Gemini + Data Table Procurement Validation n8n Workflow

Stop bad procurement data at the door—instantly validate, detect recurring errors, and use Google Gemini for anomaly checks

The Gemini + Data Table Procurement Validation n8n Workflow receives procurement line-item records via a webhook, runs deterministic validation (required fields, math, tax, and unit normalization), then—only when needed—uses Google Gemini to spot contextual semantic anomalies. It returns an immediate JSON accept/reject response and learns from new error patterns by upserting them into an n8n Data Table.

What this workflow does

• Ingests procurement data through an HTTP POST webhook.
• Normalizes and validates fields using configurable rules, including unit aliases and deterministic checks for:
  • Required field presence
  • Quantity and price validity
  • Subtotal math
  • Tax calculations
• Builds an error “pattern signature” from key fields and searches an n8n Data Table for prior occurrences.
• Reuses trusted corrections when a high-confidence recurring pattern (with enough occurrences) is found—skipping AI review for efficiency.
• Runs Google Gemini when no trusted pattern exists to identify contextual anomalies (e.g., category, coding, supplier-item, or unit inconsistencies) and suggested corrections.
• Combines deterministic results + history + Gemini findings into a single decision, returning JSON that indicates whether the record is accepted and what issues were detected.
• Improves future detection by upserting newly observed error patterns (counts, confidence, and summary) into the n8n Data Table.

Use cases

• Validate procurement line items before they enter ERP/finance workflows.
• Auto-detect repeated vendor or unit coding mistakes across submissions.
• Catch subtle semantic mismatches that deterministic rules can miss.

Technical details

• n8n nodes: if, set, code, merge, webhook, data table
• AI integration: Google Gemini for semantic anomaly detection
• Data learning: n8n Data Table for recurring pattern matching and upsert

Setup note: create an n8n Data Table to store and reuse procurement error pattern signatures.