AI-Native Wine Distributor
20 AI agents across 7 operating units
Central-European wine distributor, 150 SKU, regional operations
Problem
A traditional wine distributor was sitting on 101 months of inventory in the budget segment and carrying material receivables from a handful of debtors. Scaling with headcount was off the table — margins could not support it.
Solution
Designed an AI-native operating system where 20 agents run across 7 units: SOMMELIER and HUNTER on supply, CLOSER and NEGOTIATOR on sales, CELLAR and DISPATCHER on ops, COLLECTOR and BOOKKEEPER on finance, plus Taste Graph, Demand Prediction, Winemaker Intelligence and Pricing Engine as the data layer. Stack: Supabase, Claude API, Telegram gateway.
Impact
Architecture and approach signed off, build in progress.
- Architecture signed off by founder — equity structure agreed
- Tier-1 agents (HUNTER, CLOSER, COLLECTOR) scoped as first build block
- Infrastructure cost modelled at ~$300/month in Y1
- Data flywheel designed as the long-term defensibility moat
Stack
Client details are anonymised where disclosure was not authorised. Metrics are either directly measured or clearly marked as targets in the project log.
Want something like this for your business?
Book 30 minutes — free