An AI support agent now resolves more than 22,000 customer tickets each month without staff involvement. Last month it completed 65% of tickets end to end, cut the average initial response time to 60 seconds and received positive ratings on 70% of the conversations it handled.
The agent works with text, screenshots and PDFs inside the company’s ticketing system. It can look up account details, take actions and escalate cases to support staff. Existing standard operating procedures for refunds, billing and voice-cloning questions were converted into repeatable workflows, with structured procedures used for critical steps such as issuing refunds.
Webhook tools provide account information including subscription tier, credit use and owned voices. ElevenLabs also tests the system against roughly 50 common support scenarios and reruns those simulations after configuration changes to catch regressions before deployment.
These figures come from ElevenLabs’ own deployment of its product and have not been independently audited. The implementation nevertheless shows that automation is not just a chatbot layer: reliable resolution depends on explicit procedures, controlled access to customer data, scenario testing and a path to human escalation when the agent cannot safely finish the task.