The ‘Zero Trust Exchange’ may be the backbone of the multi-billion-dollar cybersecurity company Zscaler, but at its flagship Zenith 2025 summit, another theme shared the spotlight: agentic AI. The company, which processes over 500 billion transactions daily, positioned the next phase of the AI revolution as not just about speed but smarter, autonomous systems that can rewire how businesses work. Moreover, India has emerged as a crucial market in this evolving landscape.
Zscaler, known for its cloud-native products and zero-trust security systems, has a growing presence in India. It supports clients such as Wipro, HDFC, Edelweiss, and other top banks and IT service providers.
Jay Chaudhry, co-founder, CEO and chairman, Zscaler at the keynote speech of Zenith Live 2025. Source: AIM
“India is now our fifth-largest market by sales, up from 13th or 14th place just six years ago,” co-founder and CEO of Zscaler, Jay Chaudhry, told AIM at the recently concluded Zenith 2025.
“India has adopted a lot of good technology. It’s refreshing to see how fast things move. A lot of bureaucracy goes away, and a lot of corruption goes away. So we are investing heavily in India for our development of back office functions as well as sales,” Chaudhry added.
When asked about India’s unique customer behaviour, he called Indian customers “tough negotiators”.
While Indian customers play an important role in sales, technical prowess has been their ever-evolving AI ecosystem.
Agentic Trust
“Generative AI has tremendous applications for us to leverage in our products. But there’s also an entire domain around how we help secure our customers’ adoption of generative AI,” Adam Geller, chief product officer at Zscaler, said in an exclusive interaction with AIM.
Zscaler continues to leverage AI, expanding into both agentic and generative AI use cases. With agents acting as digital proxies, either as digital twins of users or autonomous entities, Geller emphasised the urgent need for frameworks that can establish identity and trust for agents. He describes it as a fundamentally new interface that challenges traditional security paradigms.
Geller also explained how Zscaler is approaching generative AI across two key fronts: helping customers securely adopt third-party GenAI tools, and protecting custom-built AI applications from threats like prompt injection, model poisoning, and misuse.
“If all of your employees are using a generative AI application to get the latest cooking recipe, that might not be a concern,” he said. “But if they’re using it to send a code, that should be a very big concern. [Especially] if you’re using a public AI application that likely has policies that say, ‘Once I have this data, it’s mine, I can use it for training, and there’s no way to get that back.’”
Internally, Zscaler is using generative AI to improve how it identifies sensitive data, by understanding context the way a human would, and to build copilots that help users navigate its tools more easily. However, Geller highlighted that the effectiveness of such tools hinges on how well users are trained to prompt and interpret AI responses.
From Predictive to Generative AI to Distributed AI Agents
“Remember, we process 500 billion events per day. There was no way we would be able to do that if we did not have AI before,” Claudionor Coelho Jr., chief AI officer at Zscaler, said in a conversation with AIM. “As Jay said, we are an AI-first company, and we have AI embedded into the products from the very beginning.”
Coelho highlights that in Zscaler, previous AI applications were largely predictive AI. “As we are moving more into generative AI and LLMs, we are seeing more and more people using LLMs to do classification,” he said.
However, he has also been critical of LLM’s capabilities and the risk associated with agents.
A significant concern Coelho raised was the shift from LLMs to distributed AI agents, which introduces a new and less understood layer of risk. “People are starting to send confidential data to an agent endpoint, but that agent may be communicating with other agents. At that point, you’ve completely lost control of where your data goes,” Coelho warned.
He believes that this evolution means organisations can no longer treat AI as a simple tool, and urged them to rethink data governance, trust boundaries, and security protocols in a world where agents act independently across systems.
Based on the interactions and insights shared by Zscaler’s leadership, agentic AI will only get bigger, forming a cornerstone of cybersecurity. However, it is yet to be seen how it will navigate the growing threat of rogue AI agents.
