Chen Deng

~/projects /lamose-ai-workforce

LAMOSE — an e-commerce brand run by an AI workforce

role

Owner / operator

period

2024 — present

status

live

16+

autonomous agents

21

scheduled jobs

~979

etsy listings managed

LAMOSE is my custom-engraved drinkware brand — tumblers, bottles, mugs, hockey pucks, name tags — selling through Shopify, ~979 Etsy listings, Faire wholesale, and Amazon, with four international trade shows behind it. It’s also my proving ground for a thesis: a single operator can run a six-figure e-commerce business if AI agents do the daily work.

The workforce

16+ agents run on cron and launchd (21 scheduled jobs on a Mac mini), each owning a lane:

  • Forecast — revenue pacing against monthly targets with P50/P90 projections
  • Inventory — days-of-supply tracking against vendor lead times, reorder proposals
  • Pricing — competitor monitoring against YETI, Stanley, and Hydro Flask
  • Ads — Google Ads management (its own case study — see the AI Ads Manager)
  • Etsy Guardian — daily listing title/tag optimization and competitor keyword intelligence
  • Mockups — AI-generated product mockups from real reference photos for listing refreshes and B2B previews
  • CRM, Outreach, Reviews — B2B pipeline, review responses, wholesale follow-ups
  • Night Ops, Growth, Analytics, Compliance, Production, Communications — everything else

Each morning the fleet posts briefings and proposals to Slack. I read, approve, or redirect. The business runs while I build.

The governance layer

Letting agents touch a real business without guardrails is how you wake up to a disaster. Every agent operates under a written autonomy framework with three tiers:

  1. Auto-execute — low-risk, reversible actions (report generation, data syncs)
  2. Propose with reasoning — anything touching money, customers, or public content waits for my approval in Slack
  3. Weekly strategy — direction-level recommendations batched for review

There’s a hard “never-touch” list, brand-voice rules, and an Airtable decisions log with one rule: no log entry, no action. Every agent decision is auditable after the fact.

Why it matters

This isn’t a chatbot bolted onto a store. It’s an operating model: the agents do the recurring cognitive work — monitoring, drafting, reconciling, proposing — and the human does judgment, taste, and relationships. The same pattern now runs my pharmacy and my consulting practice.

stack: Claude · DeepSeek · Python · launchd · n8n · Shopify API · Etsy API · Airtable · Slack

Want a system like this? → hire me