Joval treats voice and chat agents with artificial intelligence as self-driving cars

What do artificial intelligence, voice agents and self-driving cars have in common? Their performance can be evaluated in the same way, claims Brooke Hopkins, former chief technology officer at Waymo. Hopkins’ new startup, Joval, is trying to do just that.

“When I left Waymo, I realized that most of these challenges we were having at Waymo were the same ones the rest of the AI ​​industry was facing,” Hopkins (pictured above, center) told TechCrunch. “But everyone said this is a new paradigm, we have to come up with experiments from first principles, and basically we all have to reinvent everything. I looked at it and said, wait, we’ve spent the last 10 years driving around figuring out how to do this.”

He decided to start work in 2024 KovalA platform that builds simulations for AI voice and chat agents that test and evaluate how they perform tasks, much like Hopkins is testing self-driving cars at Waymo. Coval can run thousands of simulations simultaneously, such as an agent making a restaurant reservation or an agent answering an indirect customer service question.

Coval’s technology evaluates agents against a common set of metrics, but companies can also customize what they’re looking for and use Coval to continue evaluating regressions. Users can also take this data and the insights they spark from it and deliver it to their end customers as a demo or monitoring tool to show them that the agent is working as intended.

“One of the biggest barriers to agents being adopted by businesses is the lack of confidence that it’s not just smoke and mirrors,” Hopkins said. “Choosing between vendors is really complicated for these executives because it’s so hard to know what you’re asking or how to prove that these agents are doing what you expect. So it gives our companies an opportunity to really show and demonstrate that.”

Hopkins actually formulated the idea behind Joval during the Y Combinator Summer 2024 party before launching the product publicly in October 2024. He said demand was strong and explosive in the past two months, with customers asking how soon they could get their agents. is evaluated.

The San Francisco-based startup is now announcing a $3.3 million round led by MaC Venture Capital, with participation from Y Combinator and General Catalyst. The startup will work with capital to build its engineering team and product-market fit. Hopkins added that the company will also work to enable its users to evaluate other types of AI agents in the future, including web-based agents.

Joval enters the scene at a time when both momentum and hype around AI agents is at an all-time high. Enterprise technology leaders like it Marc Benioff He praises the technology (and its marketing), saying that Salesforce will deploy more than one billion AI agents next year. OpenAI rumor has it that he will reveal his thoughts on the AI ​​agent very soon.

There are also plenty of startups in the space. Among Y Combinator’s three 2024 cohorts alone, there were more than 100 startups building AI agents. Some AI agent startups have also raised large venture funding rounds. one, /dev/agentsIn November 2024, it raised a $55 million seed round valued at $500 million, less than a year after it was founded.

This momentum means there will be more companies looking for help evaluating their agents. Hopkins said Coval has a good chance to stand out from the pack because, unlike inevitable newcomers, Coval has a head start.

“I think where we really stand out is that I’ve been in this field for half a decade and I’ve built these systems many times,” he said. “We built a lot of iterations and saw how they failed and how they scaled, and we’re building the same insights into Joval and all of these learnings.”