The Problem
A Saudi DTC home goods brand was selling on Salla, Amazon.sa, Noon, and Instagram. Customer service was running through a 6-person team handling Arabic and English in roughly equal mix; response times averaged 3 hours during the day and 14 hours overnight. Product listings were maintained separately on each marketplace. when the catalogue changed, three people spent a day each updating their channel.
The brand was profitable but couldn't scale customer experience without doubling the CS team. which would erase the margin gain.
What I Built
1. Arabic + English customer service AI. A Claude-powered agent handles WhatsApp and Instagram DMs in Modern Standard Arabic, Saudi dialect, and English. It has read access to:
- Order status (lookup by phone or order ID)
- Product info (specs, materials, dimensions, care instructions)
- Returns and shipping policy
- Mada/STC Pay payment status
Routine questions (40-65% of volume in any given week) are auto-handled. Anything outside scope is escalated to a human CS agent during business hours, with a polite "we'll get back to you within 2 hours" template after-hours.
2. Multi-channel listing automation. A central product database (Postgres) is the source of truth. When a product is added, edited, or stock changes, n8n workflows push to Salla (API), Amazon.sa (SP-API), Noon (Marketplace API), and Instagram Shopping. New product launches that used to take 3 days now take 2 hours.
3. Automated post-purchase upsell. Order confirmation triggers a tailored follow-up at day 5 (ratings + reviews ask), day 21 (bundle recommendation based on the customer's purchase), and day 60 (reorder reminder for consumable items). Personalised in Arabic where appropriate.
Outcome
The brand absorbed 3× monthly order volume in 9 months without growing the CS team. AOV climbed 52% from the post-purchase upsell flow alone. The founder's biggest comment: "We finally feel like a brand, not a startup that responds to messages with two-day delay." The brand is in expansion talks for UAE; the same automation stack is being prepped for the new market.