Kenyan Startup Fastagger Runs AI Offline for 6,000 Small Businesses

About 6,000 Kenyan businesses have signed up for Auni, a business-intelligence assistant that runs AI offline on low-cost phones — and it got there by plugging into M-Pesa, Kenya's largest mobile money service.

Auni is the first major product of Fastagger, a Nairobi startup founded by Mutembei Kariuki, Jude Mwenda and Stephanie Njerenga. It launched in 2024. The tool helps barbershops, restaurants and small manufacturers analyze their transaction data and get marketing recommendations. For a salon owner who used to track every sale on paper, that shift is concrete: the app now shows him his busiest hours and flags his loyal customers.

The stakes reach beyond one startup. Fastagger is part of a wider effort to build AI around the technological realities of developing economies — inexpensive phones, unreliable connections, and business records that may consist of little more than mobile-money transactions — rather than adapting products designed for wealthier markets.

Why did Kenya need a different kind of AI?

Kariuki started thinking about AI in his home country in 2019. He was at a conference on the future of AI at the University of Tokyo, and the talk was all transformation. "But nobody mentioned Africa at all," he says.

Even then, before OpenAI and Anthropic released their chatbots to the public, he could see the obstacles. Internet access in Africa was slow and unequal. Computing resources for training models were scarce. High-quality, structured training data was thin. And most people reached the internet through outdated, low-memory cellphones, not computers or high-end smartphones.

So Fastagger set out to build models that could run offline and on low-power devices. The founders began with the data problem. "How do we get Africa to catch up fast?" they asked themselves. "How do we create training data fast enough, and then how do we actually get it into people's hands fast enough?"

The answer started with mobile money. More than 80 percent of Kenya's adults use mobile money apps, according to the 2024 FinAccess Household Survey by Kenya's central bank. That makes the apps a rich data source — if you can read it.

How do you extract data from bad PDFs on cheap phones?

Fastagger's first technical move was adapting an open-source optical character recognition model to read financial statements from mobile money apps. Existing OCR models choked on the low-quality PDFs, which had unfamiliar formatting and sometimes contained stamps and errors.

Because users often had unreliable internet, the team made sure key processing steps — like organizing extracted data into tables — could run on the device itself. Standard OCR software pulls out the text, and a small neural network figures out how each document is structured.

Reading sales statements, analyzing data, and visualizing results all work offline on low-cost phones, without cloud queries. More detailed analyses and question-answering about the data still require an internet connection.

Before writing any features, the founders spent entire days watching small-business owners work. The constraint that emerged wasn't technical. Owners had almost no time to think beyond day-to-day operations.

"They are standing up the whole day cutting hair," Kariuki says of salon owners, now among Auni's main users. "You have to think about that dynamic of the user journey, and how they live, and what will be going on in the background."

What does Auni change for a shop owner?

Peter Chege owns a salon and barbershop in Nairobi. He started using Auni last year. Before that, he kept all his sales on paper. "We were not able to translate that information into a graph," he says.

Now a simple chart told him mornings were his slowest hours. So he runs errands early or keeps fewer barbers on shift until around 1 p.m., when business picks up. If a regular customer stops coming, he can send a personalized message encouraging them to return.

The automation piece matters as much as the analysis. At a fish restaurant in Nairobi, identifying loyal customers for discounts used to mean going through SMS sales receipts one by one — a tedious task that made customer-retention marketing impractical. With Auni, the restaurant identifies its top customers automatically and offers them discounts.

Is this a template for AI in developing economies?

Daniel Björkegren, an economist at Columbia University who studies AI's implications for developing economies, sees real potential in the approach. Sub-Saharan Africa's prevalence of basic phones over smartphones and computers, he argues, could open the door to different kinds of AI use.

"The history of technology should suggest that this is not going to be people running Claude Code or Codex on their laptop," he says. "You will have people like that, but for much of the population, we're going to have to figure out different usage modalities."

Fastagger isn't alone in tailoring AI to local constraints. Other examples include an app for rural schools without internet in Chile, a medical triage chatbot for rural patients in India, and translation models for 19 African languages that run on the basic phones people already own. Innovations born of constraint could help democratize access to AI — though the constraints themselves don't disappear, and they still limit development.

Despite the M-Pesa integration, Fastagger remains a three-person company: Kariuki, Mwenda and Njerenga. It has received support from Google, Microsoft, and NVIDIA accelerator programs. The team is now building agents that could automatically implement marketing actions for business owners, moving from analysis to action.

The company has also branched out. It built a workforce management tool for Kenya's security companies, which employ hundreds of thousands of people. Security guards often work in remote locations without reliable internet and carry low-end phones. Fastagger's AI agent communicates with them via SMS to track routes, detect anomalies, and generate actionable alerts for operations teams.

Could small models for poor markets become a two-tier system?

Björkegren flags one risk in the trend: a widening split between small models designed for emerging economies and ever-larger models for wealthier regions. He's enthusiastic about systems built for populations poorly served by Silicon Valley, but wary of a hard division.

"I am worried that that could lead to a bifurcation," he says. "We need to enable a large menu of different options and really make sure that we have access and then see what works."

For now, Fastagger's bet is that the menu itself is the product. Auni's next step — autonomous marketing agents working over SMS on cheap phones — will test whether constraint-driven AI can graduate from dashboards to doing the work.