Fresh off raising over $60 million in a mix of equity and debt from Globespan Capital and EvolutionX, Gupshup is charging ahead to make conversational AI not just an interface, but the new internet.
For Beerud Sheth, founder and CEO of Gupshup, the agentic AI wave is not a fleeting trend—it’s the moment he’s been preparing for over two decades.
“We’re at an inflection point where AI Agents are transitioning from experimental technology to business-critical infrastructure,” Sheth told AIM. The investment, announced in July, will help Gupshup expand its platform across India, the Middle East, Latin America and Africa.
Unlike many AI startups that have sprung up overnight, Gupshup’s edge lies in what Sheth describes as its “combination of incumbency and innovation”. With over 50,000 clients across 130 countries, Gupshup has a moat that’s hard to replicate: deep enterprise integrations, pre-cleared security and legal audits, and a messaging infrastructure that already handles 120 billion messages annually.
“If a new AI startup wants to sell to a large Indian bank, it could take a year. We can do it in two weeks,” he said.
The $1.4 billion unicorn’s roots in SMS and WhatsApp messaging might seem far from today’s world of GenAI-powered agents. But Sheth disagrees. “Messaging is the foundation for conversational AI. You can’t talk to an AI agent without a messaging layer—text or voice,” he said.
That’s the infrastructure Gupshup has been perfecting since its inception.
The Biggest Asset
In a world where foundation models like GPT-4 and Claude dominate headlines, Sheth believes the real value lies in the application layer.
“AI has made product building easier, but it hasn’t made go-to-market any easier. Sales cycles are still tough,” he said. That’s where Gupshup’s existing relationships with enterprises become its biggest asset.
However, staying relevant in AI also means constant evolution. Gupshup’s ACE LLM isn’t a proprietary model, but a fine-tuned gateway that connects to multiple LLMs—open-source, proprietary and global.
For industry-specific use cases like banking, Gupshup fine-tunes existing models and uses multi-agent orchestration.
So, where does that leave Indian AI startups that are building their own agents or LLMs? While Sheth doesn’t think they’re doing anything “wrong”, he believes the lack of scale is the real issue.
“Big funding comes from big revenues, and right now, all the adoption is with foundation models. Indian companies are in the application layer, which is harder to scale because it requires deep customisation,” he said.
Moreover, when it comes to Indic languages, the reality is even tougher. Most enterprises and early adopters still operate in English. “Indic AI is still early. You can maybe boost campaign conversions by translating marketing messages into local languages, but it is voice that really holds promise for Indic, because people don’t like typing in their own language,” Sheth pointed out.
This challenge hasn’t deterred Gupshup from pushing Indic use cases. Through initiatives like CitizenLink, Gupshup is working with government agencies, from Tamil Nadu’s e-Governance Agency to the Ayushman Bharat Health Account and the Election Commission of India, bringing AI agents to citizen services via WhatsApp.
“What happens on websites in the US should happen on WhatsApp in India,” Sheth said. His vision is simple: to allow any Indian, regardless of digital or language literacy, to talk to the government as easily as chatting with a friend.
The effort isn’t purely commercial. “In entrepreneurship, you never know when the draw happens. You just carry the lottery ticket in your pocket,” he added. The scale may be missing today, but Sheth believes the groundwork will pay off as AI agents become ubiquitous.
Beyond All of AI
Gupshup’s ambitions extend beyond just government partnerships. The platform offers everything from click-to-chat ads to AI campaign copilots—tools that automate customer interactions from discovery to support.
“Our stack is built for scale—10,000 TPS during peak loads, with dynamic routing, fast lanes for OTPs and AI orchestration for replies. It’s a high-performance backbone,” Sheth said, describing a messaging engine engineered more like a traffic control system than a chatbot.
While Indian startups debate whether to build new foundation models, Sheth believes the IndiaAI Mission should take a dual-track approach. “I think it’s good to invest in both layers. Foundation models are expensive, but open-source models or incremental training can reduce the cost. DeepSeek already proved that,” he said.
As for Gupshup, the focus will remain on building AI agents that actually get used.
“The most exciting thing about AI agents is that they make computers behave like humans. Earlier, humans had to behave like computers—clicking buttons, opening browsers. But now, even someone in a remote village who can’t read or write can talk to an AI and get access to services,” Sheth said. “That’s transformative—for society, for the economy and for every individual.”
Sheth remains bullish on the future and on India’s role in shaping it. “This is the ultimate goal of technology—making lives better. And I think there’s nothing bigger than what AI agents can do,” he concluded.
The post Everyone is Gupshup-ing appeared first on Analytics India Magazine.
