Rime has raised $24 million in Series A funding as it pushes further into one of voice AI’s most competitive opportunities: helping large organizations handle phone calls.
The San Francisco-based startup is entering a market where enterprises are already testing or adopting systems from voice model developers, infrastructure companies, and customer support-focused AI providers. Rime’s bet is that better data, better pronunciation, and lower-latency speech models can give it room to stand out.
Why enterprise calls remain hard for voice AI
Voice AI startups have found a major opening in sales, marketing, and customer support calls. Large organizations want to reduce the burden of handling high volumes of phone interactions, and the category has attracted companies including ElevenLabs, Deepgram, Vapi, Retell, LiveKit, Decagon, and Sierra.
But the opportunity is not the same as an easy replacement cycle. According to Rime co-founder Lily Clifford, enterprises still often prefer legacy IVR implementations because AI voice technology has not yet matched IVR’s effectiveness for many phone workflows.
Her view is that large language models have made it easier to build working voice applications, but they have not fully changed the end-user experience. In plain terms, a voice AI system can sound better than an old phone tree and still feel awkward if it pauses at the wrong time, misunderstands background noise, or fails to respond naturally in a live conversation.
That gap matters because phone calls are unforgiving. Unlike text-based support, callers expect immediate timing, clear turn-taking, and accurate recognition of what they are saying. A small delay or an unnatural handoff can make the interaction feel mechanical.
Rime’s data strategy starts in its own studio
Rime was founded in 2022 by former Stanford PhD student Lily Clifford, ex-Amazon Alexa engineer Brooke Larson, and Stanford engineer Ares Geovanos. Instead of relying on scraping the web for audio, the company has built a recording studio in San Francisco to collect its own conversational data.
That choice is central to how Rime describes its technical edge. The company says its voice AI models are trained on conversational data it records, with the goal of reducing the amount of customization clients need to do before deploying the technology.
For enterprises, customization can become a serious obstacle. A customer support system may need to say brand names, product names, medical terms, financial terms, airline-related language, or food service terminology correctly. If the model struggles with those terms, the call can lose credibility quickly.
Rime says it focuses on tuning its voice models to handle the pronunciation of brand entities and industry-specific terms. It uses a phoneme-based architecture designed to adapt to different pronunciations, so customers do not have to retrain models for every industry-specific use case.
The shift from model pipelines to speech-to-speech
Rime did not begin with a single end-to-end approach. The startup initially used a pipeline made of separate models for speech-to-text, text-to-speech, and a large language model.
That structure reflects a common pattern in voice AI applications: one system listens, another interprets or generates a response, and another speaks it back to the caller. The drawback is that every handoff can introduce complexity.
Rime is now shifting its focus toward better speech-to-speech models. The company says the goal is to reduce latency, improve turn-taking, and address problems such as background noise.
The change also reduces reliance on orchestration. Rather than managing a collection of models that must coordinate smoothly during a call, Rime wants a more direct voice model approach that can support a faster and more reliable interaction.
That matters in the enterprise market because the experience is judged in real time. A voice agent that responds too slowly, interrupts, or struggles when conditions are imperfect can make automation feel like a weaker version of existing phone systems.
Funding, customers, and hiring plans
The $24 million Series A was led by M13 Ventures. Twilio Ventures, Corazon Capital, Unusual Ventures, and other existing investors also participated.
Rime says it has customers in food service, healthcare, airlines, and fintech. The company also says its training data and model positioning help customers stay longer on calls, which has supported enterprise contracts with clients including Mayo Clinic, Dialpad, Upstart, and Asurion.
The new funding will go toward expanding a team of 35 people. Rime plans to hire across model development, engineering, and partnerships.
The company also recently brought on Rafael Valle as chief scientist. Valle worked on audio understanding at Meta Superintelligence Labs and Nvidia’s applied deep learning audio research team.
As part of the fundraise, M13’s Morgan Blumberg is joining Rime’s board. Blumberg told TechCrunch that Rime stands out because of its focus on a model with low latency and high reliability in a regulated environment.
Rime had previously raised $5.5 million in a seed round last May. The Series A gives the company more capital to pursue a technical path that is focused less on being another application layer and more on improving the underlying voice experience for enterprise calls.
What Rime’s raise says about the market
The broader signal is clear: enterprise voice AI is crowded, but investors and customers are still looking for systems that can handle calls more naturally and reliably.
Rime is positioning itself around several linked claims. It records its own conversational training data. It tunes for brand and industry pronunciation. It is moving toward speech-to-speech models. And it is targeting enterprise environments where reliability, latency, and call quality matter.
Those are practical concerns, not cosmetic ones. If voice AI is going to replace or meaningfully improve enterprise phone workflows, it has to do more than sound polished. It has to understand callers, respond at the right moment, and handle the specific language of each business without forcing every customer into heavy retraining.
Rime’s $24 million round does not settle whether its approach will win in a market filled with strong competitors. It does show that the next phase of enterprise voice AI is likely to be judged by the details of the call itself: timing, pronunciation, reliability, and whether the person on the other end wants to keep talking.