Multinational corporations hold significant influence over the global Artificial Intelligence (AI) landscape, shaping its development through their resources, technologies, and strategic priorities. In Africa, this influence is amplified by under-regulation and the persistent effects of digital divides, which create conditions favorable to corporate power. While these corporations can support local ecosystems through initiatives such as training and skill-building, their profit-driven motives often take precedence. By prioritising proprietary datasets and restricting access to research, they limit opportunities for broader ecosystem growth and innovation.
In Ghana, two pioneering AI companies, Mazzuma and Ghana NLP, are challenging this model. By showing how smaller startups and community groups can fill an ‘empowerment gap’, their work reflects alternative approaches to technology development, emphasising cultural preservation, local investment, and sustainable innovation within Africa’s tech ecosystem.
AI consultant Kojo Apeagyei conducted semi-structured interviews with the founders of both companies to explore their approaches to AI development. This blog, by Apeagyei, is part of an independent study examining alternative models of AI development that empower local communities and contribute to sustainable growth in Africa.
Mazzuma and Ghana NLP: Building localtools with global reach
Mazzuma is a mobile payment platform integrating AI and blockchain to facilitate transactions. Headquartered in Accra, its flagship product, MazzumaGPT, generates smart contract code—digital agreements for managing specific actions or events—from natural language prompts. Mazzuma was founded to enable the financial inclusion of “businesses and developers” in Ghana and beyond, by facilitating their participation in the global payments ecosystem. Since the launch of MazzumaGPT in 2023, the platform has grown to serve users in 85 countries, enabling participation in the global blockchain economy.
Ghana NLP, based in Cape Coast, focuses on natural language processing (NLP) for Ghanaian languages and their application to local challenges. It was founded based on the “realisation that language technologies have been leaving behind… most African languages”. Its flagship product, Khaya, is the first AI translator tailored for languages such as Twi, Ga, and Fante, and it is expanding to include languages from neighbouring countries, translating languages spoken in Nigeria, Burkina Faso, Kenya, and Tanzania.
These companies demonstrate how locally developed and tailored technologies can address both community-specific and global challenges.
Empowering through Open Source
Open data and local resources have been instrumental in allowing Mazzuma and Ghana NLP to empower communities. Both Kofi Genfi (co-founder of Mazzuma) and Paul Azunre (founder of Ghana NLP) are advocates for open source—but with varying degrees of emphasis. Genfi highlights the use of open-source data in making AI development accessible, enabling developers to build specialised, relevant applications. While Azunre views creating local data sources and training Ghanaians in AI as a necessary investment into the digital ecosystem, he also cautions that the benefits of open data depend on communities retaining control of their data. Without such control, Azunre warns, big tech companies can use these datasets to build solutions that are sold back to the same communities, perpetuating a value-extraction model.
This nuanced approach to open-source data addresses a significant challenge in today’s digital ecosystem—much of Africa’s linguistic and cultural heritage is missing from global AI datasets, and where present its ‘ownership’ is contested. Africa has over 1 000 spoken languages, yet fewer than ten are typically used in training AI models. The absence of high-quality local datasets forces African AI systems to rely on foreign datasets, often lacking cultural and contextual relevance. By developing high-quality, culturally aligned datasets, Ghana NLP and Mazzuma are addressing these gaps. To this, Azunre describes Ghana NLP’s work as “setting a benchmark that keeps larger tech companies honest” ensuring that these corporations cannot ignore the importance of culturally attuned and locally controlled AI solutions.
The role of the public sector
An open-source ethos aligns with a growing movement among African AI practitioners and advocates calling for community-focused foundational systems for data and AI. However, limited public-sector involvement has hindered progress. For instance, Ghana’s Open Data Initiative, launched in 2012 to promote the reuse of government data, has faded into obscurity, reflecting a reluctance or inability to share data despite the protections of the Right to Information Act.
Public-sector participation is critical to advancing the African AI ecosystem. Government bodies could prioritise the creation and sharing of local datasets—or incentivise private-sector contributions—to drive innovation not only in the tech industry but also in sectors like transportation, education, and healthcare. Unfortunately, public officials often hesitate to share data due to governance and resource constraints, according to discussions had at a recent workshop by GFA Consulting Group and the Africa Center for Economic Transformation,
Genfi emphasises the need for public-private collaboration to bridge these gaps. “It should be public-private partnerships where companies and governments sit down and design systems together,” he explains, advocating for governance frameworks that enable secure and equitable adoption of technologies like AI and blockchain, maintaining trust and longevity.
An alternative vision
While there is always the chance of regulatory capture, where corporate interests dominate public interest, without envisioning AI development in Africa differently there may never be AI products that fully meet the needs of people on the continent. For Genfi, this means curating datasets that might otherwise remain inaccessible, empowering developers to create localised and specialised applications. “We use a two-pronged approach, training AI models with publicly available data while gathering feedback directly from companies and communities to improve our models,” he explains.
Azunre’s vision for Ghana NLP is similarly rooted in self-determination. He critiques “technological colonisation”, where foreign entities extract data and profits while offering limited value to local communities. “We strive to invest in and build an AI ecosystem that generates direct benefits for local communities”, Azunre explains, emphasising the importance of retaining resources like data within the continent.
Reclaiming agency in Africa’s AI ecosystem
The efforts of Mazzuma and Ghana NLP highlight how local initiatives can reclaim agency within Africa’s tech ecosystem. By prioritising open-source development, fostering community partnerships, and advocating for public-sector collaboration, these companies are addressing systemic inequities and creating pathways for meaningful innovation. Their work demonstrates that AI can be used by Africans to shape their futures.
Kojo Apeagyei is an AI consultant who holds an MSc in Media and Communications (Data and Society) from the London School of Economics. His research explores the ethical design and deployment of AI technologies in African contexts.
