Smart cities are increasingly becoming distributed digital systems.

Transportation networks, public services, utilities, emergency-response systems, IoT infrastructure, digital twins, data platforms, and citizen-facing applications are producing continuously changing streams of information. Artificial Intelligence is already being introduced across these environments for prediction, optimization, monitoring, and decision support.

The next evolution is Agentic AI: intelligent systems capable not only of analyzing information, but also of invoking APIs, coordinating workflows, interacting with digital infrastructure, and taking actions on behalf of people and organizations.

That shift introduces a fundamental architectural challenge:

How can cities allow AI systems to act across critical infrastructure without sacrificing governance, security, resilience, transparency, or human accountability?

This session presents a practical, vendor-neutral architecture for introducing governed Agentic AI into smart-city environments.

The session introduces the concept of a Smart City AI Control Plane—an architectural layer that connects AI models and autonomous agents with city data platforms, APIs, digital twins, IoT systems, enterprise services, identity, policy, observability, and human governance.

Participants will explore architectural patterns for:

connecting AI agents to smart-city APIs, IoT platforms, digital twins, data spaces, and existing enterprise systems;
establishing machine identity, delegated authority, least-privilege access, and policy enforcement for autonomous systems;
maintaining semantic consistency as agents interact with data produced by different agencies, systems, vendors, and infrastructure platforms;
observing and tracing agent decisions, tool calls, workflows, and actions across distributed urban systems;
introducing human approval and escalation controls for sensitive or high-impact decisions;
designing resilient agentic workflows that account for unreliable models, unavailable APIs, sensor anomalies, network failures, and downstream infrastructure outages;
evaluating autonomous-agent behavior before granting access to operational systems; and
incrementally introducing AI capabilities without replacing existing smart-city or municipal technology platforms.

A practical reference architecture will demonstrate how AI agents, digital twins, IoT infrastructure, urban data platforms, APIs, identity, cybersecurity, policy, observability, and human oversight can function as one governed system.

The session will also examine an important evolution in smart-city architecture: APIs and digital infrastructure will increasingly be accessed not only by people and applications, but by autonomous digital actors operating on behalf of transportation systems, utilities, government services, infrastructure operators, and citizens.

This changes the requirements for identity, authorization, interoperability, observability, cybersecurity, operational resilience, and accountability.

Rather than advocating a particular AI platform or model, the session focuses on architectural principles that can be applied across technology vendors and city environments.

Attendees will leave with a reusable framework for evaluating how Agentic AI can be safely introduced into smart-city ecosystems while maintaining the reliability and public trust required of critical digital infrastructure.

The objective is to help technology and public-sector leaders move beyond:

“Where can we use AI in a smart city?”

toward the more important systems question:

“How do we allow intelligent systems to act across city infrastructure while keeping those actions secure, governed, observable, resilient, and accountable?”