Reconcile Two CSV Exports with Gemini Drift Diagnosis

Reconcile two CSV exports and pinpoint Gemini drift causes in one n8n run

This n8n workflow collects two CSV exports, matches them by a user-defined key and amount column, and—when differences appear—uses Google Gemini to diagnose the most likely systemic reasons for “drift,” returning either a balance confirmation or a downloadable Excel reconciliation workbook.

What this workflow does

• Accepts an n8n form submission where you describe the reconciliation, choose the linking key and amount columns, and upload two CSV files.
• Extracts and structures both CSVs into comparable rows.
• Compares datasets by the key column to identify:
  • agreements
  • values that disagree
  • records only in the “source of truth”
  • records only in “your records”
• Calculates reconciliation metrics including counts, totals, and the net gap from the grouped differences.
• Diagnoses drift with Google Gemini using the reconciliation summary and representative examples, then parses the result into a structured diagnosis with likely causes, checks, and materiality.
• Returns results:
  • If everything balances, it shows a completion page confirming there are no differences.
  • If differences exist, it generates an Excel workbook starting with the diagnosis and listing every differing row by bucket, provided as a download from the form.

Use cases

• Monthly reconciliation between two finance systems exported as CSV (bank vs. ledger, ERP vs. accounting).
• SaaS operator drift checks after migrations, fee changes, or sync updates.
• Automation engineers validating data consistency across two independently exported reports.

Technical details

• Google Gemini (PaLM) API credential for drift diagnosis.
• Uses n8n nodes including form, form trigger, if, set, code, and supporting notes via sticky note.

Ideal for n8n automation: turn messy CSV reconciliation into a structured, AI-assisted drift diagnosis workflow with an Excel reconciliation deliverable.