The unsexy 80%: integration, not models
People expect AI case studies to be about clever models. The Vigor project was 80% integration plumbing and 20% AI. The win came from making 7 separate tools — Shopify, Gorgias, Klaviyo, ShipStation, QuickBooks, Slack, Google Sheets — talk to each other reliably.
"We thought we'd need to hire an ops manager. Instead Growvate built one. It's been running for four months and we forgot it existed — in the best way." — Jordan K., Co-founder, Vigor Beans
The five automations that did the work
- Inventory sync. Every Shopify sale decrements inventory across both warehouses, triggers reorder thresholds, and posts daily Slack summaries.
- Support inbox. 86% of customer emails (mostly "where is my order") are auto-answered with order-aware context pulled from Shopify and ShipStation.
- Newsletter generation. Every Sunday, a Claude prompt drafts that week's email — featured roast, customer story, brew tip — then posts to Klaviyo as a scheduled send. A founder approves in 60 seconds on Sunday night.
- Roast scheduling. The system pulls 30-day demand trends and pre-orders, then generates a roast schedule the founders approve each Monday.
- Subscription save bot. When a subscriber clicks "cancel", they get a Claude-powered short conversation offering a 1-month pause, a different roast, or a discount. Churn dropped from 11% to 6.4%.
What we'd do differently
We'd build the founder approval surface first. Vigor's founders are hands-on — they wanted to see what the automations were doing, not just trust them. The first version we shipped without an approval dashboard, and they didn't trust it. Two days of building a one-screen "what AI did today" dashboard fixed that completely.