Coval

Coval builds a simulation and evaluation platform for voice and chat AI agents. Engineering, product, QA and operations teams use it to test agents before deployment, monitor live conversations and review results with human evaluators. Teams can simulate thousands of conversations to detect regressions as prompts, models and workflows change. Its voice-specific analysis covers audio quality, telephony latency, transcription errors and agent workflows, helping teams investigate failures that a handful of manual test calls can miss.

Founded in 2024 by Brooke Hopkins, its founder and CEO, Coval adapts simulation methods from autonomous vehicles to conversational AI. Hopkins previously led evaluation job infrastructure at Waymo. That background informs its approach to testing changes systematically before deployment. Coval also publishes reproducible speech-provider benchmarks with public methodology, code and input data. Its text-to-speech tests measure time to first audio, latency variation and intelligibility through word error rate, giving teams concrete measurements for evaluating providers.

In June 2026, Coval reported adoption by over 60 organizations, including Zoom and Deepgram. It raised a $28 million Series A led by Norwest that month, with participation from Base10 Partners, Twilio Ventures and Y Combinator, bringing its announced total funding to $31 million since launch.