NAIROBI – Business is booming at Mandevu, a men’s beard products manufacturer, and at Master Stylists Hair Salon and Barbershop, both in Nairobi.

One reason is the explosion of interest in personal grooming among men of Kenya’s fast-growing middle class. Another reason is Auni, an app that uses AI, powered by Microsoft, to deliver business intelligence.

“I like living my life better and simpler,” says Njoki Njoroge, founder and CEO of Mandevu Beard Care, in the bustling district of Thika Road Mall in Nairobi. “I have an app for everything. If I see simple tools that will make my work life better, especially when I can see there’s clarity and I can make better decisions, I’m there, sign me up.”

Mandevu (a portmanteau of “man” and “ndevu,” or “beard” in Swahili) began as a digital-first enterprise selling beard products online. Njoki, and later her staff, kept detailed spreadsheets, but she didn’t have the digital tools or IT staff to glean real insight.

The Auni app uses AI to turn Njoki’s mobile money PDF statements into data that can be analyzed. Auni, which means “help” in Swahili, runs on Microsoft Azure and uses its integrated tools, such as optical character recognition, as well as Microsoft’s Phi-3 and Phi-4 small language models and open-source tools on GitHub such as Table Transformer, ONNX and BitNet.cpp that mostly work on the edge device.

The Auni app, created by a company called Fastagger, is available on Safaricom, Kenya’s largest mobile phone and technology company. Safaricom is also the market leader in money transfers, which is how many Kenyans do transactions of every size.

Kenya is widely recognized as a global leader in mobile money. According to the Global System by Mobile Communications Association (GSMA), the country has set the standard for mobile financial services, while a World Bank report calls the region a pioneer in this space. Traditionally – and still today – few Kenyans had bank accounts. That used to mean time-consuming trips to deliver cash – whether to siblings needing to pay school fees or to a business partner across the country; a country that is a bit bigger than France and nearly the size of Texas.

In 2007, Safaricom was testing a mobile money lending program but pivoted to transfers, creating M-Pesa (“pesa” means “money” in Swahili, and the M is for mobile). Today, M-Pesa has 91 percent of the market. Even pushcarts selling bananas or tomatoes have hand-written cardboard signs giving the vendors’ M-Pesa information for payment.

Because of the prevalence of older phone models, mobile apps in Kenya need to be barebones. M-Pesa statements are in PDF form. While they contain a wealth of information – customer name, address and phone number, amount, date and time of sale, as well as the same information for payments to suppliers – it isn’t in a usable format without costly, time-consuming data entry. That’s where Auni comes in.

Auni uses a kind of AI called optical character recognition to extract the data. It then uses another AI tool on GitHub called Table Transformer to organize the data into a table for easy analysis. The company plans to add generative AI to enable business owners to ask questions in natural language, no IT or coding knowledge necessary.

Auni is deployed as a mini app within Safaricom’s M-Pesa business super app, which has more than a million downloads. In Auni’s first three months as part of the M-Pesa super app, 3,500 businesses signed up.

“The overall idea was how can we democratize access to AI and the benefits of AI?” says Mutembi Kariuki, co-founder and CEO of Fastagger, the Nairobi startup behind the Auni app. The highest impact is in micro, small and medium enterprises, but “a lot of places, especially in emerging markets, have challenges with connectivity and the kind of devices that they have. We realized that you could actually run AI models on their four-gigabyte smartphones and enable them to have the AI equivalent of an MBA intern on the phone that can analyze their mobile money transactions.”

Kariuki and co-founder Jude Mwenda had founded other startups, but they realized that in addition to a lack of connectivity and the latest technology, Africa also lacked data. So, together with another data scientist, Stephanie Njerenga, in 2019 they founded Fastagger, to speed up the creation of data sets for AI by hiring data experts to manually tag data – hence the name, combining “fast” and “tagger.”

