Coval Raises $28M to Scale Voice AI Testing

The big picture: Brooke Hopkins, an ex-Waymo engineer, founded Coval to bring rigorous autonomous vehicle testing standards to AI voice agents. The company today raised $28 million in a Series A round led by Norwest, bringing its total funding to $31 million since launching in 2024.

Why it matters:

• Growing demand: The voice AI market is rapidly expanding, with over $7 billion invested in Q1 2026 and projections to exceed $20 billion by 2031, as companies integrate voice agents into customer service, healthcare, and financial services.
• Infrastructure gap: Despite rapid adoption, most enterprises lack the robust infrastructure needed to confidently deploy and manage AI voice agents, often relying on unscalable manual testing methods.
• High stakes: Failures in voice AI agents can have serious consequences, mirroring the critical need for reliability seen in autonomous vehicles, where Hopkins previously worked on evaluation systems.

How it works:

• Simulated testing: Coval runs tens of millions of simulated tests for voice agents, checking for issues related to accents, interruptions, background noise, and unusual conversational situations before agents interact with real users.
• Continuous monitoring: Post-launch, Coval continuously monitors live agents and automatically feeds failed calls back into the testing environment, ensuring ongoing reliability and performance improvement.
• Efficiency gains: The platform significantly reduces manual quality checks by up to 30x and accelerates deployment timelines by up to 10x, allowing companies to launch and update voice agents with greater speed and confidence.

The catch: Coval faces competition from specialized players like Hamming, which focuses on regulatory edge cases, and Roark, which excels in replaying failed conversations. A key industry question remains whether voice AI evaluation will stay independent or become integrated into larger platform offerings. Twilio’s investment in Coval suggests a preference for independent evaluation tools, at least for now.