AI-Powered Financial Inclusion for Kenya's Informal Economy

Transform spoken transactions into verified financial records using Amazon Nova AI, enabling informal traders to build credit histories and access formal banking services.

83% of Kenya's workforce (14M+ people) operates in the informal economy, conducting millions of shillings in daily transactions. Despite this economic activity, they lack formal financial records, making them "unbankable" and unable to access credit for business expansion.

Biashara Act uses multiple Amazon Nova capabilities:

1. 🤖 Agentic AI — Nova 2 Lite with tool-use parses transactions, validates numbers, categorizes the business, and assesses profitability in a multi-step agent loop
2. 📷 Multimodal Understanding — Nova 2 Lite vision parses photos of handwritten receipts and transaction records
3. 🖥️ UI Automation — Nova Act automates data entry into Wave Accounting for formal bookkeeping
4. 📊 Credit Scoring — Algorithm calculates loan readiness from transaction history

cd backend
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
# Edit .env with your AWS credentials
python app.py

cd frontend
npm install
npm run dev

Open http://localhost:3000

The parser uses the Converse API with toolConfig to run a multi-step agent loop:

User Input → Nova 2 Lite Agent
  ├── Tool: validate_transaction (fix quantity × price = total)
  ├── Tool: categorize_business (classify by items traded)
  └── Tool: assess_profitability (margin analysis + advice)
→ Validated transactions + business insights

Nova decides which tools to call and in what order. The agent loop continues until all tools have been used and a final summary is generated.

Upload a photo of a handwritten receipt → Nova extracts structured transaction data:

response = bedrock.converse(
    modelId='us.amazon.nova-2-lite-v1:0',
    messages=[{
        'role': 'user',
        'content': [
            {'image': {'format': 'jpeg', 'source': {'bytes': image_data}}},
            {'text': 'Extract transactions from this receipt...'}
        ]
    }]
)

Automates entry of transactions into Wave Accounting:

with NovaAct(starting_page="https://my.waveapps.com/login", headless=True) as nova:
    nova.act("Navigate to Transactions page")
    nova.act("Click 'Add Income' button")
    nova.act(f"Fill in description: Sale: pineapples x40")
    nova.act(f"Set amount to 2400")
    nova.act("Click Save")

biashara-act/
├── backend/
│   ├── app.py                  # Flask API
│   ├── nova_parser.py          # Agentic parser + multimodal (Nova 2 Lite)
│   ├── nova_act_bookkeeper.py  # UI automation (Nova Act)
│   ├── mock_data.py            # Mock transactions & scoring
│   └── requirements.txt
├── frontend/
│   ├── src/
│   │   ├── App.jsx             # Dashboard with text + image input
│   │   ├── main.jsx
│   │   └── index.css
│   ├── package.json
│   └── vite.config.js
└── README.md

• ✅ Agentic AI — Multi-step tool-use agent loop with validate → categorize → assess
• ✅ Multimodal Understanding — Receipt image parsing with Nova vision
• ✅ UI Automation — Nova Act automates Wave Accounting data entry
• ✅ Freestyle — Financial inclusion for 14M+ informal traders

Grades: A (80-100) Loan Ready · B (60-79) Almost There · C (40-59) Building History · D (0-39) Keep Recording

# Required
AWS_REGION=us-east-1
AWS_ACCESS_KEY_ID=...
AWS_SECRET_ACCESS_KEY=...

# Optional: Nova Act bookkeeping
NOVA_ACT_API_KEY=...
WAVE_EMAIL=...
WAVE_PASSWORD=...

Without NOVA_ACT_API_KEY, the bookkeeping endpoint returns simulated results for demo purposes.

Built for Amazon Nova AI Hackathon 2026