ITTech Pulse Exclusive Interview with Anurag Gurtu Co-Founder and CEO of Airrived

Anurag Gurtu, Co-Founder & CEO of Airrived, discusses how Agentic AI is enabling autonomous enterprise operations through governance, reasoning, and intelligent execution.

You’ve spent over two decades in cybersecurity and AI, from an early startup acquired by Sun Microsystems through Cisco, FireEye, and Splunk. What pulled you into security, and how did that lead to Airrived?

Cybersecurity has always fascinated me because it’s one of the few industries where the problem changes every single day. Attackers innovate constantly, which forces defenders to do the same. Throughout my career from startups through Cisco, FireEye, Splunk, and several AI companies I kept seeing the same pattern that every new security tool generated more data, but not necessarily more action. The industry became exceptional at detection but struggled with execution. Analysts were drowning in alerts, playbooks, dashboards, and disconnected workflows. That realization led to Airrived. We didn’t want to build another AI assistant that answers questions. We wanted to build an intelligence layer that can understand context, reason across enterprise systems, and autonomously execute work under governance. That’s the gap Agentic AI is designed to solve.

You moved from classical machine learning around 2013 into natural language processing by 2017, and helped define categories like UEBA. How has that shaped your approach to building production-grade enterprise AI?

Enterprise AI isn’t about having the biggest model, rather it’s about solving real operational problems reliably. Coming from machine learning, anomaly detection, UEBA, and later large language models, I learned that intelligence alone isn’t enough. Production systems need memory, deterministic workflows, reasoning, security, governance, observability, and the ability to integrate with hundreds of enterprise systems. That’s why we built Agentic OS rather than another chatbot. Models will continue to improve, but orchestration, governance, and execution are what determine whether AI actually creates enterprise value.

Read More: ITTech Pulse Exclusive Interview with Mike Armistead, Co-Founder and Chief Executive Officer of Pulse Security AI

Airrived was just named a Gartner Emerging Tech AI Vendor in its Agentic AI use-case report, following two Tech Innovator recognitions. What does this validation signal about where the agentic AI market is heading?

It’s a strong signal that the industry is moving beyond AI as an assistant toward AI as an autonomous workforce. For years, organizations focused on generating content and answering questions. The next phase is AI that plans, reasons, collaborates with other agents, takes action, and continuously improves. Recognition from Gartner validates something we’ve believed from day one that enterprises don’t need more copilots, they need intelligent systems that execute work safely, transparently, and at scale.

The Gartner case study cites autonomous handling of 90% of SOC alert triage and an 80% cut in mean time to respond. How does Airrived’s Agentic OS achieve those outcomes in a live SOC?

We don’t automate individual tasks, we automate decision making.Our agents correlate telemetry across the security stack, gather additional evidence, perform threat reasoning, assess business context, determine confidence levels, and execute the appropriate workflow. Multiple specialized agents collaborate while operating under predefined policies and human approval where required. The result is that analysts spend their time investigating the truly complex incidents rather than manually processing thousands of repetitive alerts.

You’ve called much of the market copilots and wrappers dressed up as intelligence, comparing your Agentic OS to VMware’s ESXi. What makes an operating system for agents categorically different from today’s AI copilots?

A copilot waits for instructions but an operating system coordinates intelligence. Just as VMware abstracted physical infrastructure so thousands of virtual machines could run securely on shared hardware, Agentic OS abstracts enterprise complexity so thousands of AI agents can operate securely across applications, data, and workflows. It provides identity, memory, policy enforcement, orchestration, communication, lifecycle management, governance, observability, and execution. The underlying LLM becomes just one component of a much larger autonomous operating environment.

Your flagship AetherClaw centers on autonomous systems governed by design. As enterprises hand more decisions to AI agents, how do you build in policy enforcement, explainability, and human oversight without slowing autonomy down?

Governance cannot be bolted on after deployment, it has to be part of the architecture. Every agent operates within defined policies, permissions, risk thresholds, and approval workflows. Every action is explainable, fully auditable, and reversible where appropriate. Human oversight is adaptive—routine, low-risk tasks remain autonomous, while higher-risk decisions automatically require review. The goal isn’t maximum autonomy. It’s the right autonomy for the level of risk.

Looking ahead to 2027, how do you see the shift from AI-assisted workflows to fully autonomous operations playing out across security and IT, and what will separate the enterprises that get it right?

By 2027, I believe autonomous execution will become the default operating model for many enterprise functions. SOCs, IT operations, identity management, vulnerability management, compliance, and service desks will increasingly be managed by teams of collaborating AI agents supervised by humans rather than driven by humans. The organizations that succeed won’t necessarily have the best models, they’ll have the best governance, clean enterprise data, standardized workflows, and platforms capable of orchestrating thousands of agents safely. The competitive advantage will come from operational architecture, not model selection.

Read More:ITTech Pulse Exclusive Interview with Douglas Gourlay, Chief Executive Officer and President of Qumulo

For the security leaders, founders, and AI practitioners following your work, what’s your one piece of advice on adopting agentic AI responsibly while still moving fast enough to outpace AI-enabled attackers?

Don’t ask whether AI can replace people, rather ask which decisions humans should no longer be making manually. Start with repetitive, high-volume, low-risk workflows where governance is straightforward. Build trust through measurable outcomes, then progressively expand autonomy. The organizations that learn to govern autonomous AI—not just deploy it—will define the next decade of cybersecurity. Attackers are already using AI at machine speed. Defenders can’t afford to respond at human speed alone.

Thank you, Anurag, for taking the time to share your insights with us.

Write to us [⁠wasim.a@demandmediaagency.com] to learn more about our exclusive editorial packages and programmes.