AI Bank Statement Parser & Data Extractor using Google Gemini

Unlock Automated Financial Data Extraction with this AI Agent

This AI Agent automates the challenging task of extracting structured information from bank statement PDFs, including those that are scanned and not text-selectable. It acts as a sophisticated 'document understanding' specialist for your financial records.

First, it retrieves your bank statement PDF (e.g., from Google Drive). Then, it intelligently converts each page of the PDF into an image using the Stirling PDF service (you can use the public service or self-host for enhanced privacy). These images are then fed, one by one, to Google Gemini's powerful vision-language model. Gemini 'reads' each page image and transcribes its content into clean, well-structured markdown, adeptly handling tables and complex layouts.

Once the entire statement is converted to markdown, the agent employs another Google Gemini instance, this time focused on 'information extraction'. It parses the markdown to pinpoint and pull out specific data you need, such as all deposit transactions, formatting them into a structured JSON output. This significantly cuts down manual data entry and makes financial data readily available for analysis, reporting, or integration with other systems.

Key Features & Benefits

• AI-Powered Document Transcription: Employs Google Gemini's vision capabilities to accurately convert PDF pages (including scans) into markdown, preserving tables and text.
• Handles Scanned & Complex PDFs: Overcomes limitations of traditional OCR, making it ideal for image-based or poorly formatted bank statements.
• Structured Data Extraction: Uses AI to intelligently identify and extract specific financial details (e.g., deposits, withdrawals, dates, descriptions) based on your defined schema.
• End-to-End Automation: From PDF retrieval to structured data output, the entire process is automated.
• Flexible PDF Source: Easily adaptable to fetch PDFs from various sources, not just Google Drive.
• Customizable Prompts & Schema: Tailor the AI's transcription and extraction behavior to your precise requirements and statement formats.
• Supports Data Privacy: Offers guidance on using a self-hosted PDF-to-image converter (Stirling PDF) for sensitive documents.

• An n8n instance (Cloud or self-hosted).
• Google Gemini API Key (for models like gemini-1.5-pro-latest) and a configured 'Google Gemini (PaLM) API' credential in your n8n instance.
• Credentials for your Google Drive account, configured in n8n, to fetch statement PDFs.
• Access to a Stirling PDF instance. The workflow defaults to http://stirling-pdf:8080 (for a local/Docker setup). You can modify this to use the public stirlingpdf.io API or your own self-hosted instance URL.

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Configure the 'Get Bank Statement' (Google Drive) node with your Google Drive credentials and specify the PDF file ID or how your workflow will locate it.
4. In both 'Google Gemini Chat Model' nodes (connected to 'Transcribe to Markdown' and 'Extract All Deposit Table Rows'), select or create your 'Google Gemini (PaLM) API' credential using your API Key.
5. Verify the URL in the 'Split PDF into Images' (HTTP Request) node. The default is http://stirling-pdf:8080/api/v1/convert/pdf/img. If you're not using a local Stirling PDF instance at this address, update it to your Stirling PDF service endpoint (e.g., public API or your self-hosted URL).
6. Review the prompts within the 'Transcribe to Markdown' (LLM Chain) node to ensure the transcription instructions meet your needs.
7. Customize the system prompt and input schema in the 'Extract All Deposit Table Rows' (Information Extractor) node to match the specific details you want to extract and your desired output structure.
8. Test the workflow with a sample bank statement PDF by clicking 'Test workflow'.
9. Activate the workflow for ongoing use or connect it to your preferred trigger.