‘Why Nigeria’s AI ambitions need more than innovation’

As Nigeria’s federal and state governments accelerate the integration of artificial intelligence into public infrastructure, from loan disbursement platforms to social benefit registries and employment systems, deployment may be outpacing the governance needed to support it.

“The conversation in Nigeria has been dominated by what AI can do,” says Jessica Aletor, a UK-based AI and data scientist specialising in data governance and responsible AI. “What we have not invested in nearly enough is the question of what AI does when it fails, and who bears the consequences.”

When the data reflects the past

Aletor, who holds an MSc with Distinction in Artificial Intelligence and Data Science and is a recipient of the Office for Students AI Scholarship for Women, brings a perspective shaped by technology environments across the UK and emerging markets. She argues that the core risk is not the technology itself but the data infrastructure it depends on.

“When we automate a decision, whether it is who qualifies for a government intervention or a small business loan, we are encoding the patterns of the past into the decisions of the future,” Aletor explains. “If the historical data reflects existing inequalities, the system may reproduce and potentially amplify those inequalities. Not because it is malicious, but because that is what it was trained to do.”

This dynamic can be particularly pronounced in contexts like Nigeria, where data collection has historically been uneven across regions, sectors and demographics. Systems trained on incomplete or unrepresentative data may produce what Aletor describes as “silent failures”—errors that do not announce themselves, accumulate over time, and may disproportionately affect the communities least equipped to challenge them.

“A silent failure is not a crash. It is a loan that was never approved, a business that was flagged without reason, a person who simply fell through the system and never knew why.”

The infrastructure problem

The challenge is structural. Nigeria’s National AI Strategy, published in 2024, outlines ambitious goals for AI-driven public services, particularly in large-scale programmes such as social investment schemes and digital loan platforms. But implementation may be moving faster than the governance architecture required to support it.

In practice, analysts have identified potential gaps: limited requirements for explainability in automated decisions affecting citizens, evolving mechanisms for appeal, and the absence of structured impact assessments before deployment in some contexts.

In the UK, these conversations are increasingly embedded in regulatory and operational frameworks. In Nigeria, they are still largely confined to policy and academic discussions and may benefit from deeper integration into implementation as systems continue to develop.

Large-scale public programmes provide a useful lens. Economic recovery funds and similar interventions have demonstrated what is possible when digital systems are deployed quickly and at scale for public good. They have also highlighted potential risks when speed outpaces verification, communication and accountability. In some cases, decisions may be made faster than the systems designed to explain or challenge them.

Resilient AI: a different measure of success

Aletor advocates for what she calls “Resilient AI”, a framework that shifts the measure of success from speed and scale to transparency, auditability and local relevance.

“We have been measuring AI success by how fast it processes applications, how many decisions it automates, how much it reduces human workload. Those are efficiency metrics. They are not accountability metrics. A system can be extremely efficient and still produce unfair outcomes if safeguards are not in place.”

Resilient AI requires three things. First, systems must be built on data that more accurately represents the populations they serve, including investment in local data collection and validation. Second, automated decisions should include human oversight mechanisms, particularly in high-stakes contexts such as financial access, employment and social benefits. Third, affected communities should have clear and accessible routes to understand and challenge decisions made about them.

“AI systems do not operate in isolation. Their performance is inseparable from the quality of the data they are trained on, the governance structures around them, and the ability of institutions to monitor and correct them over time.”

A window that will not stay open

Nigeria is at an inflection point. The country has a young, digitally engaged population, a growing base of technical talent, and increasing government appetite for AI-driven public services. But the window for establishing strong governance norms may be narrowing.

“The decisions being made now about how AI is deployed in Nigeria’s public sector will shape outcomes for years,” Aletor says. “It is far more difficult to retrofit accountability into a system that is already running than to build it in from the start.”

The countries that will lead in artificial intelligence will not be those that move the fastest, but those that build systems people can trust. That trust is not a byproduct of innovation. It is a design decision.