1
0 Comments

I built an AI executive assistant that runs my car dealership. No code team. No vendors. Just me and Claude.

I'm a New Car Manager at a high-volume Kia store in San Antonio. We push 200+ cars a month. Before I built this, my mornings looked like this:

  • 45 minutes pulling numbers from 4 different platforms
  • 30 minutes clicking through 800+ stale CRM tasks
  • 20 minutes copy-pasting into a desk log spreadsheet
  • Random email fires I didn't catch until noon

That's almost two hours before I even started managing people.

So I built AIDEN.

AIDEN is an AI assistant that handles the repetitive operational work at my dealership. Morning report lands on my phone at 6:30am. CRM tasks get triaged automatically. Desk log fills itself. Email gets classified and the urgent stuff surfaces immediately. Social media content generates weekly and posts on schedule.

The stack:

  • Claude (Anthropic) as the brain
  • OpenClaw as the runtime/orchestration layer
  • n8n for mechanical automation (posting, data routing, alerts)
  • Supabase for metrics storage
  • Browser automation for legacy dealer tools that don't have APIs (VinSolutions, kDealer, vAuto)
  • Runs on a Mac Mini sitting in my house

What it actually saves me:

  • ~2 hours/day of manual data work
  • Morning report: 45 min → 0 (done before I wake up)
  • CRM cleanup: 35 min → 5 min review
  • Desk log: 20 min → 0
  • Email triage: 30 min → 5 min scanning the AI summary

What I learned building it:

  1. AI on top of broken process just creates bad data faster. Fix your systems first, then automate.

  2. Browser automation is fragile. Sessions expire, UIs change, modals pop up. Build retry logic into everything.

  3. The 80/20 is real. 5 automations handle 90% of the value. The other 15 I built are nice-to-haves.

  4. Nobody in the car business is doing this. The "AI for dealerships" market is all chatbots and lead scoring. The operational side is wide open.

The business angle:

I packaged everything I learned into a playbook ($297) and now do consulting for other dealer managers who want the same setup. Not a SaaS. Custom builds for each store.

stevengarcia.me/playbook
stevengarcia.me/blog (12 articles breaking down each piece)

Happy to answer questions about the build, the car business, or why dealership software is stuck in 2009.

on April 4, 2026
Trending on Indie Hackers
I built a tool that shows what a contract could cost you before signing User Avatar 120 comments The coordination tax: six years watching a one-day feature take four months User Avatar 76 comments My users are making my product better without knowing it. Here's how I designed that. User Avatar 65 comments A simple LinkedIn prospecting trick that improved our lead quality User Avatar 54 comments I changed AIagent2 from dashboard-first to chat-first. Does this feel clearer? User Avatar 39 comments Why I built a SaaS for online front-end projects that need more than a playground User Avatar 18 comments