Africa’s AI story has spent the last few years mostly in the register of promise — pilots, hackathon winners, funding rounds. This week’s news suggests a shift is underway: four separate stories, four different countries, and in each one, AI is no longer the pitch — it’s the product already in someone’s hands.
The Fintech That Gave Up on Banks to Beat Them
Precious Arikeri’s problem started the way a lot of fintech problems do: during COVID lockdown in 2020, she couldn’t figure out where her money was going. Working with software engineer Wahab Balogun, a former colleague at Lagos IT firm Vascon Solutions, she built Trackfundx in 2022 — a budgeting app that pulled transactions directly from Nigerian bank accounts using the open-banking API Mono.
It worked, until it didn’t. Mono was charging roughly ₦200 per API call, a cost that scaled painfully with every user. Linking an account sometimes logged users out of their own banking apps. And when Arikeri moved to the UK for a master’s degree in 2022, the app she’d built to solve her own problem stopped working the moment she needed it most — it had been built entirely around Nigerian bank rails.
So the founders threw out the premise. Instead of parsing bank logs to guess what someone had bought, Balogun proposed letting AI read the receipts directly — a method with no dependency on any country’s banking infrastructure at all. The rebuild, launched in 2024 as Expense AI, let it work anywhere a phone camera could photograph a slip of paper. Today the company says it has crossed 10,000 users in more than 100 countries, processing over 80,000 expenses, still run day-to-day by just the two founders. Subscriptions start at $8.99 a month, priced locally at ₦2,500 in Nigeria. The next target: a business-focused version, a million users, and $100,000 in annual revenue.
The Farmer Credit Score Nobody Else Was Building
In Kampala, Sandra Nabakka was solving an older, more stubborn problem: smallholder farmers with land, experience, and functioning markets who still can’t borrow money, because they lack the formal payslips, fixed assets, or credit history conventional lenders require. Her startup, SANDI AI, uses farm and agricultural data to give lenders a different way to assess a farmer’s risk — building the equivalent of a credit history out of information the formal banking system was never designed to capture.
“Farmers should not be limited by systems that were not designed around their realities,” Nabakka said. The approach won SANDI AI the $50,000 Grand Prize at the 2026 GoGettaz Agripreneur competition, held at the Africa Food Systems Forum in Kigali this month — recognition that arrives as the startup is also applying AI to irrigation, combining soil-moisture sensors and weather data to help farmers use newly financed equipment more efficiently once they have it.
The Sign Language System Built From a Childhood
The most personal of this week’s stories comes from Pietermaritzburg. Akhil Hansrajh grew up with two parents who were both born hard of hearing, and spent much of his childhood accompanying them to banks, government offices, and municipal counters to speak on their behalf. For his final-year computer engineering project at the University of KwaZulu-Natal, he built something to change that: a system that watches a person sign through a camera and speaks what they’ve signed, out loud, in real time.
The technical challenge was steeper than it might sound. South African Sign Language — recognised only recently as the country’s twelfth official language — differs meaningfully from American Sign Language, which is what most existing research and tools are built around. Hansrajh trained his own AI model on a custom dataset of his own recordings, tracking hand, arm, and facial movement over time to identify each sign before a text-to-speech layer converts it to spoken words. He built it while working full-time to cover his final year of study after running into financial difficulty — and earned a highly commended award from UKZN’s engineering faculty for the finished system. He’s not stopping at the prototype: his next phase focuses on recognising full words and producing more fluent, natural speech.
The Government Trying to Get Ahead of All of This
The fourth story is the odd one out — not a founder, but a regulator, and it lands with real weight given where AfriWave’s readers have watched this conversation unfold before. Mozambique’s Minister of Education and Culture, Samaria Tovela, announced this week that the country is in advanced stages of drafting binding AI rules for its entire education system, primary through university, with particular focus on generative AI. The framework is due before the Council of Ministers by year’s end.
The timing is unusual in a telling way. Mozambique’s broader National AI Strategy — in development with UNESCO support since December 2025 — hasn’t been adopted yet. A sector-specific rule is on track to reach Cabinet before the national framework it should theoretically sit beneath, meaning education policy will end up setting precedents the wider strategy later has to accommodate, rather than the other way around. The stakes are real: preliminary 2025 data shows distance education now accounts for 26.6% of Mozambique’s total higher education enrolment — over 75,000 students, learning in exactly the remote, hard-to-monitor settings where generative-AI use is hardest to see and easiest to reach for.
It’s also, as far as available reporting shows, the first time any African government has taken binding, national, cross-level AI education rules to Cabinet. The approach differs sharply from what AfriWave covered in Google’s expanded free-AI offer for students across 27 African countries: Egypt has issued voluntary teacher guidance; Ethiopia is folding AI into its curriculum without a teacher framework; South African universities have handled it individually, institution by institution, mostly by retreating from AI-detection software toward disclosure requirements after the tools were shown to misfire on second-language English writers. Mozambique is betting that a single national rulebook, done early, beats a patchwork built after the fact. Whether that bet pays off depends entirely on details not yet public — what disclosure looks like in a remote exam, how a university verifies authorship at a distance, and what happens to students who can’t access the same tools as their better-resourced peers.
The Throughline
Put side by side, these four stories aren’t really about AI as a technology — they’re about who gets to decide what problem it solves first. A founder in Lagos decided it should read a receipt. A founder in Kampala decided it should read a farm. A graduate in Pietermaritzburg decided it should read his parents’ hands. A ministry in Maputo decided someone needs to decide, before 75,000 remote students decide for themselves.
For AfriWave’s community, the throughline is a genuinely optimistic one, with a catch attached: the tools being built are solving problems diaspora families recognise instantly — unreliable banking infrastructure, farmers locked out of credit, a language barrier inside your own family, a classroom that no longer has walls. But in every case, the technology arrived before the institutions meant to govern, fund, or protect its use — and only one of this week’s four stories was authored by an institution at all.
Reader, which of these stories would most change something in your own family’s life back home — and is anyone building the rules fast enough to keep up with it?
Sources: TechCabal (“Day 1 to 1000,” Opeyemi Kareem, 19 September 2026); iAfrica.com; Business Tech Africa; IOL / The Post; Pulse Uganda; Disrupt Africa; TechMoran; ChimpReports.
