How cognitive infrastructure is turning detection into action

Veena Mahesh, General Manager for Cities & Critical Infrastructure at Intel, explains why agentic AI is critical to moving city services from smart systems that perceive conditions to cognitive infrastructure that can reason and act

A tree root breaks into a sewer pipe. Someone has to find it, judge the damage and arrange a repair before it becomes a blockage or a burst. Cities rarely lack data. What they lack is a fast path from detection to action.

This is where I see an immediate role for cognitive technology in urban infrastructure. Cities can get more from existing infrastructure by connecting what it detects with the people who can act.

Today, information often sits within separate systems, making it difficult to coordinate a response. Cognitive infrastructure connects sensing, reasoning and action across those systems, helping cities identify maintenance needs earlier, direct staff where they are needed and respond more effectively to disruption.

Sewer inspections that inform repairs

Consider sewer inspections. A camera mounted on a remotely operated rover inspects conditions inside a pipe, revealing cracks, breaks or roots growing through the structure. Traditionally, someone then reviews the footage, identifies defects and prepares a report. Across miles of underground infrastructure, that creates a substantial workload before repairs can even be prioritised.

DC Water introduced Pipe Sleuth to automate the analysis of inspection footage and assess pipe condition. The system identifies defects and generates reports that help staff locate problems and determine where maintenance is needed. It also produces documentation useful for audits and compliance, reducing another manual task for operational teams.

The wider opportunity is to connect findings and kick off workflows that need to happen or notify someone who can assess its urgency and arrange a response. That is what the shift towards cognitive infrastructure should mean operationally: systems that sense conditions, interpret them and help initiate an appropriate action.

Inspecting streets on everyday journeys

Cities can also find problems through activities already taking place above ground. Buses and refuse trucks travel through neighbourhoods every day, passing potholes, dumped waste and pavements. Cameras and onboard computing can turn those journeys into opportunities to inspect streets. These applications focus on identifying infrastructure problems, allowing operational teams to respond.

One mobile sensing deployment uses wide-angle cameras and rugged computers on buses to identify road and kerbside problems. An obstruction that seems minor from a passing vehicle may prevent someone using a wheelchair from using the pavement. Detecting these problems gives city officials information they can use to investigate and clear it.

The practical advantage is coverage during everyday operations. Vehicles already making their rounds can provide observations that supplement dedicated inspections or residents’ reports.

Governing how AI takes action

Cities can establish the rules for action before deploying these systems: AI flags an issue, a person approves the response, and AI acts within agreed limits. For example, a system might identify conditions that require a valve to close or a pump to stop. The authority to act, and the conditions that require human approval can be defined beforehand.

As agentic AI takes on these responsibilities, security becomes critical. Cities must ensure that only trusted, authorised agents can act on public systems, with clear controls over what each agent is permitted to do. Intel and our ecosystem partners are also working to ensure that critical systems have zero exposure to actors who want to introduce unauthorised activities into these systems.

Responding when networks are disrupted

During an emergency, local processing becomes particularly important. Wildfires and floods can strain communications or leave responders without reliable connectivity. A system that depends on sending all its data to the cloud may struggle precisely when information is most urgently needed.

Aerial systems offer one way to support teams on the ground. Axcelia’s SpecterEye platform uses onboard AI for applications including wildfire detection and first-responder coordination. Drone footage, thermal imagery and cameras worn by responders can contribute different views of the same incident. Processing those feeds locally can help a command team operating from a nearby vehicle assess conditions as they change.

Digital twins for everyday building operations

There are quieter applications inside buildings. Digital twins, virtual representations of physical assets, become more useful when they incorporate current information from cameras, temperature sensors and other equipment. If only one zone of a floor is occupied, why cool the entire floor as though every desk were in use? Operators can use that information to adjust cooling, heating and lighting to actual conditions.

Putting agentic AI to work

For most cities, introducing these capabilities means working with equipment already installed. Industrial controllers and supervisory control and data acquisition systems produce information about field assets. Connecting those systems to predictive maintenance pipelines deployed on edge computing can help identify developing faults while retaining equipment that continues to do its job.

Agentic AI can connect more of the subsequent steps, helping analyse an issue, assemble evidence and initiate a maintenance workflow. But responsibility for the response remains essential. AI is useful at detecting patterns and flagging potential problems; staff bring knowledge of local conditions, city policies and service priorities before kicking of actions in response.

City leaders can start with one maintenance workflow, trace a problem from detection to an approved work order and measure the time saved. They can then use those results to identify where the approach can help other teams, giving the people who maintain our cities better information while there is still time to act. Our goal is to help cities ensure that systems and technologies work harder and smarter and to help them become truly cognitive.

Main image: Wirestock | Dreamstime.com