This n8n workflow demonstrates an approach to parsing bank statement PDFs with multimodal LLMs as an alternative to traditional OCR. This allows for much more accurate data extraction from the document especially when it comes to tables and complex layouts. Multimodal Parsing is better than traditiona OCR because: It reduces complexity and overhead by avoiding the need to preprocess the document into text format such as markdown before passing to the LLM. It handles non-standard PDF formats which may produce garbled output via traditional OCR text conversion. It's orders of magnitude cheaper than premium OCR models that still require post-processing cleanup and formatting. LLMs can format to any schema or language you desire! How it works You can use the example bank statement created specifically for this workflow here: https://drive.google.com/file/d/1wS9U7MQDthj57CvEcqG_Llkr-ek6RqGA/view?usp=sharing A PDF bank statement is imported via Google Drive. For this demo, I've created a mock bank statement which includes complex table layouts of 5 columns. Typically, OCR will be unable to align the columns correctly and mistake some deposits for withdrawals.
Tags
Compression
Edit Image
Google DriveGoogle Gemini Chat ModelRelated workflows
See all AI Automation→
Verify Japanese invoice registration numbers from Gmail using Gemini, NTA API, Sheets and Slack
Quick Overview This workflow monitors Gmail for unread invoice emails with PDF attachments, extracts invoice details with Google Gemini, verifies Japanese qualified invoice registration numbers against the National Tax Agency (NTA) Invoice Web-API, logs results to Google Sheets, and sends Slack alerts for invalid, missing, or inconclusive checks. How it works Polls Gmail hourly for unread emails matching “has:attachment invoice” and downloads the first attachment. Extracts text from the attached invoice PDF. Uses Google Gemini to parse the invoice text into structured fields such as the T-registration number, issuer name, invoice date, and amounts. Normalizes and validates the registration number to ensure it matches the required “T + 13 digits” format. If the format is valid, calls the NTA Invoice Web-API to verify the registration number and interprets the response as valid, invalid, or not found. Appends the invoice details and verification status to a Google Sheets worksheet. Routes the outcome by sending the appropriate Slack alert for invalid/not found/inconclusive results or marking the invoice email as read when the number is valid. Setup Connect your Gmail OAuth2 credentia
Check Japanese ad copy compliance with Google Gemini, Sheets, Slack and Gmail
Quick Overview This workflow runs daily to review Japanese ad copy in Google Sheets using Google Gemini, writes compliance risk results back to the sheet, and alerts the right team in Slack (and drafts a Gmail revision request) based on high/medium/low risk. How it works Runs every day at 9:00 on a schedule. Reads ad copy rows from Google Sheets to find items that still need a compliance check. Sends each copy item to Google Gemini with the configured product category, target laws, and extra rules to generate a risk level, violations, and a compliant rewrite. Parses and normalizes the Gemini output into fields like risk level, violation count, and safe rewrite, and posts a Slack alert if the AI check fails. Updates the corresponding Google Sheets row with the compliance results to keep a record of the review. Routes the item by risk level and posts to Slack for legal blocking (high), marketing review (medium), publish-ready notification (low), or manual review when the result is inconclusive. For high-risk copy, also creates a Gmail draft requesting the author to revise the copy using the suggested compliant rewrite. Setup Add credentials for Google Sheets, Google Gemini (PaLM) API
AI: Summarize podcast episode and enhance using Wikipedia
The workflow automates the process of creating a summarized and enriched podcast digest, which is then sent via email. Note that to use this template, you need to be on n8n version 1.19.4 or later.