SaaS

10 agents, 400 conversations/day, fully automated AI routing

Context

A B2B SaaS company with 10 support agents on Intercom, handling ~400 conversations/day across multiple product lines and customer tiers.

4 weeks

Challenge

No automated assignment whatsoever. Every conversation landed in a shared inbox. VIP customers waited alongside free-tier questions, workload was unevenly distributed, and agents spent time cherry-picking conversations instead of resolving them. Manager had zero visibility on who was handling what.

What was delivered

Built an AI classification layer using OpenAI to analyze each incoming conversation (topic, urgency, customer tier) and route it to the right agent or team automatically.

Implemented priority rules: VIP customers get routed first, spam and junk filtered out before reaching the queue.

Created a custom agent-facing app (Zite) for toggling availability, viewing assignment history, and monitoring real-time workload distribution.

Added capacity balancing so no single agent gets overloaded while others sit idle.