The algorithm-based welfare state: how Africa is developing the automated safety net

By Andrea Menabò

When people talk about AI, they relate it with a useful tool that makes an already efficient and effective system perfect. But from the African’s perspective,  AI is seen as a pivotal tool in the creation and development of the welfare state. And while the EU debates the ethics of the AI, Africa is implementing it in the development of the “Automated safety net”, pushing the entire continent in “leapfrogging” towards an automated and algorithmic-based welfare state. 

“This strategy puts forward an Africa-centric, development-oriented and inclusive approach around five focus areas notably: harnessing AI's benefits, building AI capabilities, minimising risks, stimulating investment and fostering cooperation.”

By the African Union Continental AI Strategy

By focussing on this strategy is it possible to understand how Africa has the capacity to develop a “Digital Public Infrastructure” to improve the quality of its digital services, and first and foremost, to guarantee citizens a better standard of living, especially to the vulnerable communities. This process allows states to allocate resources more effectively, improve decision-making and policy-making, and better manage the maintenance of essential infrastructure (water and electricity networks).

Source: generated with Gemini AI

The digitalization process of the public sector is also a useful way to train AI algorithms, and through the creation of national open data portals, government-led digital datasets can be made accessible to the public. The African Union’s strategy, therefore, aims to develop interconnections between the public sector, the private sector and universities, making it the fundamental centre of Africa’s development, increasing intra-continental cooperation to achieve the UN Sustainable Development Goals. 

Moving forward from the African Union’s strategic framework to better understand how AI is implemented to create the “Automated Safety Net”, it is necessary to analyse the report by the World Food Program “Saving lives, time and money: Evidence from anticipatory action”. 

The report highlights that for every 1 USD invested in anticipatory action, up to 7 USD in losses are avoided. It further notes that this approach helps keep inflation stable, that time savings has been empirically proven in the field, and that anticipatory actions are a pivotal tool for fostering human dignity and resilience. However, what is truly relevant are the two mechanisms powering these actions, namely: “Machine Learning for Anticipatory Actions (ML4AA)” and “SKAI”. 

ML4AA shifts the paradigm from reaction to prevention. By fusing cutting-edge AI with expert climate insights, this initiative sharpens droughts and floods predictions to ensure that warnings and life-saving assistance reach vulnerable communities before disaster strikes. For example, in South Sudan ML4AA predicted a flood, protecting roads and transporting supplies before flooding was four times cheaper than airlifting them after.

SKAI, instead, is an open-source tool used for disaster response and humanitarian aid, developed in collaboration with Google. SKAI leverages advanced AI to analyse satellite imagery, automatically assessing building damage on a massive scale in near-real-time. By providing immediate situational awareness, it drastically shortens the window between disaster and relief, ensuring humanitarian assistance reaches affected households faster than ever before.

The final piece to understand the development of the “Automated Safety Net” is the Word Bank’s “Novissi Togo” project. This project is today considered the continental “gold standard” because it has solved the problem: “How do you find the poor when there are no tax returns or registries?”. Fusing a low-tech approach (Novissi 1) for the urban areas, a high-tech approach (Novissi 2) for the rural areas, and geospatial data it was possible to create a “fine-grained poverty maps to locate the least wealthy cantons”. The government of Togo has collaborated with a solid number of stakeholders, such as Mobile network operators, academics and research teams, call centre, Ministry of Economy, National Statistics Office and many others. 

But did the project work? By analysing the relevant data it is possible to consider that the “Novissi Togo” project had a massive success and has developed a global benchmark in reaching and helping the “invisible” by using AI to create a “poverty score” combining phone metadata, satellite images and call logs.

In conclusion, it can be said that today Africa is not merely observing the AI revolution; but rather considers it a pivotal tool to build its future. Countries have understood the potential of this tool and are ready to get involved to develop and achieve concrete goals. The “leapfrog” is not an isolated phenomenon, but it’s grounded in the principle of “common but different responsibilities”, considered the foundation of sustainable development and international cooperation.

Sources:

“Continental Artificial Intelligence Strategy | African Union”. https://au.int/en/documents/20240809/continental-artificial-intelligence-strategy.

“High-Level Summary of the AI Act | EU Artificial Intelligence Act”. https://artificialintelligenceact.eu/high-level-summary/.

“ML4AA | WFP Innovation”.  https://innovation.wfp.org/project/ml4aa. 

“Novissi Togo: Harnessing Artificial Intelligence to Deliver Shock-Responsive Social Protection”. https://documents1.worldbank.org/curated/en/099751009222330502/pdf/IDU0e83f857301ff1047bf082710a8d21ddf42c3.pdf

“SKAI | WFP Innovation”.  https://innovation.wfp.org/project/SKAI. 

“2025 – Saving Lives, Time and Money: Evidence from Anticipatory Action | World Food Programme”. 19/05/2025. https://www.wfp.org/publications/2025-saving-lives-time-and-money-evidence-anticipatory-action.