01
Morning
Phone in the apron
She keeps her phone in her apron pocket, or pairs a Bluetooth headset or earpiece.
Voice Ledger Ghana
Your sales, recorded while you sell.
Voice Ledger listens to the conversations at your stall and keeps your books for you, in Twi, Ga, Ewe, and English.
A trader in Accra, phone in her apron, serving a customer
The problem
Millions of traders keep their books in their heads.
Across Ghana's markets, most traders, most of them women, track sales from memory or in exercise books. At the end of a busy day nobody knows exactly what sold, what is owed, or whether the day made money.
Without records, traders are shut out of credit and can't see how their business is really doing.
How it works
She sells as she always has. We keep the books.
02
All day
It listens through the day
Voice Ledger picks up her conversations with customers as she sells, in the language she already speaks, including mixed Twi and English and local units like "olonka" and "bag".
03
Every sale
Every sale captured
- Item
- Quantity
- Price
- Buyer or debtor
The built-in intelligence recognises each sale and writes it down, including who bought it or who owes.
04
Anytime
Check anytime
She opens the app whenever she likes to see today's total, confirm each sale was recorded right, and correct anything.
05
Closing time
End-of-day summary
A plain-language summary in her own language: total sales, best sellers, money owed, and spending.
What makes it different
Built for the market, not the office.
Hands-free
No typing, no buttons while serving customers.
Speaks Ghanaian languages
Twi, Ga, Ewe, and English, including code-switching.
Offline first
Works in the market without data. On-device speech and speaker identification (TensorFlow Lite), with encrypted local storage.
The trader is in control
She reviews and edits every entry. Recordings stay on her phone.
Export when needed
CSV, PDF, and Excel records she can show a bank, susu group, or microfinance lender.
The intelligence
Built with Claude, to understand how markets really talk.
Claude turns messy, code-switched market conversations into structured sales records.
It writes daily and weekly business insights in the trader's own language, and powers a simple assistant she can just ask.
Ask the assistant
"How much did I make this week?"
"Who still owes me?"
Built on
An Android app that does the heavy lifting on the phone.
- App
- Kotlin · Jetpack Compose
- On device
- Speech and speaker identification
- Understanding and summaries
- Claude
- Pilot Planned
- Kumasi and Accra markets
Help us collect market voices
We're gathering real market conversations, recorded with everyone's agreement, to teach Voice Ledger how Ghanaian trading really sounds. Have a phone and a market nearby? Record a sale in under a minute.
Founder
Christian Dzidula Dotsey
Founder · Computer Engineering, KNUST
Christian's background is in machine learning for Ghanaian languages, including Twi and Pidgin speech, and in embedded systems.
Join the pilot