All work
E-commerce, 15k SKUsEU5 weeks build · 2 weeks supervised rollout

92% of support tickets resolved without a human.

One AI agent handling Facebook, Instagram, TikTok, YouTube, and on-site chat for a 15,000-product catalog.

92%
tickets resolved without a human
<60s
median first response, 24/7
5 → 1
inboxes consolidated
3x
DM response volume increase
Challenge

The state we found.

Five inboxes, three support reps, fifteen thousand products. The team was answering the same product questions hundreds of times per week, and DMs from TikTok and Instagram were going unanswered for 12+ hours.

Every channel had its own UI. There was no single place to see a customer's history. Outside business hours, nothing moved.

Solution

What we built.

A single agent with full catalog access answers product, stock, and shipping questions across every channel. It only escalates complaints, refund requests, and edge cases — with full conversation context attached for the human.

Architecture
// 6 steps · production
  1. 01
    Catalog ingestion from WooCommerce into a vector index, refreshed hourly
  2. 02
    Meta Graph API webhook handlers for Facebook, Instagram, and WhatsApp inbound
  3. 03
    TikTok and YouTube comment polling via headless workflows
  4. 04
    Claude agent with tool access: stock check, order lookup, shipping ETA, FAQ
  5. 05
    Confidence + intent classifier: refund / complaint / human-requested → escalation queue
  6. 06
    Unified inbox in Supabase with full thread history per customer across channels
Stack
ClaudeMakeMeta Graph APIWooCommerceSupabase
Next case study

Reservations, deposits, and reviews — fully automated.

Read next
Final word

Book the call. Bring the bottleneck.

30 minutes. No deck. We either see the automation in your process — or we tell you it is not worth building.