n8n Expense OCR Workflow: Mindee + GPT-4o, Save to Postgres

Turn receipt uploads into categorized expenses—automatically

This n8n Expense OCR Workflow takes receipt submissions from an n8n form, extracts expense data using Mindee, classifies and validates it with OpenAI, stores the receipt in Google Drive, saves records to PostgreSQL, and sends Slack alerts—plus a daily analytics summary.

What this workflow does

• Receives an expense submission via an n8n form with a receipt file and optional amount/notes.
• Validates the upload, then uploads the receipt to Google Drive and enqueues it for Mindee OCR.
• Polls Mindee for OCR results and extracts key fields such as merchant, amount, date, category, receipt number, and raw text.
• Handles OCR failures and incomplete data by sending Slack alerts for manual review.
• Uses OpenAI to classify the expense (final category, business vs personal, expense type, and confidence), and blocks low-confidence results.
• Prevents duplicates in PostgreSQL by checking receipt number and amount before inserting.
• Saves the finalized expense record to PostgreSQL, including the Google Drive receipt URL, and posts a success message.
• Runs daily analytics by querying PostgreSQL spending totals and posting a dashboard-style summary to Slack.

Use cases

• Automate expense capture for SaaS ops and bookkeeping teams using receipt uploads.
• Speed up review workflows by flagging OCR or low-confidence classification in Slack.
• Maintain an auditable history by linking each expense record to its stored receipt in Google Drive.

Technical details

• Integrations/Credentials: Mindee (HTTP header auth), OpenAI, Slack, Google Drive OAuth, PostgreSQL.
• n8n nodes used: if, code, wait, merge, slack, postgres.