Autonomous Automotive Sales Intake System
An automated intake workflow for automotive inquiries: extract the request, check pricing, update the CRM and route the next action. Built in n8n and tested across eight scenarios.
Simplified architecture · 33-node n8n workflow
The challenge
Bookings, sales questions and complaints arrived as unstructured emails in one inbox. Each needed a consistent process and a traceable record.
Objective
Capture, classify, price and route inquiries, with human review for sensitive cases and an audit trail from receipt to response.
The solution
A 33-node n8n workflow connects Gmail, Gemini, Google Sheets, HubSpot and Slack. It skips duplicates, assigns tracking references and extracts customer, vehicle and service details.
- Price estimates come from an approved Google Sheets matrix; missing data is marked insufficient.
- Refund, legal and fraud-related inquiries route to human review without an automatic reply.
- HubSpot records, Slack notifications and a Sheets activity log keep the process traceable.
Key features
- Inquiry identificationEvery message is assigned an inquiry ID and tracking number, so a case can be referenced end-to-end — in Slack, the CRM and the activity log.
- AI data extractionStructured extraction of customer, vehicle and service data with an explicit 'never guess' rule — missing data is marked insufficient, never invented.
- Pricing without hallucinationA vehicle-tier and service lookup against an approved pricing matrix returns real estimates and refuses to price what it can't match.
- Sensitivity-aware routingRefund, legal and escalation cases are flagged automatically and routed to human review with no auto-reply; standard inquiries get an immediate, professional answer.
- Full traceabilityHubSpot deals embed tracking references, and a Master Activity Log records status at every stage — from the original message to the final reply.
- Idempotent processingA data-store check prevents the same message from being processed or replied to twice, even with a crash or retry.
How the workflow flows
- 01 — Incoming emailGmail trigger (1-min poll)
- 02 — Normalize & dedupeDuplicate messages skipped
- 03 — ID + trackingINQ-*, AUTO-* references
- 04 — ExtractionAI pulls vehicle, service, intent
- 05 — Tier + pricingMatrix lookup in Google Sheets
- 06 — Sensitivity checkRefund / legal / fraud routes
- 07 — Record + notifyHubSpot deal + Slack alert
- 08 — Review or replyManual review — or auto Gmail reply
Current status
The workflow and integrations are built. Eight end-to-end scenarios passed: booking, sales, complaint, refund, legal threat, missing vehicle data, unknown service and duplicate email.
Delivery status
- Completen8n workflow (33 nodes)
- CompleteGmail trigger + normalization
- CompleteDeduplication / idempotency
- CompleteInquiry IDs + tracking numbers
- CompleteHubSpot contact & deal creation
- CompleteAI extraction (Gemini)
- CompletePricing matrix lookup
- CompleteSensitivity classification
- CompleteSlack team notifications
- CompleteAuto-reply vs manual-review routing
- CompleteMaster activity log in Sheets
- CompleteEnd-to-end test execution (8 scenarios)