← All case studies
E-commerce

60% faster support response with an AI agent

A conversational agent with tool calling that handles support chats and return requests end to end.

Read the architecture
The problem
A growing e-commerce support queue pushed first-response times far past acceptable limits during peak season.
The solution
We built an LLM agent with function calling over the order and returns APIs, with confidence-based escalation to human agents.
−60%
Response time
68%
Auto-resolved
4.5/5
CSAT
Stack
GPT-4oFunction callingPostgresRedis queue