Nicholas Leonard and Derek Caneja wished to construct AI voice brokers, however after they went to construct the product, they felt many of those voice brokers had design flaws.
A few of these brokers have been being constructed with no-code instruments, which means transport to manufacturing was quick, however the high quality of the product was usually low. Different brokers have been being made by firms that had the time and sources to spend months constructing specialised instruments. “Builders and enterprises wanted another,” Leonard informed TechCrunch, including that he and Caneja additionally realized that the way forward for software program could be “coded, validated, and optimized by coding brokers.”
“These two insights and a historic realization gave us the inspiration for VoiceRun,” Leonard, the corporate’s CEO, mentioned. Caneja is the corporate’s CTO.
Final 12 months, they determined to launch VoiceRun, a platform that lets builders and coding assistants launch and scale voice brokers. Proper now, many of those low-code platforms let individuals construct voice brokers with visible diagrams, the place individuals click on via dialog flows and write prompts into packing containers that then dictate how the agent ought to behave. All of that may be exhausting to handle, Leonard mentioned.
VoiceRun, however, lets customers code how they need their voice brokers to behave, giving them extra flexibility in creating the product they need. Code is the native language of coding brokers, Leonard defined. “They’ll do a much better job working in code than in a visible interface,” Leonard mentioned.
Moreover, with visuals, there are restricted configuration choices, so, for instance, if somebody wished to construct a voice agent that would converse in a unique dialect, it may be more durable to do if the maker of the visible interface didn’t construct a characteristic that may deal with that activity.
“However in code, it’s extremely easy to do,” he mentioned. “There’s a lengthy tail of tens of millions of examples of little belongings you may need to try this aren’t supported by the visible interface.”
Techcrunch occasion
San Francisco
|
October 13-15, 2026
Except for coding brokers, VoiceRun additionally lets customers carry out A/B testing and deploy immediately with one click on.
The corporate is geared towards enterprise builders, serving to firms, for instance, incorporate AI into their buyer providers, or assist tech firms launch voice-based merchandise. He talked about, for instance, working with a restaurant-tech firm launching an AI cellphone concierge for meals reservations.
The corporate introduced on Wednesday the closing of a $5.5 million seed spherical led by Flybridge Capital.
There may be plenty of competitors within the AI agent area. Startups on this space final 12 months nabbed billions of {dollars} (out of the various billions that flooded into AI firms typically). Leonard feels his firm is up in opposition to two ends of the market: There are the no-code voice builders, like Bland and ReTell AI, he mentioned, that lets consumer construct fast demos. There are additionally extra refined instruments, like LiveKt and Pipecat, which give builders “most management.” He feels Voicerun sits in the midst of these two ends.
“We offer international voice infrastructure and an evaluation-driven lifecycle, whereas maintaining possession of enterprise logic code and knowledge within the buyer’s palms,” he mentioned. “The important thing distinction is that we’re closing the loop for end-to-end coding agent improvement. We count on builders to be supervising coding brokers that write code, run checks, deploy, and suggest enhancements.”
In some methods, Leonard is hoping his product helps builders create voice agent instruments that can, in flip, assist individuals really feel extra comfy with automated voices. Clients as we speak “really feel reduction” when a human solutions the cellphone, “as a result of voice automation has been brittle and ineffective.”
A survey from Five9 last year showed that three-fourths of its survey respondents nonetheless desire speaking to a human in terms of customer support issues. Leonard mentioned he desires to alter this notion as a result of “human brokers as we speak have their very own limitations,” like language obstacles or making individuals really feel judged.
“There have been nice vehicles earlier than the Mannequin T, however automobiles didn’t develop into ubiquitous till the meeting line,” Leonard mentioned. “There are nice voice brokers as we speak, however they received’t be ubiquitous till the voice agent manufacturing facility is constructed. VoiceRun is that manufacturing facility.”
