ElevenAgents is going into A/B testing mode.

ElevenLabs launches Experiments for ElevenAgents to A/B test voice agent prompts, workflows, and guardrails on live traffic using CSAT and cost metrics.

ElevenLabs is launching Experiments within ElevenAgents, a module for testing different versions of a conversational agent on real traffic. Prompts, workflows, voices, or guardrails can be modified, then exposed to a controlled percentage of users. Performance is measured via concrete metrics: CSAT, conversion rate, response time, cost per resolution. Each variant is versioned, traceable, and reversible. Objective: to replace intuition with production data, and continuously optimize voice agents in an enterprise environment.