What do AI voice assistants and autonomous vehicles have in common? Their performance can be assessed using a similar approach, says Brooke Hopkins, a former tech lead at Waymo. Hopkins’ new venture, Coval, is designed to tackle this very challenge.
“When I left Waymo, I noticed that many of the issues we faced there were also affecting the broader AI industry,” Hopkins explained in an interview. “But everyone seemed to believe they were starting from scratch, building testing frameworks from the ground up. I thought, Wait, we’ve spent a decade solving these kinds of problems for self-driving cars—why not apply those lessons to AI agents?”
In 2024, Hopkins launched Coval, a platform that creates simulations to test and analyze how AI voice and chat agents perform various tasks—similar to the way she evaluated self-driving systems at Waymo. Coval’s platform can run thousands of simulations simultaneously, assessing how an AI agent handles tasks like making a restaurant reservation or responding to an indirect customer service query.
Coval’s technology provides a set of general performance metrics, but companies can customize these to fit their specific needs. The platform also allows continuous evaluation, helping businesses identify regressions and monitor performance. Additionally, the insights gained from these tests can be used to showcase the agent’s capabilities to clients, either through demos or as ongoing proof of reliability.
“Building trust is one of the biggest challenges businesses face when adopting AI agents,” Hopkins explained. “Executives often struggle to determine whether these tools truly meet their expectations or if they’re just flashy demos. Coval gives businesses a way to validate and demonstrate that their agents are effective and reliable.”
Hopkins developed the concept for Coval during the Y Combinator Summer 2024 cohort and publicly launched the product in October 2024. Since then, demand has surged, with companies eager to evaluate their AI agents as quickly as possible.
The San Francisco-based startup recently secured $3.3 million in seed funding, led by Mac Venture Capital, with additional support from Y Combinator and General Catalyst. The funding will be used to expand Coval’s engineering team and refine the product to achieve a strong market fit. Hopkins also plans to extend Coval’s capabilities to include other types of AI agents, such as web-based systems, in the future.
Coval enters the market at a time when enthusiasm for AI agents is reaching new heights. Industry leaders like Salesforce’s Marc Benioff have highlighted the transformative potential of these systems, with Salesforce planning to deploy over a billion AI agents next year. Meanwhile, companies like OpenAI are rumored to be preparing their agent-based offerings.
The sector is buzzing with activity, with more than 100 AI agent startups emerging from Y Combinator’s 2024 cohorts alone. Some, like dev/agents, have already attracted significant investment, such as a $55 million seed round at a $500 million valuation in late 2024.
Amidst this momentum, Coval aims to stand out by leveraging Hopkins’ extensive experience. “I’ve spent years building and iterating on these systems, learning from both their successes and failures,” she said. “We’re taking all of those lessons and embedding them into Coval to create something truly impactful.
