An AI agent that resolves support tickets end to end
A tool-using support agent that checks orders, processes returns within policy and hands complex cases to people.
- Client
- Online retailer
- Industry
- E-commerce
- Timeline
- 12 weeks
- Year
- 2025
The challenge
Support volume tripled during seasonal peaks. Most tickets were routine order and return questions, but agents spent minutes per ticket switching between the helpdesk, the order system and the payment provider.
Our approach
We built an agent with scoped access to order lookup, return creation and refund APIs. Policies are enforced in code, every action is traced, and anything above a refund threshold or below a confidence threshold goes to a human with a drafted reply.
Results
- Tickets fully automated
- 62%
- Customer satisfaction
- 4.6/5
- Median resolution time
- 90s
- Routine tickets resolved without human touch
- Peak-season backlog cleared without temporary hires
- Every agent action traceable in an audit log
Services
Technology
- Python
- Claude
- LangGraph
- PostgreSQL
- Zendesk API
- AWS
Screenshot placeholder 1
Screenshot placeholder 2
Screenshot placeholder 3
Next project
One operations platform for a growing logistics fleet
Let's build something remarkable.
Tell us where you want to be in twelve months. We'll show you how software, AI and automation can get you there.