How the World Food Programme is using AI

In a sector where every dollar must have an impact, financing to experiment with artificial intelligence technology in the humanitarian and development space can be hard to come by. “If we invest, some of these investments will work out, some of them won’t. That's definitely a tough reality in the sector we’re working in where nobody wants to waste money,” says Bernhard Kowatsch, head of the innovation accelerator at the World Food Programme. For him, part of the answer is to co-design solutions with local communities and program officers as well as technology experts, to ensure they have the best chance of achieving impact. “This is where human-centered design [and] lean, start-up thinking comes in. We need to test and we need to develop with people so that we actually don't waste money,” he said. Then “once you prove [a solution] is working, all of a sudden, you are able to get traditional funders, like some of the larger government donors, to actually fund those types of innovative approaches.” Speaking to Devex, Kowatsch explained the role he believes AI can play in aid — and how one project has saved the organization $50 million. This conversation has been edited for length and clarity. What do you think the role of AI is within the global development sector? AI has the potential to transform [many] aspects of humanitarian and development work as we know it, in terms of how we improve operations. For WFP, it’s [about] identifying where hungry people are, how hungry they are, and what type of support they need. It’s essentially answering questions … [for] vulnerability analyses: How can we be quicker and more accurate in providing people with what they need and what they want? This is also an aspect of supply chain optimization and data optimization, [creating] accountability systems for the individuals that we’re serving. … In a development setting, I think what’s quite interesting is that all of a sudden there are opportunities to democratize services and change types of programs. [For example], we have programs where, based on AI, you now can provide insurance solutions to smallholder farmers in developing countries who were uninsured before. Do you think the development and humanitarian sectors are keeping up with the implementation of AI? Generally speaking, it’s a constant challenge. We've been lucky … we had initial funders – the German government and also the German Foreign Office in Bavaria. Now, we’re supported by the German government, the United States government, Luxembourg, Austria, the Bill & Melinda Gates Foundation, Google, and Salesforce; a lot of different organizations, both private sector entities and individual funders. But it’s a constant challenge and you need to find the right people with the right understanding and risk appetite in the sector. Our new executive director, Cindy McCain, came in and said she has three priorities … [including] innovation and partnerships. … I think the tone at the top from the CEO and leadership is very important. Is WFP already implementing AI solutions? There's one example, which is called Project Optimus. This is optimizing supply chain routes and warehouses where food is purchased, including the food baskets that people receive, using big data analytics. Optimus has been used in 20 countries already, including in emergency [contexts] like in Afghanistan, Ethiopia, and Ukraine. Using big data, the tool has led to $50 million of savings already. Another example is … a tool that's called Hunger Map. It's actually public [and] what you can see is the near real-time food security situation, even on a district level, where we are actually overlaying field services. A third one is Meza, an AI tool that digitizes paper records from nutrition programs. … In a lot of cases in remote areas, we might still receive paper records from the health clinic. Meza is a tool that’s already active in Kenya, where we’re using it to digitize the data and, based on the data, optimize the programs. Can these interventions be rolled out anywhere or does there have to be a certain enabling environment? If you’re developing something where everybody needs to have constant internet access and a smartphone, it may work in some countries or in some urban areas. … If connectivity is available as a basic technology and people can afford data, then that’s of course great because it opens a lot of opportunities. At the same time, what we are seeing in reality in developing countries is that oftentimes there might be just one entrepreneur in the village, operating the smartphone. … So you need to think about offline capabilities. … This is where I think actually building a solution that’s relevant so that it’s solving a real problem, fit for purpose, but also affordable to people [is key]. What advice would you give to smaller organizations that perhaps don’t have the resources of WFP about how to start dabbling in the AI space? What’s the organization’s strategy and how can innovation technology best support this? That’s the foundational thought. … [Then you need] to encourage all staff to come up with ideas or submit innovation, which is actually a lot more about the organizational culture. We started in 2014 to run innovation challenges. … Our executive director sends out publicly, but also internally, a message to everybody asking for new solutions, new ideas. I think the step change for most organizations is when they have a person whose job description is to coordinate or work on innovations. It doesn’t have to be a full-time role, but at least one person that’s ideally somebody who has a background in working on digital products, on new business models, or has been starting new things before.