Smart Cities Technology and Data

A new open-source model published last month from the International Institute for Applied Systems Analysis allows city planners to continuously track how urban vegetation is distributed and changes over time.

Using satellite imagery and machine learning trained on data from cities around the world, the model found that the Green View Index —  a street-based indicator of canopy coverage — has been decreasing 0.3% to 0.5% per year. It also found high heterogeneity across cities and regions and “stark inequalities” in street green space within cities.

According to Giacomo Falchetta, an IIASA research scholar who led the study, the model “can shed light on the more structural progress or decline in canopy cover due to processes such as land use change or vegetation health status, and hence on the changing quantity and quality of street greenery in a city.” 

Urban vegetation mitigates heat islands through shade and evaporative cooling, provides flood control and carbon sequestration, preserves biodiversity, affects human health and happiness, and is positively associated with real estate prices, according to the study.

Cities face a range of barriers to increasing street greenery, including identifying suitable areas, regulations, securing funding for planting and maintenance, water availability, and side effects like increased pests and vector-born diseases, Falchetti said in an email. “Many cities lack consistent, street-level data on vegetation coverage, making it hard to prioritize areas based on existing conditions and inequalities.”

The researchers studied within-city GVI distribution in the 10 most populated urban areas in 19 world regions to create the model, which allows city planners to track population and characteristics like social vulnerability in each district of an urban area to “support inclusive planning of street greenery by addressing areas with low exposure but high population density,” Falchetti said.

The model allows for targeted planning by pinpointing where low GVI intersects with high population density and “supports the prioritization of existing vegetation maintenance for areas of cities where GVI is on a declining trend,” Falchetti said.

Because the model is open-source and designed to work with publicly available satellite and ancillary data, cities need only moderate technical infrastructure and capacity — a kick-off process to set up the local workflow and a two- to three-day training for urban planning officers — to utilize it, he said.

How is the largest U.S. state transportation agency planning to use artificial intelligence? When the Texas Department of Transportation announced in December a three-year strategic plan for implementing artificial intelligence across the agency, it had identified 230 potential use cases. 

AI “can fundamentally revolutionize what we're doing and create that transformation in the transportation industry,” Anh Selissen, chief information officer at the department, said on a June 2 webinar on its plan hosted by the Eno Center for Transportation.

The 2025-2027 plan sets a road map for the department to integrate advanced analytics and intelligent systems into its operations and transportation system. 

Among the possible uses of AI, the department sees the technology as a way to prioritize infrastructure projects, manage budgets and contracts and avoid conflicts with utilities when planning project locations. AI could also provide decision support for emergencies, such as through weather monitoring, identifying the best evacuation routes and alerting and responding to roadway incidents and wrong-way drivers. 

The department also envisions a role for AI in workforce training and managing its vehicle fleet. “Our next step [in the process] will be bringing in a large amount of the traffic data so we can actually do some stronger AI analytics around traffic and congestion information from a stakeholder experience,” Selissen said. 

So far, the department is using AI in 22 aspects of its work, said Darran Anderson, the department’s director of strategy and innovation,  including for real-time and predictive traffic data analysis and crash detection. Layering AI on top of the department’s highway camera network is reducing incidence response time for emergencies or roadside assistance, he said.

Selissen said the department sees “millions of attacks” daily into its computer systems, making cybersecurity essential for any AI implementation. “If a hacker gets into this system and basically permeates your network, it is going to be a world of hurt for your organization,” she said. The department’s strategy, as laid out in its three-year plan, includes educating staff of the risks and ensuring AI systems are designed, tested and deployed in ways that ensure security.

The department’s guidelines include privacy protections along with checks and balances to ensure results are unbiased and provide equity across all users. Selissen said about AI, “You've got to learn it. You’ve got to figure out its capacity for use, and you’ve got to understand the boundaries of what that use should be.” She recommended trying pilot programs and starting with small tasks.

Sponsored content

By CivicPlus

Located in North Central Indiana, St. Joseph County blends industrial heritage with modern innovation. Home to 270,000 residents and institutions like the University of Notre Dame, the county has experienced rapid growth, driven by major tech investments. To meet residents' evolving needs, the county prioritized transparency, engagement, and efficiency through digital transformation.

