How it works

This workflow automatically processes bank statements from various formats and extracts structured transaction data with intelligent categorization using AI.

Key Steps

1. File Upload - Accepts bank statements via webhook upload (PDF, Excel, CSV formats).
2. Smart Format Detection - Automatically routes files to appropriate processors (PDF text extraction or spreadsheet parsing).
3. AI-Powered Extraction - Uses GPT-4 to extract account details, transactions, and balances from statement data.
4. Data Processing & Categorization - Cleans, validates, and automatically categorizes transactions into expense categories.
5. Database Storage - Saves processed data to PostgreSQL database for analysis and reporting.
6. API Response - Returns structured summary with transaction counts, expense totals, and category breakdowns.

Setup steps

Setup time: 8-12 minutes

1. Configure OpenAI credentials - Add your OpenAI API key for AI-powered data extraction.
2. Set up PostgreSQL database - Connect your PostgreSQL database and create the required table structure.
3. Configure webhook endpoint - The workflow provides a /upload-statement endpoint for file uploads.
4. Customize transaction categories - Modify the AI ​​prompt to include your preferred expense categories.
5. Test the workflow - Upload a sample bank statement to verify the extraction and categorization process.
6. Set up database table - Ensure your PostgreSQL database has a bank_statements table with appropriate columns.

Features

• Multi-format support : PDF, Excel, CSV bank statements
• AI-powered extraction : GPT-4 extracts account details and transactions
• Automatic categorization : Expenses categorized as groceries, dining, gas, shopping, utilities, healthcare, entertainment, income, fees, or other
• Data validation : Cleans and validates transaction data with error handling
• Database storage : PostgreSQL integration for data persistence
• API responses : Clean JSON responses with transaction summaries and category breakdowns
• Smart routing : Automatic format detection and appropriate processing paths