top of page
  • Writer's pictureJulian Garratt

AI Potential in Africa: Reflections from the 2023 IMF & WB Annual Meetings

Updated: Apr 15

Written by Julian Garratt, 2023 IMF/WB Fellow

Leading into the 2023 World Bank (WB) and International Monetary Fund (IMF) Annual Meetings in Morroco, I expected economics and international relations to dominate the discussion. Little did I predict that Artificial Intelligence (AI) would become a leitmotif in the numerous sessions throughout the weeklong conference. However, one perspective remained salient: AI’s ability to leapfrog potential in Africa. AI is estimated to contribute 15.7 trillion dollars to the global economy by 2030 with the greatest economic gains predicted to be captured by China and the US, accounting for 70% of the global economic impact (PwC, 2017). More recently in an analysis by the IMF, advanced economies are 36% more exposed to AI than low-income countries (Georgieva, 2024). Although advanced economies will likely see a tightening job market and lower wages, more automation implies enhanced productivity. Low-income countries are likely to see less of an effect on both fronts mostly deriving from weaker infrastructure, investment and a smaller labour market. However, with Africa constituting 42% of the global youth by 2030 (Habti, 2022), the continent often overlooked as an economic powerhouse is poised to capitalise on the AI revolution.

Despite a smaller workforce equipped to develop AI systems, there has been a distinct push to upskill the working population in the new AI frontier. AI Saturdays is a global movement focussed on upskilling communities of enthusiasts and professionals alike in cutting-edge AI technology. In Africa, AI Saturdays is run by communities in Nairobi, Kenya and Kigali, Rwanda. In conjunction, the 2025 1 Million Women in Intelligent Automation Initiative is an AI development and leadership program aimed at up-skilling a million women in AI and Intelligent Automation by 2025. Aside from grassroots movements, Google has launched a 10-week equity-free accelerator program for African startups developing Africa-centred solutions that utilise advanced technology such as Machine Learning and AI6. 

That is not to say that AI hasn’t already impacted African industry. UjuziKilimo is a Kenyan-based agriculture company that uses Machine Learning and various analytics to supply farmers with data-driven decisions. In the health sector, Ubenwa is a Nigerian startup to address the high rate of newborn mortality using AI to detect Perinatal Asphyxia. Africa’s AI footprint also spans globally with the record-breaking sale of the Tunisian-founded AI company InstaDeep for approximately 500 million euros (InstaDeep, 2023). These examples aren’t outlier successes, however, the African AI landscape bears numerous challenges regardless.

Applying AI to a problem is tumultuous regardless of the country, the barriers only seem to be more exacerbated in Africa. When we use AI products like ChatGPT, most people write in English, however, Africa is estimated to have up to 2000 languages (Harvard University, n.d.) which are mostly underrepresented in modern large language models (LLMs) (Ojo et al., 2023). This partially stems from a need for more data on African dialects required to train LLMs but ignores the wider barriers to electricity and internet access in Africa that would allow for an efficient transfer of local knowledge and data. Aside from Africa’s connectivity constraint, a significant challenge to building AI-powered tools is the demand for highly educated workers. Illustrated in a session at the 2023 WB & IMF Annual Meetings titled “Knowledge Café - Skills and Workforce Development for the Green and Digital Transformation”, 1 billion workers will need to be upskilled in the next 10 years. This is anticipated to arise from digital forms of education such as Khan Academy as the primary engine for a skilled workforce diverges from formal education. 

Africa’s internal issues aren’t the only forces acting against AI adoption in Africa. Graphic Processing Units (GPUs), the main workhorse for building and using AI applications, aren’t available in Africa on the most popular cloud platforms run by Amazon, Google and Microsoft as of February 2024. Although the underlying reason may stem from a lack of AI investment in Africa, the true reason may derive from a combination of challenges mentioned previously which in aggregate has reduced the demand for AI in Africa. Yet, the biggest threat to leapfrogging potential in Africa is the centralisation of AI in large tech companies. When the algorithms for AI tools are hidden from the public, so too are the underlying biases of the creators. Instead, as we see currently, a large majority of state-of-the-art LLMs are already available to the public free of charge in a movement known as “open-source”. For entrepreneurs, researchers and enthusiasts alike, open source promotes total accessibility without the imposed biases which has the potential to progress Africa’s AI economy.

As profits and media attention spur global interest, it is not unexpected that the conversation has turned towards regulation. African countries such as Egypt, Nigeria, Angola, Kenya and South Africa have all either introduced an AI ethics framework or begun developing bills related to AI and robotics (Sun, n.d.). In the meantime, AI systems will continue to improve and are more likely to outgrow the regulations and frameworks poised to contain them. This is no different in Africa, which at the 2023 Annual Meetings was referred to as “an early AI startup”. Hence, it should be unsurprising that African AI companies and researchers should crave the same investment and acceleration as AI in advanced economies. However, leapfrogging AI potential in Africa is more than just investment or regulation, rather it’s about allowing solutions to be created in Africa, for Africa.

References: Georgieva, K. (2024). AI Will Transform the Global Economy. Let’s Make Sure It Benefits Humanity. International Monetary Fund. 

Habti, H. E. (2022). Why Africa's youth hold the key to its development potential. 

Harvard University. (n.d.). Introduction to African Languages. 

InstaDeep. (2023). BioNTech Completes Acquisition of InstaDeep.

Ojo, J., Ogueji, K., Stenetorp, P. and Adelani, D.I. (2023). How good are Large Language Models on African Languages? [online] doi:

PwC. (2017). Sizing the prize: What’s the real value of AI for your business and how can you capitalise?

Sun, R. (n.d.). Global AI Regulation Tracker. 

19 views0 comments


bottom of page