A digital overhaul for a modern county

St. Joseph County sought to move beyond traditional, in-person government interactions by creating a seamless digital experience. The county implemented a comprehensive suite of digital solutions, including a responsive website, a mobile-friendly 311 CRM system, an online code library, and a mass notification system. These tools enhance efficiency, engagement, and accessibility.

Enhancing resident experience through a unified website

The county’s website serves as a digital hub where residents can access public information, submit service requests, and stay informed. Its intuitive interface and mobile-first design make navigation easy for tasks like permit applications and event searches. A biannual redesign ensures it remains modern and user-friendly.

According to Frank Fotia, Economic Development Planner:

"The design team is great. After having just a couple of meetings, we went through some examples and saw what we liked from some other municipalities or counties, and what we got back was beyond what we were expecting, it was fantastic."

Fotia also highlighted improved collaboration across departments:

"At the beginning, staff put something on the website and then maybe looked at it once a year and unless something changed, kept it very static. Now, we’ve gotten several departments buying into putting more things out there, sending us stuff for social media, and really increasing that online presence."

Improving accessibility with a mobile-first solution

A major milestone was the launch of a mobile app that allows residents to report issues like potholes or graffiti in real time. This direct line of communication has increased responsiveness and engagement.

Within just three weeks of its launch, over 200 reported issues were successfully addressed. Fotia emphasized:

"Getting people to find and download the app and being able to report issues to their government I think is incredibly important. Being able to meet the residents where they are, on their phones, they don't even have to go to our website to navigate there. They can just put the app on their phone and when they’re out and about in the community, if they see something, they can reach right out and let us know so that we can get it resolved."

Strengthening transparency with online codification

Ensuring open access to government documents and ordinances was another priority for St. Joseph County. The county’s online code library provides a user-friendly, centralized platform where residents can quickly search for resolutions.

Fotia explained:

"For transparency purposes, being able to put all our contracts, agreements, resolutions, and ordinances passed right up there on the website for anybody to see and read. I think it’s incredibly important, especially with all of the development and activity going on in this county, it’s very important for our residents to be able to have that ability to go online and see exactly what agreements we’re entering into and all of the contracts that we’re binding ourselves to, and as far as cost outlays, all out there, open book, anybody in the public can go online and see what we’re doing."

Personalized communication for a more connected county

The County has embraced a multi-channel communication approach to keep residents informed. A mass notification system allows the county to send alerts via email, text, and social media, ensuring messages reach residents in their preferred format.

Fotia recognized the benefits of this approach:

"Everyone has their preferred way of receiving information and communicating, whether it’s going to a website, or social media, or getting emails, or getting text messages or whatever else."

A future-ready approach to government service

St. Joseph County’s digital transformation has created a more responsive, transparent, and accessible government experience. The county is finalizing an integrated community development platform that will centralize permitting, code enforcement, and health inspections, ensuring 24/7 access to essential services.

Fotia expressed optimism about the future:

"Being able to have these tools in place, that are scalable and allow us to communicate and be transparent, is important for the county and important for our citizens as well. It does take some work, but I think at the end you’re going to say, 'I can’t believe we didn’t do this sooner.'"

As experts warn of potentially dire consequences from National Weather Service budget cuts and staffing shortages, local emergency managers are looking at how emerging tech and public-private partnerships could help ensure public safety during extreme weather events.

“Staffing shortages at NWS offices around the country mean that jurisdictions are at risk of not receiving tailored weather forecasts or timely weather alerts and warnings,” said Carrie Speranza, president of the U.S. Council of the International Association of Emergency Managers, in an email. 

The remaining NWS staff members are working overtime as NWS regional offices shift staffing to account for the shortfall, she added, noting that “staff burnout and fatigue bring additional risks to any operation.”

Speranza advised city officials concerned about the impact of NWS cutbacks to consider integrating data from “a plethora of private-sector companies that provide weather forecasting services” into their dashboards, planning tools and emergency protocols.

Tara Goode, vice president of strategic partnerships and radar operations for radar network company Climavision, said public officials are opening up to public-private partnerships with companies like hers. “There are things the private sector can do more cost effectively and a lot faster,” she said.

Climavision is building a proprietary network of low-level radars to fill coverage gaps in the federal next generation weather radar, or NEXRAD, system, especially in areas where the earth’s curvature and radar angles limit visibility near the ground. It has partnered with counties, cities and universities to install radars in 29 cities so far, providing data for early wildfire smoke detection, snow squall warnings and tornado verification, Goode said. The company’s goal is to install 200 radars, which it says would fill every low-level gap in the U.S.

