4 practical applications of AI in global development

We often hear about the potential for artificial intelligence, or AI, to transform global health and development. But how are large language models, which respond to prompts using considerable amounts of data and deep learning algorithms, already impacting the real world? As organizations optimize AI solutions for global health and development challenges, there are more and more examples that demonstrate what the AI hype is all about. For Devex Pro week, experts from organizations including the Gates Foundation, UN Women, and OpenAI — the AI research company that developed the popular chatbot ChatGPT — gave live demos of practical applications of AI in global health and international development. Here are four ways AI is being used. Improving data access and decision making While ChatGPT does not require technical skills, beyond providing a clear prompt, users with technical skills are customizing GPTs to meet their needs. “We don't necessarily build solutions in the global development space,” said Rebecca Distler, strategic health partnerships lead at OpenAI. “Rather, partners build using our tools and leveraging our models.” She demonstrated how IDInsight, a global advisory, data analytics, and research organization focused on global development, used OpenAI tools to develop Ask-a-Metric. IDInsight wanted to make it easier to access the data needed to make decisions, by meeting them where they are: on WhatsApp. Distler presented the example of a publicly available database of COVID-19 cases in the Tamil Nadu state in India. Instead of leaving the data on dashboards, best accessed on a desktop, IDInsight created a chatbot that allows users to interact with the data in real time, on their phones, with the model replying in the language of their choice. Streamlining proposal writing AI also makes it possible to write compelling and compliant requests for proposals, or RFPs, quickly and efficiently, said Mustafa Hasnain, CEO of Grant Assistant, adding that his own personal challenges navigating this process helped lead him down this path. “It's a really interesting challenge for somebody with dyslexia to be involved in the proposal writing and international development space,” he said, “especially when words kind of move around for you.” He began considering how an AI tool like Grant Assistant could help him, and local organizations worldwide, which have fewer resources to compete for RFPs than the groups that tend to win these contracts. Hasnain showcased a 207-page request for proposals from the U.S. Agency for International Development, demonstrating how the Grant Assistant tool can assist teams in deciding whether to apply, identifying their value proposition, and crafting their proposal. Improving efficiency in the workplace Emad Karim, advocacy, campaigns, and innovation coordinator at UN Women, also developed a customized AI tool in response to challenges he faced in the workplace. “A lot of us are struggling with juggling different hats,” he said of his work within the United Nations. “Regardless of our specialization, we all have to do different segments of work related to communication, knowledge management, managing staff, admin work, and the higher you get at the managerial level, you also get to do a lot of admin and repetitive tasks.” The U.N. has its own language, context, and sensitivities that were not captured by public AI tools, Karim explained. So he developed “UN Style” GPT, which helps create materials that adhere to U.N. guidelines and aids job seekers applying for roles at the U.N. Karim demonstrated how the tool can serve as a personal assistant, improving work output without replacing the human element. It can even design a social media campaign that reflects a particular U.N. agency’s tone and style. Providing health care advice mDoc, a Nigerian health tech startup, leverages AI to provide virtual healthcare guidance to people living with chronic diseases. For his demonstration, Zameer Brey, deputy director for technology diffusion at the Gates Foundation, played a video of Nneka Mobisson, co-founder and CEO of mDoc. She explained how mDoc created a chatbot that would answer inquiries from women with hypertension using an existing dataset of questions and answers from their human coaches and community members. “We recognize that really kind of building and investing in the agency and self-efficacy of people could really transform that behavioral change that they needed to really live healthier, happier, and more productive lives,” she said. By using data from hundreds of thousands of interactions to fine-tune ChatGPT, mDoc was able to build a custom chatbot to deliver real-time, accurate, safe advice to end users, Brey explained. He shared early results showing that the GPT outperformed the human health coach. The challenge mDoc faces is scaling from 100,000 members to millions of users across Nigeria. But large language models allow them to reach many more people than they could through human interactions alone, Brey explained. “If we think about these kinds of solutions in direct-to-consumer applications, we are excited because they provide a high potential of providing end users with accurate, timely information in a way that's easy to relate to and reaches them where they are,” he said. Watch the recording to see these demos, as well as a discussion on questions around the cost and accessibility of AI tools, ethical considerations, and other considerations for organizations looking to fully leverage the potential of AI in their work.