How We Cut a Client's Support Volume by 70% in 30 Days
A real breakdown of how we deployed a customer support AI agent for a 40-person SaaS company, the hurdles we hit, and the exact results by week.
The client's problem
A SaaS company with 3,000 active users was receiving 200+ support tickets per week. Their two-person support team was drowning. Response times had crept to 48 hours. NPS was declining. They needed relief fast — but couldn't justify hiring a third support person for what was largely repetitive Tier-1 volume.
The solution we built
We deployed an Inbound Triage Agent connected to their Zendesk account. The agent read every incoming ticket, classified it by type and urgency, and resolved the 60% that matched known patterns automatically using their documentation. Complex or sensitive tickets were flagged and routed to the human team with a suggested response draft.
Week by week results
Week 1: Agent live, handling 40% of volume. Week 2: Confidence threshold tuned — handling 58%. Week 3: Documentation expanded based on gaps — 67%. Week 4: Full month result — 71% deflection rate. Human team response time dropped from 48 hours to under 4 hours on the remaining tickets.
What we learned
The biggest surprise was how quickly the documentation gaps became visible. The agent's uncertainty flags showed exactly where the knowledge base was missing. Within two weeks, the client had a more complete support knowledge base than they'd had in three years — as a side effect of the deployment.
Want results like this? Our Inbound Triage Agent is available in the marketplace, or we can build a custom version as part of a Build My Agent deployment.
Book a Free Call →