"The future of AI is private,” says KOGO AI CEO as firms seek data control

As companies race to deploy artificial intelligence (AI) agents across core operations, an Indian startup is pitching a contrarian thesis: the future of enterprise AI won’t live in the cloud at all.

KOGO AI, whose customers include Tech Mahindra, Michelin and the Indian Army, says large organisations increasingly want AI systems that never leave their own infrastructure, even private clouds, amid rising concerns over data leaks, intellectual property exposure and regulatory scrutiny.

“The future of AI is private,” Raj K Gopalakrishnan, chief executive officer and co-founder of KOGO AI, told Business Today in an interview. “We don’t bring data to AI. We take AI to where the data is.” 

The company’s core product, KOGO OS, is designed as a “private agentic operating system” that replaces multiple layers of AI tooling, from model management to deployment and monitoring, while running entirely inside the end customer’s ecosystem.

Unlike software-as-a-service platforms that process data on vendor-controlled infrastructure, KOGO installs its stack directly on enterprise systems. “It’s not a SaaS-based platform,” Gopalakrishnan said. “The platform in its entirety goes and sits inside a customer’s infrastructure.” 

Earlier in April 2025, the Bengaluru-based startup partnered with chipmaker Qualcomm to develop a full-stack private AI solution for enterprise deployments. 

Sovereignty over scale

The approach is finding buyers in sectors like banking, defence and healthcare, where strict compliance rules govern data use.

KOGO said it is growing revenue by roughly 600% to 700% year-on-year, fuelled by deployments in these sectors. 

At the ongoing India AI Impact Summit in New Delhi, the company showcased installations spanning private data centres, edge systems and air-gapped networks, environments where internet connectivity is restricted or prohibited.

The company also launched CommandCore, a hardware-software bundle built with Arinox AI, a security-focused AI platform builder, on Nvidia infrastructure that allows organisations to run AI agents entirely offline. The system targets defence agencies, government bodies and critical infrastructure operators that “must operate without internet connectivity, public cloud access, or external data exchange.” 

“Sovereign AI requires sovereign infrastructure and a sovereign operating system working together,” Gopalakrishnan said in a statement following the launch. 

Horizontal agent factory

KOGO’s platform acts as an orchestration layer over hundreds of AI models rather than building its own foundation model. Enterprises can integrate open-source or proprietary systems and create autonomous agents tailored to internal workflows.

“You don’t train agents,” Gopalakrishnan told Business Today. “Models are trained. We are the orchestration layer on top of models.” 

The approach allows companies across sectors, from supply chains to finance to HR, to build specialised systems without exposing proprietary data externally.

More than 25,000 agents have already been created on the platform by customers, he said. 

Gopalakrishnan said that KOGO targets large enterprises rather than smaller businesses or consumers. “We are a large enterprise deep tech,” he said. “We are not retail. We are not SME.” 

Privacy as a business metric

While much of the AI industry emphasises model accuracy or benchmark scores, KOGO frames success in financial terms, time saved, cost reductions and total cost of ownership.

“Business metric is the only metric that matters,” Gopalakrishnan said. 

The company also argues that keeping data in-house reduces one of the biggest reasons enterprise AI projects fail, which is privacy breaches during implementation. 

“What differentiates a 100-year-old company from a 2-year-old company? It’s their experience of those years,” he said. “That is their intelligence… You cannot have that intelligence breached.” 

Security arms race

Agentic AI, systems capable of making decisions and carrying out tasks, have raised concerns about manipulation attacks such as prompt injection.

KOGO claims to have embedded “red teaming” tools directly into its platform to simulate adversarial attacks against deployed agents, a capability Gopalakrishnan says is rare among commercial systems.

“We are the only agentic platform that has a red teaming module built into the platform,” he said, describing automated attempts to “attack, hack and break” AI systems to test resilience. 

Betting on a private AI future

Gopalakrishnan believes concerns over surveillance, data leakage and compliance will push both organisations and individuals toward localised intelligence.

“The future of AI is private,” he said, adding that conversational systems accumulate context that users may not fully realise they are sharing.