The problem
A landlord with ten rooms, or a caretaker running three hundred units, faces the same question every month: who has paid, and who owes? Tenants pay by M-Pesa, sometimes to the wrong account number, sometimes in parts. Records live in SMS inboxes, notebooks, and WhatsApp chats. Arrears grow quietly, and disputes follow.
Context
- Most rent in Kenya is paid by mobile money, but most small landlords still reconcile it by hand.
- The people doing the work are often caretakers on a phone, sometimes on a weak connection, and some tenants only have a feature phone.
- It started close to home. The apartment building where our founder lives still runs on traditional records, with no automatic messages about balances or utilities. Work gets repeated by hand, and everything is slow.
Approach
We designed it around the phone and the payment, not around a desktop dashboard.
- Make the payment do the work. Tenants pay to a Paybill, and each payment is matched to the right tenant and invoice automatically. Hand-typed transaction codes are checked against M-Pesa before they're accepted.
- Reach everyone. Tenants use an installable web app, or USSD on any phone. That includes paying.
- Work where the signal is weak. Meter readings and repair reports can be captured offline and synced later.
- Keep it safe. Each organisation's data is strictly separated. Roles limit who can do what. Every action is logged.
The solution
- Properties, units, tenants, and leases, with automatic monthly invoices and recurring charges
- M-Pesa Paybill matching and receipts
- Arrears follow-up by SMS, WhatsApp, and push. Quiet hours and tenant consent are built in.
- Electricity and water meters, with offline reading rounds
- Repairs, expenses with approval, letters, documents, and inspections
- Tenant portal (web app) and USSD
- Reports: expected vs collected, arrears, vacancies, profit and loss, cash flow, and owner statements
- An AI assistant for staff that always asks for confirmation before acting, and answers to tenant questions over WhatsApp
- An API with webhooks, for agencies that need to connect other systems
Technology
Built with: Python · Django · PostgreSQL · Redis · M-Pesa (Daraja) · Africa's Talking (SMS, USSD) · Anthropic Claude · PWA
Engineering quality:
- automated tests across the system (116 test files), with continuous integration;
- load-tested with Locust;
- backup restores tested;
- a security review covering content security policy, MFA, and access checks on every route.
Result
No user results yet. LandlordPMS hasn't been used by a real landlord.
What we can say is that it's fully built and tested, and ready for a pilot. The pilot will measure:
- how many payments are matched automatically;
- how much time landlords save on reconciliation each month;
- whether arrears go down;
- whether pilot landlords are still using it after three months.
What we learned
We built before we interviewed.
We put our energy into building the product before spending enough time validating the workflows with landlords. The system is complete and well tested, but the most important question is still open: which of these features do landlords actually need, and in what order?
That changed how MLLabz works. We now understand the problem and the people first, then build. For LandlordPMS, that means interviewing landlords and caretakers before the pilot, and letting what they say decide what the pilot tests.
- Also learned: mobile-money reconciliation, offline sync on weak connections, keeping each organisation's data strictly apart, and making AI actions safe all turned out to be engineering problems, not features you can simply add on.
What's next
- Interview landlords and caretakers first, then pilot with five landlords, once hosting and M-Pesa go-live are done.
- Then intelligence, once real data exists:
- early warning on tenants likely to fall behind;
- smarter payment matching;
- meter anomaly and leak detection;
- rent and vacancy insight;
- a Swahili voice assistant.