$10M on top: what exactly happened
The Indian startup Ringg extended its Series A round, adding another $10M from Peak XV Partners. The money didn't come as a new round, but as an add-on to an already closed one: earlier this year the company raised $5.5M, and now the total Series A stands at $15.5M.
For the voice agent category, which over the past couple of years has become one of the most overheated, the size of the check isn't the main thing here. Far more interesting is the fact of the top-up itself: an investor who had already seen the inner workings decided to increase the bet rather than dilute it with a next round. That's usually what happens when they see not a pitch deck, but a working product under real load.
And there is load: Ringg handles around 20 million call attempts per month. That's no longer a pilot with a couple hundred conversations.
From speech synthesis to enterprise agents
The company's starting point was completely different. Initially it was a text-to-speech project called DesiVocal: the team was trying to build its own speech models. The endeavor ran into economics — training your own models is so expensive that there's practically nothing to recoup it with at an early stage.
Instead of continuing to bang their heads against the wall, the founders moved up the stack and took on voice AI agents for business. The company's co-founder is Siddharth Tripathi.
The first major client appeared quickly: it was the Indian fintech Cred. The customer list then grew to include Flipkart, Practo, Groww, and Policybazaar — that is, companies whose phone channels have historically been overloaded.
Why India turned out to be a convenient venue in the first place
A cultural factor is at play here. According to a recent Truecaller study, more than three-quarters of Indian consumers prefer to handle matters with businesses over the phone rather than via messaging. Where in the West a call is perceived as an intrusion, in India it remains a normal and expected way to communicate. For voice agents, this is ready-made soil: demand doesn't need to be artificially created, it just needs to be served.
From simple scenarios to complex processes
Ringg's first scenarios were predictably mass-market and not particularly intelligent: outbound calling, lead qualification, loan debt collection. According to Tripathi, such tasks have low "stickiness" — a client easily switches vendors, and competition inevitably slides into a race to the bottom on price.
The company kept part of this line of business, but placed its bet on something else — on processes where a mistake is costly and where value is measured not by the number of calls made, but by a job seen through to completion:
- appointment scheduling for clinics and post-visit follow-up;
- abandoned cart recovery in e-commerce;
- onboarding and KYC checks for fintech apps;
- first- and second-line support for large services.
A telling example is healthcare. Ringg's voice agent operates in 1,200 clinics connected to the Practo app: it helps patients book a visit or receive a reminder and follow-up after an appointment. This is no longer a script with buttons, but a conversation that has to be sustained under real-world conditions — with clarifications, rescheduling, and incomplete data.
Multichannel: voice isn't the only outlet
Despite the category's name, it's not limited to calls alone. Voice conversations still generate more than 70% of Ringg's revenue, but the company is gradually adding other channels — chat and WhatsApp. The logic is simple: the client doesn't care where exactly their issue gets resolved, what matters is that it gets closed.
A separate storyline is support automation through the browser. For some customers, including Shell, the company takes on requests that employees usually fill out manually in web interfaces. Here the agent doesn't talk at all: it navigates through systems and performs actions.
This is precisely what Tripathi calls a shift in self-definition: not "voice agents for enterprises," but a platform of agents that deliver measurable results and see tasks through to completion.
The technological bet: orchestration instead of an in-house stack
Ringg takes a pragmatic stance on models. The company develops its own speech recognition and generation systems, and in the long run would like to own the entire voice stack — including infrastructure and deployment. But right now that's too expensive, so the product works as an orchestration layer: a task is routed to one model or another depending on the scenario, language, and latency requirements.
This is perhaps the main fork in the road for the entire category. Some try to be the model owner, others — the dispatcher between others' models. The second path is cheaper and faster, but requires constant work on margins: inference costs eat into profit just as much as training your own models did at the start.
The investor's position is explained by Rishen Kapoor, a principal at Peak XV. He recalls that Ringg began as a research lab that built its own models — and that technical foundation is visible in the tasks the company now takes on. According to him, this is about fully running corporate processes from start to finish: merchant onboarding, first- and second-line support — with predictable quality and repeatability.
The market: crowded in India
Voice AI in India is a direction almost everyone with the resources has entered. Model makers Deepgram, ElevenLabs, Cartesia operate here, as well as local players Sarvam and Smallest.ai. A separate layer is startups built around orchestration, such as Bolna and Blue Machines. There are also industry players like Gnani and Arrowhead, focused on finance.
In such density, the winner isn't the one with the louder model, but the one who gets a scenario to the "works without a human" state faster. Ringg is trying to establish itself precisely in complex processes, where the barrier to entry is higher and the cost of a mistake is more visible.
What's next: people, hiring, and geography
Ringg currently has 40 employees, with more than 15 joining in the last three months — a fairly aggressive hiring pace for a company of this size. They're mainly recruiting for two roles:
- forward-deployed engineers, who need both technical skills and the ability to drive a product at a client's site;
- researchers, working on reducing the cost of running models.
Geographically, most customers are in India, with some in the Middle East and the US. The company isn't building direct sales to American companies. Instead, it's betting on global capability centers — offshore hubs that multinational corporations are increasingly using for back-office and support. The idea is to sell automation as a complement to human support, not a replacement for it: that makes it easier to fit into existing budgets.
One thing is worth keeping in mind. The shift from outbound calling to clinic scheduling and KYC also means a change in metrics: instead of cost per minute of conversation, you'll have to prove the share of successfully completed tasks and the reduction in load on live operators. This is exactly the field where Ringg and its numerous competitors will clash — and this is exactly where the new tranche from Peak XV will have to be earned.
