Fastagger's Auni brings offline-capable AI to Kenyan small businesses

IEEE Spectrum profiles Fastagger, a Kenyan startup that builds AI models able to run offline and on low-power cellphones. Cofounder Mutembei Kariuki, who is from Kenya, first thought about how AI could work there in 2019, at a conference on the future of AI at the University of Tokyo. Speakers kept saying the technology would transform the world, "But nobody mentioned Africa at all," he says. Even then he saw obstacles: slow and unequal internet, scarce computing resources for training, little high-quality structured data, and most people reaching the internet on outdated, low-memory phones.

With cofounders Jude Mwenda and Stephanie Njerenga, Kariuki created Fastagger. Its first big product, Auni, is a business-intelligence assistant that launched in 2024 and got a big push when it was integrated into the business app of M-Pesa, Kenya's largest mobile money service. Since then, about 6,000 businesses have signed up. Auni helps small and medium businesses such as barbershops, restaurants and small manufacturers analyze their transaction data, and it gives them marketing recommendations.

The founders began with data. They chose to focus on mobile phones, and specifically on mobile money apps, which more than 80 percent of Kenya's adults use and which therefore offer a rich source of data. One of their first steps was to adapt an open-source optical character recognition (OCR) model to read the financial statements those apps produce; existing models struggled with the low quality of the PDFs, which had unfamiliar formatting and sometimes contained stamps and errors. Because prospective users often had unreliable connections, key steps, such as organizing the extracted data into tables, were built to run on the device. The team then spent entire days watching small-business owners work. The lesson: owners have little time or resources to think beyond daily operations. "They are standing up the whole day cutting hair," Kariuki says of salon owners, who are now among Auni's main users.

Peter Chege, who owns a salon and barbershop in Nairobi, started using Auni last year. He used to record all his sales on paper: "We were not able to translate that information into a graph." Now the app shows him his busiest days and hours and his loyal customers, and if a regular stops coming he can send a personalized message to bring them back. A simple chart showed that mornings were his slowest hours, so he can run errands then or keep fewer barbers in the shop until around 1 p.m., when business picks up. At a fish restaurant in Nairobi, finding loyal customers to offer discounts once meant going through SMS sales receipts one by one, which made client-retention marketing impractical. With Auni, the restaurant now identifies its top customers automatically and offers them discounts.

Some Auni functions, such as reading sales statements, analyzing data and visualizing results, can run offline on low-cost phones without cloud queries. They mix AI and conventional software: standard OCR software extracts the text and a small neural network identifies how each document is structured. More detailed analyses and answering questions about the data need an internet connection.

Fastagger is still small. It has received support from Google, Microsoft and Nvidia accelerator programs but has just three employees, the original founders. They are now working on adding agents to Auni that could automatically carry out marketing actions for business owners. The company has also built a workforce management tool for Kenya's security companies, which employ hundreds of thousands of people. Guards often work in remote places without reliable internet and use low-end phones, so Fastagger's AI agent talks to them by SMS to track their routes, detects anomalies and generates alerts for operations teams.

The article places Fastagger in a wider effort to build AI around the realities of users in developing economies. Other examples it lists: an app for rural schools without internet in Chile, a medical triage chatbot for rural patients in India, and translation models for African languages that run on basic phones. Such innovations could help democratize access to AI, the article says, but the constraints that produce them do not disappear or stop limiting development.

Daniel Björkegren, an economist at Columbia University who studies AI's implications for developing economies, says the prevalence of basic phones in sub-Saharan Africa 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 sees great potential in systems built for populations poorly served by Silicon Valley, but fears too clear a split between small models for emerging economies and increasingly large, powerful models for wealthier regions. "I am worried that that could lead to a bifurcation," he says.

Key facts

Fastagger, a three-person Kenyan startup, builds AI models that run offline on low-power phones; its Auni business-intelligence assistant launched in 2024.
Auni is integrated into the business app of M-Pesa, Kenya's largest mobile money service, and about 6,000 businesses have signed up.
Fastagger adapted an open-source OCR model to read mobile-money statements; reading, analysis and visualization can run offline, while detailed analyses and questions about the data need internet.
Reported benefits are anecdotal: a Nairobi barbershop owner learned his slowest hours, and a Nairobi fish restaurant now finds its top customers automatically.
Economist Daniel Björkegren sees potential in such systems but worries about a bifurcation between small models for emerging economies and large ones for wealthier regions.

Why it matters

Most AI products assume fast internet, cloud compute and capable devices. Fastagger starts from the opposite end: inexpensive phones, unreliable connections, and business records that may be little more than mobile-money transactions. Kariuki says the idea began in 2019, when a Tokyo AI conference never mentioned Africa. The article presents Fastagger as part of a broader push to build AI around the conditions of users in developing economies, with more than 80 percent of Kenya's adults on mobile money apps as the data foundation.

Who it affects

Primarily small and medium businesses in Kenya, such as barbershops, restaurants and small manufacturers, with salon owners now among Auni's main users. Also Kenya's security companies, which employ hundreds of thousands of people and whose guards work with low-end phones in remote places. More broadly, the article points to builders of AI for emerging economies, and Björkegren's concern touches how those markets are served relative to wealthier regions.

How to use it

Auni is reached through the business app of M-Pesa, where it analyzes a business's transaction data and gives marketing recommendations. In practice, owners use it to see their busiest days and hours, spot loyal customers, and message regulars who stop coming. Reading statements, analyzing data and visualizing results can work offline on low-cost phones; detailed analyses and questions about the data need a connection. No revenue, pricing or business model for Auni is stated.

How solid is it

This is a single feature article in IEEE Spectrum by Lucila Pinto, a journalist from Argentina who writes about AI, trust in science and economic development. The main evidence is the founders' account, one salon owner and one restaurant example. The 6,000 figure is given as "about" and its source (Fastagger or M-Pesa) is not named. No measured results such as accuracy or revenue uplift are given, and benefits are anecdotal.

Risks and caveats

Fastagger has three employees and, despite the M-Pesa integration, remains a small startup supported by Google, Microsoft and Nvidia accelerator programs. Offline capability is partial: deeper analysis and question answering need internet. The planned agents that would implement marketing actions automatically are still in development, with no timeline given. The article notes that constraint-driven innovations do not make the constraints themselves disappear, and Björkegren warns of a possible split between small models for emerging economies and large, powerful models for wealthier regions.

“We need to enable a large menu of different options and really make sure that we have access and then see what works.”

— Daniel Björkegren, economist at Columbia University