Support tickets stop needing a human: ElevenLabs runs 22,000 a month on ElevenAgents, 65% resolved end to end

ElevenLabs' support team runs its own customer support on ElevenAgents, an AI support agent that works with text, screenshots and PDFs, is integrated with the ticketing platform, handles long-running conversations, looks up account information, takes actions, and escalates cases to support staff when needed. Initial response previously took hours; after adding the ElevenAgents teammate, average initial response time is 60 seconds. Last month the agent resolved 65% of tickets end to end, and 70% of all conversations it handled were rated positively by users.

The build turns the support team's existing standard operating procedures into repeatable agent workflows rather than teaching the agent from scratch. The team already had detailed standard operating procedures for every major ticket category, including refunds, billing changes and professional voice clone questions, and the core of the build was translating that existing knowledge into something an agent could execute reliably. Critical actions such as issuing refunds use Structured Procedures so each step is taken deterministically, and webhook tools let the agent dynamically look up subscription tier, credit usage and user-owned voices.

Simulation tests and shadow mode gate the rollout, and Architect keeps diagnosing and improving the agent afterward. The team used simulation tests across roughly 50 scenarios its support team encounters most frequently, and Architect iterated on the system prompt and procedures until all tests passed, with the suite re-run after config changes to catch regressions. Shadow mode showed how the agent would behave on production traffic inside the ticketing system before it was rolled out, and Architect answers a single natural language query with a breakdown of top ticket categories and diagnoses whether an escalation came from a knowledge base gap, a missing tool or an internal policy that needs clarifying.

This report is for researchers and engineers and covers how ElevenLabs runs its own customer support on ElevenAgents: the agent works with text, screenshots and PDFs, is integrated with the ticketing platform, looks up accounts, takes actions and escalates to staff when needed. The reported figures are over 22,000 tickets a month resolved without needing to involve a person, 65% of last month's tickets resolved end to end, a 60-second average initial response, and 70% of handled conversations rated positively. The team validated with simulation tests across roughly 50 frequent scenarios and with shadow mode before deployment, and re-ran the test suite after config changes.

The figures come from ElevenLabs' own support operation and cover its own ticket mix and product portfolio. The team keeps humans in the loop to understand customer needs, update policy, and avoid overfitting to any given type of ticket. The text does not describe what conditions other organizations would need to reproduce the workflow.