“We want to make sure that the warning authorities, such as the weather service offices, have all of the data that they need to make the right calls when time is of the essence — when lives are at risk,” Goode said. “We also want to make sure that the local public safety officials and the state agencies that have to prepare, respond and recover from these types of disasters have the same data and can make their own decisions. Our data can help them when time is of the essence.”

Alex Stauch, head of marketing for Meteomatics, said partnerships that weave private high-resolution weather data directly into local decision-making represent “a shift from a one-size-fits-all approach to something more flexible and resilient.” Meteomatics provides services including drones and AI-driven weather models to supplement government forecasts with hyperlocal, real-time data. 

Meteomatics’ models integrate weather data from more than 110 sources, including aircraft, ground stations, drones, radars and satellites, allowing emergency officials to determine everything from when a storm will hit to how likely flooding will be, Stauch said. The data is recalculated hourly, he added.

“It’s not about replacing the public system, but backing it up and building around it,” Stauch said in an email. “The partnerships might look different from place to place, but the winning formula’s pretty consistent: real-time info, data you can actually act on, and tools that plug right into what local teams are already using. That’s what really counts when things get serious.”

Researchers at Texas A&M University have developed a tool that may be able to produce “near-instantaneous” damage assessments and recovery forecasts following a tornado.

Using remote sensing from high-resolution satellite or aerial images, deep learning technology and computational models that predict how an ecosystem will respond to restoration, the model can provide a macro-scale view of damage in areas hit by tornadoes within hours, according to a TAMU press release.

The model can be embedded into cities’ emergency response operations and long-term recovery planning so city officials can quickly identify the hardest-hit areas, prioritize emergency services and deploy resources where they’re needed most, one of the researchers said in an email.

Following a tornado, a National Weather Service team is typically sent to the site to do a full ground survey to assess damage. “Manual field inspections are labor-intensive and time-consuming, often delaying critical response efforts,” Abdullah Braik, a Texas A&M civil engineering doctoral student, said in a statement. Braik co-authored the research on the model published in Sustainable Cities and Society.

The model uses high-resolution sensing imagery and AI-generated algorithms to create damage assessments within hours, providing first responders and policymakers with “actionable intelligence,” Braik said.

Immediately following a tornado, an AI tool that has been trained using images and data from previous disasters analyzes satellite images to deliver detailed damage maps. These can help officials “anticipate which neighborhoods may face delays, due to infrastructure dependencies or social vulnerabilities, allowing targeted interventions and coordinated recovery efforts across housing, utilities and public health,” Braik said in an email.

Over time, cities can use the damage and recovery data they collect to calibrate and improve the underlying models, further refining their response strategies and strengthening recovery plans for future events. “This creates a feedback loop – turning each disaster into an opportunity to build smarter, faster and more resilient systems,” Braik said.

TAMU researchers are developing the model for use after other types of disasters such as hurricanes and earthquakes and hope to extend it beyond damage assessment to include real-time updates that track recovery over time.

Atlanta beat out coastal tech havens like San Jose, California, and Boston to rank as the top U.S. city for a “smart city future” in ProptechOS’ 2025 Smart City Index. 

The annual ranking is based on metrics related to each city’s tech infrastructure and connectivity, sustainability and tech job market. 

“Some of the benefits of scoring highly on these metrics for local citizens are that they can enjoy better public services, transport, security, economic opportunities and less air pollution,” according to ProptechOS, a software company based in Stockholm. 

Atlanta climbed from 10th place in last year’s index, where Seattle took the top spot.

In the latest ranking, Atlanta scored 88 out of 100 for tech infrastructure and 82 for its tech job market, with 643 tech jobs advertised per 10,000 people, according to the study. 

“Atlanta has the fourth-highest internet download and upload speeds of any major U.S. city,” ProptechOS stated. “Relative to its population size, Atlanta also boasts the third-highest number of companies dedicated to the Internet of Things (IoT) and the fifth-highest number of artificial intelligence companies.”

Atlanta has long been on a path toward strengthening its smart city status, joining the White House’s Smart Cities Initiative in 2015 and partnering with the Georgia Institute of Technology on smart city initiatives to ease traffic congestion, promote public safety and improve the city’s environmental sustainability. The city’s North Avenue Smart Corridor Project was recognized by the International Smart City Expo as a “living lab for active, multi-modal traffic management through technology deployment and data analytics.”

