AI is changing what software can do for a business. We are putting it to work on Tanzanian problems first: in Swahili, inside the products people already use, and always with a person in charge of the decision.
How we use AI
Useful, not novel.
Swahili first
Assistants that read, write and speak Swahili as well as English, because that is the language most Tanzanian businesses actually run in.
Inside the workflow
No chatbot bolted to the side. AI sits where the work already happens: the stock count, the loan review, the member query on WhatsApp.
A person signs off
Models draft, flag and forecast. People approve. Anything that moves money, medicine or a member balance stays a human decision.
Built for modest hardware
The intelligence runs in our cloud, not on the device, so a five-year-old Android phone gets the same help as a new laptop.
Building now
Extending products that are already live with real customers.
Ecosystem
AI across the live products
Our customers already have the data: every sale in DawaOS, every repayment in OnLoan, every contribution in CommunityOS. Most of it is only ever looked at in hindsight.
What we’ll build
Demand and expiry forecasts in DawaOS, early repayment-risk signals in OnLoan, and a Swahili assistant in CommunityOS that answers members’ balance and loan questions on WhatsApp.
Where AI helps
Forecasting on each customer’s own history, and language models that answer from the customer’s records, never from guesswork.
Why here, why now
The fastest way to make AI useful in Tanzania is to put it inside software people already open every day.
Up next
Committed work. Discovery is done and design has started.
Customer service
Swahili business agents
Shops, clinics, schools and lenders answer the same questions all day on WhatsApp and the phone: prices, opening hours, balances, appointments. It eats the time of the few staff they have.
What we’ll build
A WhatsApp and voice agent any business can switch on, trained on its own prices, policies and records, that answers in Swahili or English and hands over to a person when it should.
Where AI helps
Conversational models tuned for Swahili, connected to the business’s live data through the Afrevo products.
Why here, why now
WhatsApp is already how Tanzania does business. The missing piece is software that can keep up with the volume, in the right language.
Health
Clinics & insurance claims
Dispensaries and health centres still keep patient records on paper, and insurance claims are often rejected for missing codes or mismatched documents, money a small facility cannot afford to lose. The move towards universal health insurance will only raise the volume.
What we’ll build
A records and billing system for small clinics that grows out of DawaOS, with claims prepared, checked and submitted from the same place the patient was seen.
Where AI helps
A claim checker that catches errors before submission, and note-taking help that turns a consultation into a structured record.
Why here, why now
We already sit in the pharmacy. The clinic next door has the same problems, and a far bigger paperwork burden.
Commerce
Tax-ready retail
Every registered business must issue TRA-compliant fiscal receipts, and most also take mobile money. Owners end up running a POS, a separate receipt device and a phone, then reconciling all three by hand.
What we’ll build
One point of sale for everyday shops that issues compliant receipts, reconciles M-Pesa, Airtel Money and other wallets automatically, and closes the day in one view.
Where AI helps
Automatic matching of mobile-money payments to sales, and a plain-language daily summary of what the business actually made.
Why here, why now
Compliance is a cost every business carries. Software that removes it, rather than adds to it, sells itself.
Exploring
Problems we are studying in the field. Not every one will ship.
Agriculture
Farmer cooperatives
Agriculture employs most working Tanzanians, and crops like cashew, coffee and cotton move through cooperatives (AMCOS) that still weigh, record and pay farmers on paper. Delayed or disputed payments are the norm.
What we’ll build
A cooperative system for farmer registration, crop collection, weights and grades, input credit and mobile-money payouts, so every farmer can see what they delivered and what they are owed.
Where AI helps
Voice advice in Swahili for farmers on basic phones, and yield and price signals that help a cooperative plan collection and sales.
Why here, why now
It is the same trust problem CommunityOS solves for savings groups, at far larger scale and with far more riding on it.
Education
School operations
Thousands of private and community schools track fees in exercise books, chase parents by phone, and compile results by hand every term.
What we’ll build
Fees collected through control numbers and mobile money, results and report cards generated automatically, and parents kept informed by SMS and WhatsApp.
Where AI helps
Early flags for students falling behind, and report-card comments drafted for teachers to review and edit.
Why here, why now
Parents in Tanzania invest heavily in schooling. The schools themselves have almost no software built for them.
Finance
Credit for small businesses
Small businesses struggle to borrow because they have no formal credit history, even when their sales records show they are a good risk.
What we’ll build
With the owner’s consent, sales and repayment history from Afrevo products becomes evidence that licensed lenders using OnLoan can assess fairly.
Where AI helps
Transparent scoring models a loan officer can read and question, built within Bank of Tanzania rules for responsible digital lending.
Why here, why now
We hold the operating data on one side and the lending software on the other. Connecting them, carefully, opens credit to people the system has missed.
Infrastructure
Afrevo Cloud
Tanzania’s Personal Data Protection Act sets clear duties for how personal data is handled, yet much of the country’s business data lives on servers an ocean away.
What we’ll build
Hosting, storage, backups and identity for Afrevo products and for teams that want their data, and the AI that works on it, kept close to home.
Where AI helps
Models served from local infrastructure, so sensitive health, finance and member data never has to leave the region to be useful.
Why here, why now
As AI grows, where data lives and who processes it will matter more, not less.
Names are working titles. Plans change as we learn in the field.
Working on one of these problems?
We build best alongside the people who live with the problem. If you run a clinic, a cooperative, a school or a lender, or want to back this work, we'd like to hear from you.