Africa Needs AI That Serves Its Priorities

By Dr Girmaw Abebe Tadesse

Across Africa, some of the most important uses of AI do not look like chatbots.

They look like forecasts that warn nutrition teams about a crisis, satellite maps that show changing land use, or drought assessments that reach ministries in time to act.

Much of this work relies on data beyond language, including imagery, health records and weather information. Yet language often determines whether AI can influence a real-world decision.

Building AI That Speaks Africa’s Languages

One in six people worldwide has now used a generative AI product, according to Microsoft’s 2025 AI Diffusion Report. Africa has about 1.5 billion people and more than 1,500 languages. However, most AI models still rely heavily on English and a small number of global languages.

A farmer seeking planting advice in Dholuo or a mother looking for health information in Amharic may therefore struggle to use today’s AI systems.

Language is not everything in Africa’s AI story. Without it, however, much of the technology’s potential will struggle to reach people.

Encouraging progress is already coming from within the continent.

LINGUA Africa, an initiative involving the Masakhane African Languages Hub, the Gates Foundation, Microsoft AI for Good Lab and Google.org, supports open datasets, speech resources and language tools.

Its recent call attracted more than 800 applications from 64 countries. About 85 per cent came from Africa. The 26 selected projects cover more than 50 African languages across 47 countries.

Projects include work on Tanzanian Sign Language and efforts to make vaccination information accessible in Soussou, Pular and Maninka.

Voice also matters on a continent with strong oral traditions.

Microsoft Research Africa’s Paza project is improving speech recognition for low-resource languages. The project has developed a benchmark covering 39 African languages, alongside new models for Swahili and five Kenyan languages.

Researchers have also tested the technology with farmers using ordinary mobile phones, including in noisy environments and areas with patchy connectivity.

From Language to Local Relevance

However, fluency does not automatically make AI useful.

A system can answer a farmer in fluent Kikuyu and still offer advice that does not fit the soil, season or household budget.

That means language must go hand in hand with local knowledge and context.

Data scarcity in Africa also goes beyond having fewer examples. It can mean missing communities, outdated maps and health records that capture only people who reached a clinic.

Models trained on such data can inherit the same blind spots.

Locally led data collection, documentation and long-term data stewardship therefore deserve as much attention as the models themselves.

At the same time, Africa cannot wait for perfect datasets or computing infrastructure before building useful AI. Solutions must work within the conditions people face today.

Using AI to Solve Real Problems

At Microsoft’s AI for Good Lab, the starting point is the decision that needs to improve, rather than the model researchers want to apply.

In Kenya, the lab worked with Amref Health Africa and the Ministry of Health to combine routine health records with satellite measurements of vegetation.

The project developed a model that can forecast acute childhood malnutrition up to six months in advance. Microsoft describes the work as a promising proof of concept, rather than proof of improved health outcomes.

The same approach applies to geospatial work.

Together with the Kenya Space Agency, the team developed land-use maps tailored to local landscapes. The work found that locally trained models can outperform global, one-size-fits-all models.

Other research uses satellite imagery to estimate building density and height. This includes settlement growth around a refugee camp in Chad.

Such information can help humanitarian planners understand population changes and identify communities that may struggle to receive early warnings through existing digital infrastructure.

Giving Local Institutions a Bigger Role

For AI to create lasting impact, local institutions need to own its development, deployment and long-term stewardship.

That thinking shaped the launch of ADAPT-Kenya, a national initiative led by Kenya’s Ministry of Agriculture and Livestock Development. The initiative receives support from the Gates Foundation and Microsoft’s AI for Good Lab.

Nearly 50 partners are involved.

ADAPT-Kenya brings together local knowledge and satellite observations to develop AI-powered agricultural data products. These tools aim to improve crop monitoring, harvest forecasting, market access and trade.

They can also support insurance services and disaster risk reduction.

The approach has already faced a real-world test.

When drought affected maize-producing counties in Kenya, ADAPT partners, including NASA Harvest, assessed conditions and provided decision-makers with a timely picture of the harvest.

The team plans to apply the same approach during the coming El Niño season.

Infrastructure Is Part of the AI Agenda

The hardest part of AI is not always the algorithm.

Around 600 million people in sub-Saharan Africa still lack electricity. Connectivity, devices, skills and maintenance can determine whether an impressive demonstration becomes a service people use every day.

These challenges should not sit outside Africa’s AI agenda.

They are part of it.

Africa is also not a single dataset, market or deployment environment. Its people should help define the problems, methods and standards used to measure AI progress.

Grassroots AI communities can lead this work alongside universities, SMEs, non-profits and UN organisations.

Governments will also remain central because they play a major role across almost every sector.

Africa Needs AI That Serves Its Priorities

Building AI for Africa’s Realities

Africa should aim for more than AI that speaks its languages.

The continent needs AI that understands its realities, strengthens its institutions and helps people make better decisions.

Language belongs at the centre of that ambition. So do good data, earned trust and the capacity to act.

Africa-centric AI is already becoming an important area of research and development, with Microsoft Research describing it as an approach that starts with African contexts when defining problems, developing solutions and deploying technology.

The opportunity now is to move from promising demonstrations to systems that people and institutions can trust, use and sustain.

By Dr Girmaw Abebe Tadesse

Lead Africa team of the Microsoft AI for Good Lab in Nairobi