Atlanta also met its Better Buildings Challenge goal two years ahead of schedule, reducing the city’s energy consumption by 20% in 2019 and saving 1.3 billion gallons of water.

In May 2024 the city began enhancing its customer-facing digital products, including station screens and the mobile app for its Metropolitan Atlanta Rapid Transit Authority, a project expected to wrap up this June. 

While Atlanta dominated overall in the 2025 Smart City Index, San Francisco topped the list of cities with the best environmental metrics, in part due to its number and density of EV charging stations and LEED-certified green buildings. Oakland, California, scored highest on the index (98 out of 100) for its number of tech jobs, with 889 tech jobs per 10,000 people — enough tech jobs for approximately 10% of the city’s population, according to the study.

The index also ranked European smart cities in 2025, awarding Paris top billing, followed by Stockholm; Amsterdam; Munich; and Lisbon, Portugal.

 

Authorium, a cloud-based platform built for government administrative operations, announced the release of AuthorAI, a product that uses generative AI to streamline the procurement process.

AuthorAI generates statements of work when a user prompts it with a description of the product or service a government team is looking to procure. Authorium says it trained the AI tool on 15 million government procurement documents.

The system generates a statement of work “in minutes,” the company says, and it can include citations to original source documents. The user can then share and revise the SOW across government teams. The tool is available for immediate purchase from Authorium and AWS Marketplace.

The product launch was made possible by an $8 million infusion from investors at the start of 2025, according to Authorium. “Over the past year, Authorium has doubled the amount of government acquisitions that are processed on our platform, increasing from $25 billion in spring 2024 to $50 billion today,” Authorium co-CEOs Jay Nath and Kamran Saddique said in a joint statement.

The company also offers tools designed to streamline budget administration, grant management and legislative analysis.

Authorium was founded in 2014 under the name City Innovate, rebranding to Authorium in 2024. Sixty percent of its employees are former government employees, the company website states. Financial backers for the company include SJF Ventures and Govtech Fund.

AI in local government ‘in its infancy’

A survey of local government officials last year found that 48% viewed AI utilization as a “low priority” for their local government. Under 6% of respondents considered AI “high priority,” according to the International City/County Management Association, which surveyed 635 local government leaders in April and May of 2024.

“The use of AI as a core practice of local government in the U.S. is clearly in its infancy,” ICMA states in a report on the survey results. “There are some communities that have exuberantly embraced AI. Others are approaching this emerging technology with the utmost caution.”

The top area local government leaders considered for AI potential was “resident engagement,” with 55% of respondents seeing promise in streamlined service interfaces and AI chatbots, the report states.

A separate survey of local government IT executives in late 2024 found that 38% felt their local government was “not prepared at all” to safely incorporate AI technology, with 46% saying it was “somewhat prepared” and 9% saying their local government was “fully prepared.”

Despite local governments being in the early stages of AI adoption, major companies see its potential in the public sector. OpenAI, the developer of ChatGPT and one of the largest AI service providers, launched a version of its product specifically tailored for U.S. government agencies earlier this year.

The government procurement space has seen an abundance of new AI tools from companies like OpenGov, Appian, Icertis, Cloverleaf AI and more, Procurated Director of Marketing Steve Isaac recently wrote in a column for American City & County.

“Today, we are witnessing a second wave of AI applications for government procurement — not just speeding up how our industry works but fundamentally changing it,” Isaac said.

As wildfires ravaged the Los Angeles area in January, some residents evacuated, others opened their homes to family and friends and many became citizen scientists, using their phones to check the air quality around them.

Historically, air quality monitoring was the domain of regulatory bodies like the Environmental Protection Agency, which sets national standards for measuring and improving air quality and reducing pollution and communicates the air quality via the Air Quality Index. The EPA installs air quality monitors in areas that are densely populated or at high risk of pollution, but such monitors are by no means ubiquitous. Due in part to the price of these high-quality instruments, only about a third of U.S. counties have these monitors.

But with rapid advancements in air quality sensors over the past several years, communities are now home to networks of lower-cost sensors that collect and share air quality data essential to public health. New research also suggests that air quality data collection will only improve.

