Coval Raises $28.2M Series A to Build the Trust Layer for Enterprise Voice AI

The money isn't chasing AI voice anymore. It's chasing confidence. Every week another AI voice agent appears promising faster conversations and lower costs. Enterprise buyers aren't asking whether AI can answer the phone anymore. They're asking whether they can trust it when conversations stop following the script. That question is fueling one of the latest funding rounds in AI infrastructure.

What Happened

San Francisco-based Coval, an AI voice and chat agent simulation, evaluation, and observability platform, has raised $28.2M in Series A funding led by Norwest Venture Partners, with participation from Base10 Partners, Twilio Ventures, Y Combinator, MaC Venture Capital, Swift Ventures, and Alumni Ventures. Founder & CEO Brooke Hopkins now leads a company that has raised approximately $31M following its $3.3M seed round. A graduate of Y Combinator Summer 2024 (YC S24), Coval is expanding the infrastructure enterprises use to test, monitor, and improve AI voice deployments before and after they reach customers.

The Series A reflects more than product momentum. It represents conviction from investors including Scott Beechuk, Nikhil Goel, and Andy O'Dower of Norwest Venture Partners, alongside continued support from Harj Taggar, Managing Partner at Y Combinator. Together, the funding reinforces growing demand for infrastructure that helps enterprises deploy AI voice systems with greater confidence and consistency.

Why This Matters

Brooke Hopkins didn't stumble into this problem. At Waymo, Brooke Hopkins led evaluation infrastructure responsible for running millions of simulations across 20 petabytes of data to help validate autonomous driving systems. The challenge wasn't building intelligence. It was proving reliability under conditions nobody could perfectly predict. Enterprise voice AI faces the same problem, where the real test begins when customers interrupt, change topics, introduce background noise, or simply behave like humans instead of demos.

Coval's platform combines AI simulation, evaluation, observability, and human review into a continuous improvement loop. Enterprises can simulate thousands of conversations before deployment, monitor production calls in real time, identify failures, and feed those learnings back into future testing. According to the company, customers can reduce manual checks by up to 30x while accelerating deployments by as much as 10x. Reliability, not novelty, is increasingly becoming the purchase decision for enterprise AI.

Market Context

The platform already supports customers including Perplexity and Deepgram, while integrations with Vapi, Retell AI, and Bland position Coval at the center of an expanding enterprise voice AI ecosystem. Tens of millions of simulations and production calls every month create a feedback loop where every interaction strengthens future evaluations, improves deployment confidence, and steadily raises the standard for enterprise AI reliability.

That momentum is driven by the people building the platform as much as the technology itself. Alongside Founder & CEO Brooke Hopkins, the team includes Kobi Hudson, Dakota Mallen, Loren Phillips, Rhea Pokorny, Callum Reid, Ale Vergara, Henry Finkelstein, Rob Young, and Mallory McLoughlin, Chief of Staff. Their work sits beneath the surface of enterprise AI, but it is increasingly becoming the foundation customers depend on.

The Bigger Industry Shift

The broader market is moving in the same direction. As enterprises shift from experimenting with conversational AI to operating it at scale, evaluation infrastructure becomes less of a developer tool and more of an operational requirement. AI models may capture the headlines, but the companies building the systems that measure, validate, and monitor those models are quietly becoming the foundation underneath the next generation of enterprise software.

Infrastructure rarely earns standing ovations because success looks uneventful. Calls connect. Customers get answers. Nothing breaks. In enterprise AI, that kind of silence isn't boring. It's the product.