The Pothole Is Now a Data Point. That’s Where AI Is Earning Its Keep in Cities.

By Paul Wilson

Key takeaways:

• Municipalities are using AI to assess everyday assets, from roads and footpaths to ash trees and vacant storefronts.
• Once mapped, an asset can be linked to neighboring datasets—planning, amenities, tree health—turning periodic inspection into prioritization and, increasingly, prediction.
• AI governance has moved from an afterthought to a design requirement, as cities decide what AI should do, who’s accountable, and how to comply with privacy laws.

The steady flow of news about AI’s advances makes it easy to feel excited, or anxious, about the future. Less dramatic, and a lot more practical, is AI’s role in helping cities manage everything from potholes to tree cover, the systems that make cities thrive.

Dublin is a great example. The city uses digital models, open data, and AI to manage aspects of its fire brigade’s work, university campuses, city planning, and other municipal tasks. A collaboration with Trinity College Dublin trials new uses of AI, and the city is keen to do more—including helping other municipalities tap into the technology.

Dublin recently welcomed about 80 digital leaders from local governments across Ireland and the U.K. for an AI CityXchange workshop. The event was part of an international series that helps local government leaders, practitioners, and industry experts move from AI curiosity and experimentation to practical, scalable plans.

The workshop, held at the city’s Civic Offices in September, was hosted by Smart Dublin, a regional initiative to find innovative solutions for urban challenges, and SmartCitiesWorld, with support from Microsoft. (SmartCitiesWorld is the global media partner of the Smart City Expo World Congress, which takes place this week in Barcelona. Bentley Systems and Bentley Labs will be there.)

The event showcased successful ways local governments are now using AI and other technologies to help build, maintain, and improve their cityscapes. Here are some highlights:

Intelligent roads

Dún Laoghaire-Rathdown, the county council area just south of Dublin City, has about 500 miles of roads, 850 miles of footpaths, and 60 miles of trails, all overseen by just nine engineers. Rob Mulhall, a senior executive engineer with the council, said during the workshop that “prioritization is paramount given budgets are tight.” He noted that the county could only afford to repair or replace “somewhere between 1% and 2% of the total network each year.”

The region’s traditional approach to assessing road conditions was slow and labor-intensive. A couple of people drove around with clipboards, scoring 50-meter stretches and comparing notes. The survey took six to eight months and happened only every few years, so it could quickly become outdated. Severe weather alone could change a road’s condition in a few days.

Footpaths were beyond the reach of the two-person team. Tree roots pushing up pavement went unrecorded, creating trip hazards that cost the council €1.5 million to €2 million a year in injury claims.

To help, the council decided to experiment with AI-enabled computer vision. Vehicle-mounted cameras captured road imagery, and on-board laptops geolocated, identified, and categorized potential problems. The approach worked, enabling the council to map defects, identify clusters, and prioritize repairs. The cameras also captured signage, road markings, gullies, and benches, and footpaths were assessed for the first time.

The result was not only a better picture of the roads, but a new relationship with the network. Instead of wondering when a road or path was last inspected, the engineering team now asks, “What is changing?” They no longer need to wait for a complaint because the network monitors deterioration and is beginning to model what will happen next. “Predictive maintenance is moving from something others do to something we can do,” Mulhall said.

(Bentley’s Blyncsy platform uses AI to process anonymized images from dashboard cameras and monitor road conditions in the U.S., Europe, and Africa.)

Trees are infrastructure too

Computer vision and AI also help manage trees. This “green infrastructure” stores carbon, cools streets, and holds soil and water in place, reducing flooding.

Kieran Kelly and his team at engineering consultancy Mott MacDonald used vehicle-mounted stereo cameras and machine learning to catalog more than 13,000 ash trees for Glasgow City Council. They then used the data to identify diseased trees.

Ash trees across Europe are under attack from ash dieback, an airborne fungal disease. In Ireland, more than 90% of ash trees are expected to be lost. Kelly said an infected tree typically survives less than a decade, shedding 10% to 15% of its branches each year. Because ash trees line roads and rail lines, falling branches are a real hazard.

Tracking each tree’s species, location, volume, health, and proximity to other ash trees lets AI help in new ways. For example, proximity helps prioritize removals, and volume estimates allow accurate forecasts of timber supply. The latter helps local furniture makers and wood-chip biomass suppliers know how much timber is available, where and when, helping to build a stronger local timber economy. 

Next, the empty shop

Empty shops are the scourge of any town’s Main Street, and vacancy can be contagious: One empty shop seems to lead to another. Tim O’Driscoll, of the Tipperary County Council, showed how AI can help stop the spread.

His council’s Vacant Business Reopen Grant platform uses geospatial data to understand vacant commercial properties and their surroundings. It checks boundaries, identifies nearby amenities and planning information, and connects property owners and prospective occupiers with grant programs. It also uses generative AI to help draft reports and business proposals.

A vacant building’s location unlocks a world of adjacent information. Connect those datasets, rank and visualize them on a heat map, and use AI to write a report—and suddenly an empty property comes alive with information. Tipperary’s platform has mapped more than 320 properties and substantially reduced manual assessment and validation time, O’Driscoll said. The platform has also increased the number of grant applications the council can process.

Location is key to interoperability

Whether it’s roads or trees or vacant shops, geolocation was the key to unlocking additional information in each project and allowing digital intelligence to emerge. Linking open data through APIs was once heralded as the answer, but that proved too arduous for many cities. Now protocols—such as the Model Context Protocol (MCP), which lets AI agents plug into tools and datasets, and the companion Agent2Agent (A2A) protocol, which lets agents work with one another—are reducing the effort. Using MCP, an AI agent that knows a shop’s location can pull in its size, likely rent, nearby planning applications, and other information, and then turn the results into a report.

From location, location to agentic action stations

Michael Casement, an account technology strategist at Microsoft, laid out his vision for what he called The Frontier Council, describing how agentic AI could change the way councils and local governments operate. He sees the organizational model emerging in four stages. First, every employee has an AI assistant. Second, agents become digital colleagues, taking on specific tasks at the direction of humans. Third, humans set the direction while agents run processes and workflows, checking in when necessary. The fourth, yet to come, is physical AI acting in the real world.

Casement noted that governance questions are moving up the agenda. What should humans do? What should AI agents do? Where should the boundary be? And who is responsible and accountable?

Governance by design

A highlight of the workshop was a panel discussion addressing some of those questions. Jamie Cudden of Dublin City Council, Tommy Kavanagh of South Dublin County Council, and Theresa Cloonan of Dún Laoghaire-Rathdown County Council discussed how their councils govern AI.

The backdrop of the discussion was the European Union’s AI Act. The law’s transparency rules—which require telling people when they are dealing with an AI system and labeling AI-generated content—took effect in August. Stricter obligations for high-risk systems, covering risk management, human oversight, data quality, and accountability, were recently pushed back to December 2027.

All three panelists described how governance has become core to the design, architecture, procurement, and engineering of IT. The room listened closely, and the discussion continued in small groups.

Overall, the Dublin workshop offered practical, pragmatic, and positive examples of AI at work in Ireland’s cities and counties. It also showed a sharper focus on governance, transparency, and caution that cities across Europe could look to as they explore ways to tap into the benefits of AI.

Paul Wilson is chair of the SmartCitiesWorld Advisory Board. He helped moderate the Dublin workshop. He is also a senior advisor at Open Agile Smart Cities and an honorary research fellow at the University of Bristol’s Smart Internet Lab. He was previously chief business officer of Connected Places Catapult, the U.K.’s innovation accelerator for cities, transport, and place leadership.