Better sensors with lower costs

“We're at a point now where if a community had a few tens of thousands of dollars, it could install a community air quality network and get reasonably good results,” said Charlie Catlett, a senior computer scientist at the U.S. Department of Energy's Argonne National Laboratory, who first started an urban sensor pilot project in Chicago about a decade ago. Back then, Catlett said in a recent interview, air quality sensors were much more expensive. Plus, those that were commercially available were geared toward the household, not the community level, he said.

Today, sensors are not only cheaper, but many are designed to present a picture of air quality across communities.

While cities from New Haven, Connecticut, to Sandy City, Utah, have installed their own air quality sensor networks, these networks have also sprung up in other communities through the proliferation of simple, consumer-grade sensors belonging to individual households. In the District of Columbia, the local government loans sensors to interested residents with eligible properties.

The South Coast Air Quality Monitoring District, the agency enforcing air quality regulations in the Los Angeles area, has tested hundreds of low-cost sensors to determine their efficacy via its Air Quality Sensor Performance Evaluation Center. “Performance of these sensors varies greatly,” said Andrea Polidori, assistant deputy executive officer of monitoring and analysis at South Coast AQMD. But the sensors that detect particulate matter — particle pollution — “are actually quite decent,” he said.

One such sensor is produced by the company PurpleAir, which updates aggregate, localized air quality data every two minutes on its free maps.

PurpleAir sensors measure PM2.5, fine particulate matter up to 2.5 microns in diameter, small enough to be inhaled and travel into the lungs. PM2.5 makes up roughly 90% of the particulate matter emitted from wildfires.

Over the past several years, sensors like PurpleAir’s have also gotten better at illustrating air quality in a given area, said Karen Clay, professor of economics and public policy at Carnegie Mellon University’s Heinz College of Information Systems and Public Policy. Data hasimproved as sensors have become more common, she said, because “you get much better local data if you have more sensors.”

Air quality inequality

Air quality sensors are not equally distributed in communities, pointing to a role for cities, said Clay.

Some communities “already have pretty dense PurpleAir networks, so in those cases, having [city-provided] monitors may not add a lot of value,” she said.

But cities can make a difference by strategically placing sensors in communities where there are data gaps, Clay said, such as neighborhoods where lower-income residents are less likely to purchase their own sensors.

These communities are also more likely to be affected by PM2.5, according to a 2018 study published in the American Journal of Public Health.

A 2022 study in that journal found that though the number of PurpleAir sensors increased in California between 2017 and 2020, there were fewer sensors in census tracts with higher rates of residents of color, low-income residents and PM2.5 levels.

The study also found that the areas in California with the highest levels of pollution and socioeconomic disadvantage included southern Los Angeles County and areas downwind of it.

A network of air quality networks

PurpleAir’s air quality map shows hundreds of PurpleAir sensors in the Los Angeles area.

Clay said data from low-cost sensors bolsters the government’s more advanced data. Indeed, PurpleAir’s hyperlocal data is now integrated into the air quality reports the EPA provides, such as its AirNow Fire and Smoke Map.

But while current sensors can measure the level of air pollution, they still don’t give a complete, real-time picture of what’s in the air, said Jason Low, deputy executive officer of monitoring and analysis at South Coast AQMD. That’s why it’s important to “communicate to the public the context of air monitoring data,” he said.

The Air Quality Index “is a good indicator for smoke,” Low said, “but there are other compounds [in the air].” The AQI doesn’t take into account all toxic material, such as lead, that could be in the air. And wildfire particulate material is more toxic when fires spread through cities rather than forests because of the burning of man-made materials.

The air quality app maps that many Californians relied on in January essentially show total PM mass concentration, said Roya Bahreini, a professor of atmospheric science at the University of California, Riverside. “But what [an app map] doesn't tell you is what the particulate matter is made of.”

Bahreini is co-principal investigator of the federally funded Atmospheric Science and Chemistry mEasurement NeTwork project, or ASCENT, which continuously monitors aerosol composition at 12 sites nationwide.

During the wildfires, ASCENT’s research-grade monitors captured high levels of toxins, including lead, at the project’s sites in Los Angeles, Riverside and Joshua Tree National Park. That’s perhaps not surprising where the fires were burning houses with lead paint and cars with lead batteries.

