AI Customer Support · Personalized Shopify Jewelry

Resolving 40% of Shopify Customer Requests Before They Reach the Inbox

An AI support chatbot for a custom jewelry brand that answers live order and design-stage questions, collects required details and photos for complex issues, and reduces email back-and-forth by 75–80% when the support team takes over.

40%: Requests resolved directly in chat
75–80%: Less email back-and-forth on escalated requests
8: Guided ecommerce support workflows

Problem

The store’s support team handled all complex tickets by email. Because personalized jewelry orders go through custom design, preview approval, production, and shipping stages across separate systems, basic Shopify status bots couldn't answer customer questions.

Resolving requests about order updates, unseen design previews, missing deliveries, damaged items, or order changes required 3–4 back-and-forth emails just to locate the order, identify the affected item, and collect photos.

Solution

I built a rule-guided AI chatbot connected to Shopify, the store’s custom order-production system, Notion, and Reamaze.

Using deterministic workflow rules paired with Google Gemini, it answers routine questions directly in chat using live order, production, and knowledge-base data. When human action is required, it gathers the exact details and photo evidence first, then creates a structured Reamaze email ticket so agents can resolve the issue in a single reply.

Wrong-item support before and after: the chatbot collects the order number, affected item and photo, then creates a complete Reamaze ticket so an agent can resolve it in one reply. — For wrong-item requests, the chatbot collects order details, item selection, and photo evidence before handing a complete ticket to Reamaze.

What I Delivered

Eight guided support workflows: Order status, missing deliveries, missing design previews, wrong or damaged items, order changes, cancellations, reorders, and pre-purchase questions.
Live multi-system lookup: Real-time retrieval of Shopify order data, custom design/production stages, and shipping tracking.
Knowledge base integration: Accurate policy and product answers pulled from Notion.
Validated photo collection: Upfront image uploads in chat for damaged or incorrect item claims.
Stateful conversation memory: Multi-turn session tracking so customers never repeat their order number or issue.
Guardrails & safe handoffs: Strict workflow gates prevent AI hallucinations and package full conversation context into Reamaze tickets.

Results

40% of customer requests resolved directly in chat, with zero support ticket created.
60% of requests pre-qualified and escalated to email with order context, item selection, and photos already attached.
75–80% reduction in email back-and-forth on escalated tickets, turning 3–4 message threads into one-touch resolutions.

Stack

Python · FastAPI · Shopify API · Notion · Reamaze · Google Gemini · LangChain · PostgreSQL · Vercel

Problem

Solution

What I Delivered

Results

Stack

Connect your questions to real action