nonpartisan, independent organization and does not take institutional positions.


Image credit for social media/web: Simone D. McCourtie / World Bank