Connecting data to people

While the ASCENT data is still preliminary, the researchers released data from the Los Angeles-area sites as the wildfires were raging to inform residents about the specifics of the air pollution around them, complementing the usual AQI data. For example, if particulate material was high, residents could look at the ASCENT data to see if the pollution was due to typical emissions or if it contained urban wildfire toxins, Bahreini said.

The South Coast AQMD recently expanded its air monitoring efforts, deploying mobile monitoring vehicles to identify sites with high levels of toxins and human-made chemicals, as these areas could require further monitoring. Data it captures will be shared with the public via the South Coast AQMD’s website, experts at the agency said.

“Getting the data to people — and getting people to pay attention to the data,” is challenging for air quality networks, said Clay.

Residents could benefit from alerts when there are toxins like lead in the air, she said. For example, a digital screen in town could broadcast the AQI to a community, she suggested.

“[AQI] does show up on some weather apps,” Clay said. “But how much are people staring at weather apps?

State and local communities are facing a rise in cyber threats from nation-state-linked and criminal threat groups, which in many cases are looking to undermine confidence in public institutions, according to a report by the Multi-State Information Sharing and Analysis Center. 

Much of the nation’s critical infrastructure is owned and operated at the state or local level, which underscores the need to develop more resilient systems, share threat intelligence and prepare for coordinated incident response.

State and local governments have raised concerns in recent years about the need for additional resources, including funding, technology upgrades and trained personnel who can monitor and rapidly respond to malicious attacks and other cyber incidents.

The report comes amid an increase in ransomware, supply chain attacks and other threats facing state and local organizations in recent years.

Hospitals and healthcare providers have increasingly been targeted by ransomware groups in recent years.

Water and wastewater treatment facilities have faced a series of attacks from state-linked actors, hacktivist groups and criminal ransomware gangs. 

In December, Rhode Island officials were confronted with an attack that disrupted a key social services site that provided food assistance, healthcare and other services to vulnerable residents.

State officials say that state-linked threat groups have targeted local and state organizations as part of a larger effort to undermine public trust through the disruption of critical services. In many cases, state and local officials are fighting uphill battles to defend against malign actors armed with financial resources and plenty of time to launch their attacks.

“There is a need for sustained stable funding for cyber defenses, and with that includes the ability to hire skilled personnel,” said Josh Bauman, director of technology at the Festus R-VI School District in Missouri and a member of the MS-ISAC Executive Committee. 

Emergency response management is a high-stakes space. Each decision can mean life or death, which makes training and preparing for these scenarios critical. 

Over the course of a year, a three-person team from George Mason University worked to improve preparedness via AI-powered games for the Department of Public Safety Communications and Emergency Management in Arlington, Virginia. 

Using an iterative process and feedback from Arlington County department members, the team created two interactive games, called Go-Repair and Go-Rescue, which simulated infrastructure maintenance, resource allocation and evacuations. The dynamic learning environment provided utility managers and volunteers with a wider scope and more flexibility than traditional training methods.

“AI optimization and reinforcement learning models calculate the optimal decisions, and this allows players to assess whether their decisions are better or worse comparatively,” said Shima Mohebbi, assistant professor in the Department of Systems Engineering and Operations Research at GMU and the project’s leader.

Incorporating AI into training enables the department to better identify areas of improvement in decision-making and provides a tool to learn from mistakes without real-world consequences. To avoid subjectivity often seen in large language models, the researchers developed a model with a mathematical foundation and used reinforcement learning. 

Enterprises can also gain value from incorporating AI into training and workforce development, from adaptive learning platforms to AI-augmented skills assessments. More than 3 in 5 HR pros expect AI to wholly transform the way training is currently conducted, according to a survey by management consulting firm OC&C Strategy Consultants. Most respondents said they were looking to expand AI adoption to talent management. 

To infuse AI into training, business leaders can choose off-the-shelf options. Most corporate training providers, including Coursera, Skillsoft and LinkedIn Learning, have added the technology to their tech stacks to some degree, and CIOs can also look into developing in-house solutions. 

Mohebbi said leaders interested in developing a similar solution to Go-Repair and Go-Rescue should start by getting a clear understanding of the problem they want to address. To do that, Mohebbi’s team focused on connecting with stakeholders.

AI model decisions

Despite the current spike in interest regarding generative AI, Mohebbi’s team didn’t pursue that particular branch of artificial intelligence. 