During the Covid-19 pandemic, a Kenyan business association asked the Fastagger founders to conduct webinars to help struggling businesses stay afloat. That opened their eyes to how owners of businesses with just a few employees did most of their transactions on their phones, and sometimes not even smartphones, without internet service all the time and “on top of that, it’s expensive,” Kariuki says.

“We realized we had to build something that even if it was sitting on their own personal phone, then so that we can read the SMS for the mobile money transactions and be able to give them that insight, it would have to work offline and on device. When we say ‘on device’ we mean that you can literally use AI without having to have internet,” he says.

“Our bet has been that the African market doesn’t have money for GPU compute,” says Mwenda, who is Fastagger’s chief technology officer. “So, we compress the model, make it smaller and optimize for low-cost constraints.”

Distributed AI was the focus of Mwenda’s Ph.D. at Georgia Tech University (he got his master’s at the Massachusetts Institute of Technology), “so that worked out perfectly for us.” Fastagger pivoted away from tagging data to the Auni app, which was embedded into the M-Pesa super app. It isn’t just for small businesses – Fastagger is adapting Auni to other sectors, like health care, manufacturing and agriculture.

“We never sold this as AI to customers,” Kariuki says. “Just, ‘hey, you can retain and increase customers.’”

Master Stylist hair salon owner Peter Chege is among those who have benefited from the analytical capacity of Auni. Before Auni, he kept basic accounting books.

“We write down the name of the service and how much the client paid,” he says. “But unless you actually go through the books again, those details aren’t accessible and it’s easy to forget about the data.”

With Auni, Chege could easily see how frequently each customer came and how much they spent, grouping what had previously been seen as separate transactions where the husband would come for a trim, the wife for braids and nails and the children would regularly get haircuts.

Chege can send discount coupons to regulars who haven’t come in for a while. When Chege sees a trend of slow days, he puts more effort into social media to attract customers. Social media is especially important because the salon is in a charming but quiet shopping center without much foot traffic. Business is booming – he recently renovated and expanded the salon.

“Big corporations pay people to understand how the market is behaving,” Chege says. “With Auni, we get that for far less.” And, he adds, the app is easy to use, and the data is presented in an easy-to-understand format.

Studies have shown that a five percent increase in customer retention can increase profitability by 25 percent or more. “Because small and medium enterprises are 70 percent of employment in Africa and account for 90 percent of the businesses, if you can increase their profitability then you’re helping to secure jobs, you’re helping to increase income in families. That creates the multiplier effects,” Kariuki says.

For Njoki of Mandevu, keeping track of repeat customers is important because different products last different amounts of time.

“Beard oil, depending on size of the beard, will last two to three months,” she says. “Toner will not end at the same time as your cream. So, if we send a reminder, we get really good customers because it’s not a blind text we’re sending. It’s targeted. Of course the product is running out because it’s time. We’re able to know if they have a goatee and are bald – we aren’t going to suggest shampoo.”

Njoki uses Auni to map customer locations and, armed with that data, negotiates with retailers in key neighborhoods to carry Mandevu’s products. Having about 70 retail outlets also means easier pickup and faster deliveries in sprawling Nairobi’s dense and chaotic traffic.

“One report tells you hourly revenues and the number of transactions,” Njoki says. “Noon Thursday is peak. So we need more delivery guys, because no one is going to wait for five hours or a day for a delivery. People will send us their riders. Or we know that the stores around a certain area have to restock by end of week so they don’t run out.”

As an e-commerce business, Mandevu gets orders online, but also from calls, emails and texts. Whereas those different streams previously were separate, Auni has united them because nearly everyone pays with M-Pesa.

“My background is in marketing and communications, so I always wanted to record everything. From day one I could tell you who ordered, even what time,” she says. But success has made it impossible to continue to enter data by hand daily. “In the beginning, as founder you do everything. Over time, you’re able to have someone to run logistics, a merchandiser. I’m able to make clear decisions because I have the data. You’re moving from guessing to knowing.”