The Car Dealership AI Stack — A Real Case, Not a Demo
Car dealerships lose deals in the gaps: the lead that came in during Sunday lunch, the call that rang out while every salesperson was on the floor, the trade-in inquiry that sat unanswered overnight. Buyers cross-shop; the first dealership that responds usually gets the test drive.
This page documents the automation stack running at a real multi-brand used-car dealership in Brazil — one of the most WhatsApp-mature car markets in the world. The patterns transfer directly to dealerships (and any local business) in the US, Europe or LatAm.
What is actually automated in production
- Lead capture and instant WhatsApp reply. Every inquiry — website, marketplace listings, click-to-WhatsApp ads — gets an answer in under a minute, any hour. The bot asks which vehicle, collects the buyer's timeline, and hands off to a salesperson with context attached.
- Missed-call recovery. Unanswered phone call → automatic WhatsApp message within seconds ("Sorry we missed you — which car are you calling about?"). A ringing phone that nobody caught becomes a conversation instead of a lost lead.
- Inventory marketing on autopilot. New vehicle enters stock → photos and copy are generated and queued for social media on a schedule, with human approval before anything publishes. Sold vehicle → listings updated everywhere.
- Follow-up cadence. Interested but not ready buyers get a spaced follow-up sequence — day 1, day 3, day 7 — that stops the moment they reply or opt out.
The stack and its real monthly cost
| Job | Tool | Monthly cost |
|---|---|---|
| WhatsApp API + team inbox | Wati (or respond.io if you also handle Instagram DMs) | $49–99 |
| Workflow engine | n8n — self-hosted on a small VPS | ~$6 (VPS) or $26 cloud |
| AI copy and image handling | LLM API usage inside n8n | $5–20 |
| Voice answering (optional) | Retell AI when calls overflow | ~$0.11/min used |
Total: roughly $60–150/month for the core. Compare that against the margin on one extra car sold per month — the math isn't close. If phone volume is significant, run the AI voice agent calculator to see the AI-vs-receptionist line at your minutes.
Lessons from running it (the part demos skip)
- Approval gates matter. Nothing publishes to social media without a human tap. Automation drafts; a person approves. This single rule prevented every embarrassing mistake.
- Never let the bot pretend to be human. The WhatsApp flow says what it is, collects what it needs, and hands off. Buyers don't mind bots that are fast and honest; they hate bots that fake it.
- Log everything. Every message, publish and handoff is recorded. When something misbehaves, the log answers "what happened" in minutes.
- WhatsApp beats email for local business. Open rates aren't comparable. If your market uses WhatsApp (LatAm, Spain, India, growing fast in the US), meet buyers there.
Build it yourself
The lead follow-up and missed-call workflows are available as import-ready n8n templates on our templates page — the free versions cover the core loop; the paid dealership pack adds inventory posting and the full follow-up cadence. The general-purpose version of this pattern is documented in the lead follow-up stack.
Affiliate note: some links are affiliate links (marked sponsored). The stack is what actually runs — tool choices were made before any affiliate relationship existed.