“The AI models that we are using are based on a mathematical foundation, not just generated based on past experiences,” Mohebbi said. “We really need to have that strong mathematical foundation to avoid the subjectivity we usually see in large language models.”

Hype alone isn’t enough to warrant implementation.

CIOs have had to work to rein in generative AI use cases and prioritize those with the most potential. The costly technology is not the best or most reliable tool in every situation. 

“We want to look at AI, which is not a hammer looking for a nail, but [by being] clear in terms of what’s the business problem or context we are solving and how can AI enable us to do that,” Surabhi Pokhriyal, chief digital growth officer at Church & Dwight, said during the company’s analyst day last week.

Technology leaders who tease out generative AI’s weaknesses and lean on its strengths to prioritize use cases find success. Oftentimes that means fewer deployments. Leading organizations pursue about half as many generative AI opportunities on average compared to less advanced businesses, according to Boston Consulting Group. 

As the GMU project developed, stakeholders raised questions about the reliability of the AI models used to power the training games. 

“We added more information about these background algorithms,” Mohebbi said. “They were more comfortable afterward and enthusiastic about adopting the game.”

Implementing stakeholder feedback

Go-Repair and Go-Rescue function quite differently than they did initially. Mohebbi’s team created several versions before landing on the current functionality and capabilities. 

“We held meetings and then provided a version and they would test it out and provide us with feedback,” Mohebbi said. “We’d address their feedback and provide them with an updated version. It was a very iterative process.”

One of the first changes was also the biggest. The game was initially designed to be played offline, but security protocols within the department wouldn’t allow downloads. Mohebbi’s team then shifted to web-based development. 

Mohebbi’s team also added visual elements, improved user experience, enhanced the localized scenarios and provided more insight into user decisions and their consequences throughout the process. 

The partnership between technologists and end users was crucial to the project’s success. 

“If it wasn’t for the feedback and our interactions with Arlington County, we wouldn’t even know about the certain specifics and features needed,” Mohebbi said.

The National League of Cities and Google announced on Wednesday a report on how local governments can leverage artificial intelligence-driven technologies — and mitigate the risks they present. 

AI offers cities the opportunity to become more efficient, sustainable and livable, NLC CEO and Executive Director Clarence Anthony said in a foreword to the report. He pointed to applications such as optimizing traffic flow, reducing energy consumption and streamlining administrative tasks. 

However, not all cities are using or even have access to AI, which could turn “the digital divide that has plagued many cities into an ‘AI divide,’” Karan Bhatia, vice president and global head of government affairs and public policy at Google, said in a statement.

While the concept of AI has existed for decades, it burst into the public discourse in 2022 and 2023 with the release of generative AI applications that allowed people around the world to engage more directly with the technology for the first time, the NLC and Google report says. 

Generative AI, which can create new content like text and images based on patterns learned from existing data, is just one of numerous categories of AI. Other types of AI relevant to local government are predictive AI, which makes predictions about future events or trends, and perceptive AI, which interprets and understands sensory inputs.

Each of these types of AI brings distinct promise and pitfalls, says the report, which was spearheaded by the NLC’s AI Advisory Committee.

Generative AI, for example, can be used to translate public meetings and 311 calls or to develop chatbots that can respond to resident questions. However, generative AI can make mistakes. New York City’s chatbot made headlines earlier this year after providing residents with misinformation on topics such as housing rights and worker protections. 

AI tools may also reflect existing biases present in the data they are trained on; for example, they could reinforce existing inequitable resource allocation, the report says. AI tools also may pose data privacy and security risks and become a target for cyberattacks.

The report contains a toolkit that walks cities through the steps of getting started with AI, including conducting a landscape analysis, assessing readiness, developing use guidance and policy, and engaging the public.

Throughout the report, NLC and Google highlight how some cities are already using AI. Numerous communities have chatbots that help users find information about the city and its services, while Sunnyvale, California, gives residents who attend in-person city council meetings access to in-time closed captioning translation. Memphis, Tennessee, is using AI to find potholes by analyzing video footage from city vehicles and combining those results with existing data. Tucson, Arizona, meanwhile, is using AI software to predict which water pipes are most likely to fail next, helping the city prioritize maintenance and repairs.

But barriers to these applications remain: Resource constraints, concerns from city staff or elected officials and lack of political will to invest in technology are among the challenges cities face in implementing AI, the report says.