Nandagopal Seshagiri

When Humans—and Hand-Programmed Systems—Can’t Keep Up: AI-Mediated Trust Decisions in Smart Systems

Abstract:

Smart computing systems increasingly operate at scales, speeds, and levels of autonomy that exceed the governing capacity of both human decision-makers and traditional hand-programmed, rule-based systems. While deterministic policies and human oversight have historically anchored trust and control, they are ill-suited for dynamic, adversarial, and highly distributed environments such as autonomous infrastructure, large-scale cloud platforms, IoT ecosystems, and AI-driven services.

This keynote examines why human-centric governance models fail under decision overload, and why static, hand-coded trust mechanisms degrade silently in the face of uncertainty and evolving behavior. It argues that the next generation of smart systems must move beyond binary authorization and rigid policies toward AI-mediated trust decisions—adaptive decision processes that continuously evaluate identity, behavior, context, and risk to produce confidence-weighted outcomes.

The talk introduces a tri-modal trust architecture in which hand-programmed systems provide invariants and guardrails, AI systems infer risk and adapt decisions in real time, and humans intervene selectively when confidence thresholds are breached. Through examples drawn from smart infrastructure, machine-to-machine interactions, and autonomous access control, the keynote highlights both the promise and the challenges of this approach, including explainability, decision confidence modeling, failure containment, and long-term trust sustainability.

By reframing trust as a dynamic, computed property rather than a static assumption, this keynote outlines a research and design agenda for building scalable, resilient, and governable smart systems—systems that not only act intelligently, but also know when not to decide alone.

Profile:

Nandagopal Seshagiri is a Senior Software Architect specializing in large-scale identity, access management, and security architecture at Okta. His work focuses on designing and securing highly distributed, autonomous systems spanning cloud infrastructure, machine identities, cryptographic trust, and AI-assisted decision frameworks.

With deep experience in zero-trust architectures, just-in-time access, secrets and key management, and certificate automation, he bridges theoretical security models with real-world, production-scale systems serving millions of users. His current research interests explore AI-mediated trust decisions, human-AI cooperative governance, and the limitations of deterministic, hand-programmed systems in modern smart environments.

An active contributor to the global research and professional community, he regularly serves as a keynote speaker, technical reviewer, and industry chair for international conferences, and is engaged in advancing secure, scalable, and governable smart computing systems.