← All case studies

E-commerce · Sample

An AI support agent that resolves most tickets on its own

A customer-support agent that answers order, returns and product questions end to end, and hands off to a human with full context when it can't.

ClientD2C skincare brand
Timeline6 weeks
ServicesAI agents · APIs & integrations
62%
tickets resolved without a human
< 1 min
first response time
4 weeks
from kickoff to live pilot

The challenge

Support volume tripled during sales. The team of six spent most of its day on the same order-status and returns questions, and response times slipped past 24 hours.

What we built

We built an LLM agent grounded in the brand's help centre and product catalogue (retrieval-augmented generation), connected to the order and returns APIs so it can look up and act, not just answer. Guardrails route refunds above a threshold and angry customers to a person, with a summary attached.

The outcome

The agent now handles routine tickets around the clock, and the human team focuses on complex cases and retention.