For the longest time, AI voice agents and self-driving cars have been considered to have little to nothing in common, but lately, a former tech lead at Waymo has argued differently. She states that their evaluation can be performed similarly, as both rely on decision-making in complex, unpredictable environments. Drawing parallels between the rigorous simulation testing used to ensure the safety of self-driving cars and the performance of conversational AI agents, she identified a gap in how AI voice agents are evaluated. Therefore, with her experience with self-driving cars, she decided to launch Coval.
Coval AI is a platform that helps developers build reliable voice and chat AI agents using AI-powered simulations to generate extensive test cases. Founded by Brooke Hopkins in 2024, Coval applies principles from autonomous vehicle testing to evaluate AI agents’ performance with metrics like accuracy and latency. For example, imagine you are a hotel owner and you want to create a chatbot that will be used to interact with potential clients and help them book a room. Coval’s simulation and evaluation platform is the game changer. Instead of reinventing the wheel, you can use the Coval framework to easily tune your chatbot or AI voice agents to your wishes.
There is a general set of metrics that are used to evaluate any agents throughout the Coval testing ecosystem. Furthermore, companies can also create custom metrics to closely define what they want from the AI assistant. Why are the metrics so important and what does Coval AI make so special? One of the main reasons why Coval AI is such a huge success is because it can prove to you how good your model is. With the extensive sets of performance benchmarks and an almost infinite amount of demonstrations, you can easily be assured that your model has achieved the desired capabilities.
Coval AI joins the current AI evolution by tackling a critical challenge: ensuring the reliability and scalability of conversational AI systems. As AI adoption accelerates across industries, the need for robust testing, evaluation, and real-time monitoring of AI agents has grown exponentially. Coval’s approach, which leverages simulation principles from autonomous vehicle development, aligns with the broader push for AI safety and optimization. It enables the automation of complex testing and seamless integration into the existing pipelines.
This startup is now announcing a $3.3 million seed round led by MaC Venture Capital with participation from Y Combinator and General Catalyst. Regarding the future of Coval AI, it is easy to say that they will revolutionize the way we think about Artificial Intelligence in general. In the interview, Hopkins added apart from chatbots and voice agents, the company will start incorporating support for web-based agents by focusing on building a strong engineering team.