Coval, founded by a former Waymo evaluation infrastructure lead, raises $28 million from Norwest, Base10 Partners, and Twilio Ventures to build simulation-based testing for voice AI agents — applying the same safety discipline that made self-driving cars possible to autonomous voice systems.
In late June 2026, Coval — a San Francisco-based evaluation platform for AI voice agents — announced a $28 million Series A round led by Norwest, with participation from Base10 Partners, Twilio Ventures, and Y Combinator. The round brings Coval's total funding to $31 million since its 2024 launch. The company was founded by Brooke Hopkins, who previously led evaluation job infrastructure at Waymo, Alphabet's self-driving unit, where she built the simulation systems that tested autonomous vehicles across billions of virtual miles before they ever touched a public road.
Coval applies that same simulation-first philosophy to voice AI. Its platform runs tens of millions of automated tests against voice agents, probing for failure modes that manual QA often misses: how does the agent handle a thick regional accent, a sudden interruption, background noise from a construction site, or a customer who changes their mind mid-sentence? After deployment, Coval continuously monitors live calls and automatically feeds failed interactions back into the testing pipeline, creating a closed loop where every production failure becomes a new test case. The company claims customers can reduce manual QA work by up to 30 times and accelerate deployment timelines by up to 10 times compared to traditional testing methods.
Over 60 organizations already use Coval, including Zoom and Deepgram, alongside Fortune 500 enterprises deploying voice agents in customer service, healthcare, and financial services. Revenue has grown 10x year-over-year, reflecting the urgency as enterprises rush to deploy voice agents into production. Coval competes with Hamming, which focuses on regulatory edge cases in healthcare and finance, and Roark, a Y Combinator graduate specializing in replaying failed conversations with updated agent logic. Coval differentiates by offering a full-stack solution — pre-deployment simulation, live production monitoring, and human review workflows — all purpose-built for voice audio rather than adapted from text-based testing tools.
The company's origin story resonates because it mirrors the trajectory of autonomous vehicles: a technology that seemed perpetually years away until rigorous simulation-based testing made it production-safe. More than $7 billion was invested in voice AI during Q1 2026 alone, and the market is forecast to exceed $20 billion by 2031 — but deployment velocity has outpaced testing rigor. Coval's bet is that enterprises won't entrust customer relationships to voice agents until they can prove those agents work reliably across the full diversity of real-world speaking conditions, and that the evaluation layer will become as essential to the voice AI stack as it is to autonomous driving.
