All work
Support Triage Agent
- Client
- B2B SaaS company
- Sector
- SaaS
- Service
- AI
- Timeline
- 8 weeks
- Year
- 2025
- Stack
- TypeScript / Claude API / Zendesk / PostgreSQL
A three-person support team was triaging eight hundred tickets a week by hand.
The problem
Every inbound ticket waited for a human to read it, tag it, set priority, and route it. Urgent issues sat in the same queue as password resets, and response-time commitments were slipping.
Our approach
We built an agent that reads each ticket, checks the customer's account context, drafts a tagged-and-prioritized triage decision, and acts on the clear cases, while routing anything ambiguous to a human with its reasoning attached.
What we built
An always-on triage layer inside the existing helpdesk. The team reviews edge cases instead of processing a queue, urgent tickets surface in minutes, and every agent action is logged and reversible.
The results
11 minMedian first-response time, from 9 hours
71%Of tickets triaged with no human touch
2xTicket volume absorbed with the same team