Meridian Commerce · E-commerce · 2025

An AI support agent that took over 74% of ticket volume

74%

of tickets resolved without human touch

9 sec

median first response time, down from 12 hours

+18pts

CSAT improvement in the first quarter

$21k/mo

support cost reduction at current volume

Challenge

Support volume was scaling linearly with orders — 6,000+ tickets a month, 70% of them order-status, returns, and sizing questions. First-response time had slipped past 12 hours, CSAT was falling, and the support team of eight was burning out on questions a system should answer.

Solution

We built a conversational agent connected directly to Shopify order data, the returns policy engine, and the product catalog. It resolves order-status, returns initiation, and product questions end-to-end, escalating edge cases to Zendesk with full context. A supervised launch month tuned it on real conversations before full rollout.

Process

01

Two-week discovery: transcript analysis of 10,000 historical tickets to map resolvable intents

02

Agent build with retrieval over policies and live Shopify integration

03

Supervised soft launch — every response reviewed for two weeks, failure modes tuned

04

Full rollout with weekly quality reviews and monthly tuning cycles

Stack

Claude APINext.jsShopify APIZendeskPostgreSQL

Head of CXMeridian Commerce

Next case study

A lead engine that 3×'d qualified pipeline in one quarter

30 minutes. No pitch deck. We'll map where AI and automation create the most leverage in your operation — and give you the roadmap either way.