Publications

The Leapfrog Conditions: What Mobile Money Predicts About AI in Africa

By Natalie Gill-Mensah · August 2026

Every year brings another wave of headlines warning AI will hollow out jobs. The warning almost always lands on the same audience, white-collar workers in wealthy economies watching their spreadsheets and their writing edge toward automation. The World Bank’s new World Development Report on AI tells a different story. Economies where more of the workforce does more manual work carry less automation risk than high-income economies do. Their jobs sit closer to augmentation, workers getting help instead of losing the work altogether.

This isn’t the first time the “they’ll be left behind” story ran backwards. It happened with money.

In 2007, Safaricom launched a mobile money service in Kenya called M-Pesa, and it’s worth remembering how unlikely a candidate Kenya was for a financial technology revolution. Before M-Pesa, roughly a quarter of Kenyans had access to a formal bank account. Money moved the way it always had: cash, carried by hand, or entrusted to a bus driver heading toward a relative’s village. There was no trusted, reliable channel for something as basic as sending money home.

What Kenya did have was phones. Mobile penetration was already climbing well before smartphones existed, on a network built for calls and text, not banking. Safaricom didn’t need to build a new distribution system. It repurposed the one that already reached nearly everyone: its own airtime agents, who became the cash-in, cash-out points for M-Pesa.

The last piece was regulatory. Kenyan banks pushed back, arguing a telecom company had no business running what amounted to a financial institution. The Central Bank of Kenya disagreed and let it proceed.

The result surprised even the people who built it. Within two years, more than half the country was sending money through M-Pesa. Formal banking access rose from roughly a quarter of the population in 2006 to 68 percent by 2014. By 2013, mobile money transactions equaled something close to 43 percent of Kenya’s GDP, flowing through a service that hadn’t existed a decade earlier.

Strip the story down and four conditions made it possible:

1. No legacy system to defend. Kenya had no mature banking sector with a financial stake in keeping mobile money small.
2. The distribution layer already existed. Phones were everywhere before banking was.
3. An acute need, years in the making. Getting money safely across distance was a daily problem with no good answer.
4. A regulator willing to say yes. The Central Bank chose to let a new model run rather than protect the old one.

None of these are Kenya-specific. They’re the conditions under which any leapfrog becomes possible, and they explain why the same opportunity, sitting in front of a different country, can produce a completely different outcome.

Nigeria had the same unbanked population, arguably a larger one, and the same rising mobile penetration Kenya did. What it didn’t have was condition four. Nigeria’s central bank chose a bank-led model for mobile money, licensing banks rather than telecom operators to run it, which meant the service couldn’t piggyback on an existing distribution network the way M-Pesa had. By 2014, Nigeria counted roughly 800,000 mobile money users against a population of 178 million, a rounding error next to Kenya’s numbers. Same need, same infrastructure, same technology. One different regulatory choice, and the leapfrog didn’t happen. The conditions are a checklist, and missing even one changes the outcome.

A second case makes the pattern harder to dismiss as a fintech story. In 2016, the Rwandan government began working with a California drone company called Zipline to solve a problem that had nothing to do with money: getting blood and medical supplies to rural clinics across a country where paved roads and reliable transport were scarce. Rwanda didn’t commission a feasibility study that dragged on for years. It partnered, launched, and let the model prove itself.

It did. Zipline’s drones now deliver roughly three-quarters of Rwanda’s blood supply outside the capital, turning what used to be a multi-hour trip over dirt roads into a delivery measured in minutes. Independent research associates the program with as much as a 51 percent drop in maternal mortality in Rwanda and a 56 percent drop in Ghana, where the model expanded in 2019, along with roughly 60 percent fewer stockouts of essential medicines and a 37-percentage-point increase in immunization coverage in the areas served. The program has grown enough that in November 2025, the U.S. State Department backed a $150 million expansion meant to take the network from 5,000 to 15,000 health facilities, reaching an estimated 130 million more people.

Run it through the same four conditions. No legacy logistics system to protect, since roughly 600 million people across Sub-Saharan Africa live without reliable roads to begin with. A need that predates the technology by decades, blood and vaccines that couldn’t reach the people who needed them. A distribution layer that had to be built rather than repurposed, which is the one place this case departs from M-Pesa’s script. And a government willing to partner and move rather than study the idea to death.

Different country, different technology, different sector entirely. Same four conditions, same result.

Test those four conditions against AI in Africa today, and the picture is not uniform, which is exactly what makes it worth taking seriously instead of treating it as inevitable.

Condition one holds in more places than people assume. A meaningful share of African institutions never built the twenty-year-old enterprise software stacks that Western organizations now have to plan around and unwind. There’s no legacy system standing in the way of going AI-native, because there’s no legacy system.

Condition two is arguably stronger than it was for M-Pesa. Mobile-first, often multilingual populations are already fluent in phone-based interaction, and increasingly voice-based interaction, which happens to be the interface AI is converging toward anyway.

Condition three barely needs arguing. The list of acute, decades-old needs is long and specific, health worker shortages that predate any AI conversation, agricultural extension services that have never reached most smallholder farmers, education systems stretched past capacity, financial inclusion that, M-Pesa notwithstanding, is still incomplete.

Condition four is the honest complication, and it’s worth stating plainly rather than glossing over it. Momentum is real. Kenya launched a National AI Strategy running through 2030. Rwanda was one of the first countries on the continent to adopt a national AI policy. The regional trade bloc COMESA is now running AI strategy consultations across 21 member states, with Kenya and Zambia first at the table. But as of this year, only 15 of Africa’s 54 countries have an AI strategy in place at all. The regulatory yes that made M-Pesa and Zipline possible is forming. It isn’t settled, and it won’t arrive evenly.

That gap is the entire opportunity. The African Development Bank projects AI could add up to a trillion dollars to Africa’s economy by 2035, close to a third of the continent’s current output, along with 40 million digital jobs. That number is less a forecast than a contingency, resting on whether the fourth condition catches up to the other three, market by market, regulator by regulator.

Kenya didn’t wait until most of the country had a bank account to build a financial system. Rwanda didn’t wait until its roads were paved to build a supply chain. In both cases, the absence of an established system wasn’t a disadvantage to overcome. It was the opening.

The countries and organizations that treat AI as a chance to build the system that was never affordable before, rather than a tool bolted onto the one they’ve already got, are the ones likely to produce the next version of this story. The conditions don’t care how wealthy a country is. They care whether the need is real, whether the distribution already exists or can be built fast, whether there’s nothing propping up the old way, and whether someone with regulatory authority is willing to say yes before the rest of the world finishes debating it.

Call it a pattern rather than a prediction, one that has already run at least twice on the continent. The only open question is where it runs next, and how many people are paying enough attention to recognize it when it does.

Sources: World Bank World Development Report 2026, African Development Bank, COMESA, and reporting from the Institute of Economic Affairs and Think Global Health.

Views expressed here are my own and do not represent those of my employer or the U.S. government.