As an AI application enthusiast with limited technical background but deep cultural knowledge, a phenomenon I have seen first-hand is when types of artificial intelligence or large language models (LLMs) fail spectacularly upon encountering African contexts.
AI remains deficient when it comes to many non-western cultures, but current LLMs such as Claude, Gemini, Chat GPT and Deep Seek fall particularly short when it comes to drawing from diverse African experiences, local nuance and specific cultural context. Training data is often absent, and even if there are decent references, the cultural framework is often weak.
I am under no illusion that co-ordinating cultural knowledge across 54 countries and more than 2,000 languages is difficult, but we don't need perfection. What we need is momentum through networks where each nation builds its own cultural AI foundation while sharing methodologies and learning across borders.
As of now, the results AI shows are frequently biased, stereotypical and generic. I recently asked for specific information about Kenya, and I received responses regarding a “Harambee framework” (the concept of working together) and how Kenyan business culture is about “Ubuntu” (an operating system). This is either not correct, or depending on the context, lacks nuance and adequate explanation.
Given my own local knowledge, I can point out gaps and direct the LLM for a better response. When I asked Claude AI to critique its own cultural knowledge and training about Kenya, it responded by saying: “I could translate words but miss the cultural weight behind code-switching. I lack understanding of regional language variations and their social significance. The cultural intelligence you're building can't come from language data sets alone – it needs the lived experience of when, how and why people communicate in different languages. AI systems (including me) have superficial language knowledge without cultural depth.”
The honesty was refreshing!
But what if this weakness could become Africa's greatest AI opportunity? Talented African storytellers have an opportunity to apply deep local cultural knowledge to existing tools, or develop new applications in local language that solve problems. This also gives us narrative ownership through AI from which we can create authentic African perspectives.
I recently developed a “covering Africa” prompt for journalists to use as a smart sounding board that flags stereotypical stories about the continent. I applied 20 years of journalism experience, combined with cultural and detailed local knowledge about the continent and the way we Africans, in all our diversity, see ourselves.
The continent has transformed but stories about the African people often follow the same tedious beat. But a sensitive and culturally aware prompt offered a useful storytelling solution.
Although Africa may not have fully competitive engineering skills or large-scale investment, it has cultural depth, and when applied locally, that can be its superpower.
Here's just one example to illustrate how this can be leveraged. At The Rundown Studio, a cloud-based platform, we have developed a newsroom co-pilot that can work very well in African news operations and markets. Using the highest international editorial standards and approaches, we can empower all newsrooms in Africa and other emerging markets with world-class editorial, verification and on-air systems. This demonstrates how deep local knowledge can be applied alongside LLMs.
It is not just about working with tools though. African cultural knowledge must become the training ground for next-generation AI systems. Instead of merely fixing western models, our leaders, entrepreneurs and innovators must position Africa as the laboratory for the culturally intelligent AI that the world needs. This takes us from seeing Africa as merely a market for AI products to Africa as the architect of AI that understands humanity's full cultural spectrum.
This should start by national AI cultural councils being created in each country. Funded through public-private partnerships, here storytellers, linguists and technologists could collaborate to build comprehensive cultural datasets that become the foundation for training truly inclusive AI models.
One practical application is that every country should develop its own national prompt; their own quintessential presentation of their nation to the world, that influences a person’s perception or narratives when LLMs are queried.
I am working on a collaborative process for Kenya to create its own prompt that would be our north star. It will contain critical components that capture a diverse nation and peoples through a cultural lens. Oral culture, local nuances, languages and various contexts can be crafted with hyper local knowledge and lived experiences.
There are a lot of people and organisations already doing great work. We are not trying to build one monolithic system for Africa; what we should aim to do is create a decentralised framework where local groups operate independently but share open-source tools and governance principles. This allows organic growth rather than top-down co-ordination.
African talent needs access to AI tools, but subscription costs and data barriers remain prohibitive. We need funded programmes that democratise this technology while building cultural knowledge ecosystems.
This would allow us to create volume and put guardrails in place to either enhance existing western and Chinese LLMs or provide the data to our own language models. At the core is our own economic transformation through jobs and investment rooted in what we know best. If Africa captures 10 per cent of global AI adoption, AI could contribute $1.5 trillion to the continent’s economy by 2031, driving GDP growth from 5.2 per cent in 2025 to 8.5 per cent by 2030.
For Africa, the solutions lie in our authenticity, languages and lived experiences. We can become the template for culturally intelligent AI that the world desperately needs.
