VoiceRun nabs $5.5M to build a voice agent factory

Nicholas Leonard and Derek Caneja wanted to build AI voice agents, but when they went to build the product, they felt many of these voice agents had design flaws. 

Some of these agents were being built with no-code tools, meaning shipping to production was fast, but the quality of the product was often low. Other agents were being made by companies that had the time and resources to spend months building specialized tools. “Developers and enterprises needed an alternative,” Leonard told TechCrunch, adding that he and Caneja also realized that the future of software would be “coded, validated, and optimized by coding agents.” 

“These two insights and a historical realization gave us the inspiration for VoiceRun,” Leonard, the company’s CEO, said. Caneja is the company’s CTO. 

Last year, they decided to launch VoiceRun, a platform that lets developers and coding assistants launch and scale voice agents. Right now, many of these low-code platforms let people build voice agents with visual diagrams, where people click through conversation flows and write prompts into boxes that then dictate how the agent should behave. All of that can be hard to manage, Leonard said. 

VoiceRun, on the other hand, lets users code how they want their voice agents to behave, giving them more flexibility in creating the product they want. Code is the native language of coding agents, Leonard explained. “They are going to do a far better job operating in code than in a visual interface,” Leonard said.

Furthermore, with visuals, there are limited configuration options, so, for example, if someone wanted to build a voice agent that could speak in a different dialect, it might be harder to do if the maker of the visual interface didn’t build a feature that can handle that task.

“But in code, it’s incredibly simple to do,” he said. “There is a long tail of millions of examples of little things you might want to do that aren’t supported by the visual interface.” 

Aside from coding agents, VoiceRun also lets users perform A/B testing and deploy instantly with one click. 

The company is geared toward enterprise developers, helping companies, for example, incorporate AI into their customer services, or help tech companies launch voice-based products. He mentioned, for example, working with a restaurant-tech company launching an AI phone concierge for food reservations.

The company announced on Wednesday the closing of a $5.5 million seed round led by Flybridge Capital. 

There is a lot of competition in the AI agent space. Startups in this area last year nabbed billions of dollars (out of the many billions that flooded into AI companies in general). Leonard feels his company is up against two ends of the market: There are the no-code voice builders, like Bland and ReTell AI, he said, that lets user build quick demos. There are also more sophisticated tools, like LiveKt and Pipecat, which give developers “maximum control.” He feels Voicerun sits in the middle of these two ends. 

“​​We provide global voice infrastructure and an evaluation-driven lifecycle, while keeping ownership of business logic code and data in the customer’s hands,” he said. “The key difference is that we are closing the loop for end-to-end coding agent development. We expect developers to be supervising coding agents that write code, run tests, deploy, and propose improvements.” 

In some ways, Leonard is hoping his product helps developers create voice agent tools that will, in turn, help people feel more comfortable with automated voices. Customers today “feel relief” when a human answers the phone, “because voice automation has been brittle and ineffective.” 

A survey from Five9 last year showed that three-fourths of its survey respondents still prefer talking to a human when it comes to customer service matters. Leonard said he wants to change this perception because “human agents today have their own limitations,” like language barriers or making people feel judged. 

“There were great cars before the Model T, but vehicles didn’t become ubiquitous until the assembly line,” Leonard said. “There are great voice agents today, but they won’t be ubiquitous until the voice agent factory is built. VoiceRun is that factory.”