The next time your phone rings with an unfamiliar number asking about your shopping habits or political views, there's a decent chance the voice on the other end won't be human.
Miravoice, a San Francisco startup building AI voice agents for phone surveys, has raised $6.3 million in seed funding, according to Crunchbase News and company press releases. Unusual Ventures led the round, joined by Neo, 25madison, and a handful of angel investors from Ramp, PubMatic, Atlassian, and Google.
It's a bet on a simple premise: that artificial intelligence can do what call centers have done for decades, only faster, cheaper, and at vastly greater scale. Miravoice claims its platform cuts costs by 70 to 90 percent compared to traditional survey operations, though no independent validation of those figures is available.
For an industry long dependent on human callers navigating clunky scripts and low response rates, that's no small claim.
The Pitch to Researchers
Survey research is, by nature, labor-intensive. Running a phone survey of even modest size typically requires coordinating teams of interviewers, managing call schedules across time zones, and troubleshooting in real time when respondents pause mid-question or ask for clarification. It's expensive. It's slow. And for many organizations—universities, political pollsters, market research firms—it's essential.
Miravoice thinks it can compress that entire workflow into hours rather than weeks.
The company's AI interviewer is designed to handle structured, long-form surveys. Some run more than 120 questions and stretch past 40 minutes, a marathon by phone survey standards. The system supports 14 languages and plugs into existing survey platforms like Qualtrics, Forsta Decipher, and Voxco. Researchers can build a questionnaire, spin up a phone number, and launch a campaign without assembling a team of live callers.
The platform uses usage-based billing—customers pay only for the time the AI is on the phone. Miravoice says the system can handle interruptions, manage pauses, and respond to respondent questions, all while adhering to strict survey logic intended to prevent the kind of hallucinations that have plagued other generative AI applications.
Whether it can do all that reliably at scale remains an open question. But investors seem willing to find out.
Who's Behind It
VKL Research, Inc., doing business as Miravoice, was founded by Nishant Jain (CEO), Danny D. Leybzon (CTO), and Shreyas Tirumala (COO and president). The trio previously operated under the name Surveybot AI before settling on the current brand. The company is formally registered as VKL Research, Inc., with California filings dating to March 2025.
The startup says it completed more than 100,000 calls in 2025 and currently works with 10 to 20 customers, a mix that includes public opinion organizations, market research firms, university departments, and private companies in retail, entertainment, and logistics. Some are still in paid pilot phases; others have moved into full production.
Lars Albright, a general partner at Unusual Ventures who led the investment, pointed to the startup's emphasis on precision data collection as a key draw. Named angel investors include Karim Atiyeh, CTO and co-founder of Ramp; Rajeev Goel, CEO of PubMatic; and Amar Goel, CEO of Bito AI, along with executives from Atlassian, Crusoe, and Google. The round was first reported by Crunchbase News.
What Comes Next
The fresh capital will go toward expanding the engineering team and scaling the platform to handle tens of millions of calls. Miravoice is actively recruiting founding engineers in Palo Alto, a signal that the company is still in build mode.
"The goal is to make talking to 100,000 people as simple as creating a Google Form," Jain told Crunchbase News—a soundbite that captures both the ambition and the challenge ahead.
The startup has begun collaborating with established research organizations, presenting findings alongside NORC at the University of Chicago and SSRS at industry conferences including MAPOR 2025 and AAPOR 2025. Those sessions have explored how AI voice agents perform when deployed in probability-based survey panels, a rigorous standard in the research world.
Still, questions linger. How do respondents react when they realize they're speaking to a machine? Does the AI introduce subtle biases that human interviewers might avoid—or does it eliminate biases that humans bring? And perhaps most practically: can the technology handle the sheer unpredictability of real-world phone conversations at the scale Miravoice envisions?
For now, the company is placing its bet that researchers—facing tighter budgets and an increasingly difficult-to-reach public—will be willing to trade the human touch for speed and cost savings. Whether that trade-off feels acceptable to respondents themselves is another matter entirely.
