The cities getting the most out of AI aren’t just the ones adopting it fastest but those putting their residents first. These cities are seeing more than twice the return on their AI investments compared with their peers. This is according to ThoughtLab’s new global report, Building a People-First, AI-Powered City.
The research is based on a survey of 525 senior officials across 215 cities worldwide, alongside a review of 363 city AI plans. The report found a clear link between how cities approach AI and what they get in return. It also identified 37 AI leaders that combine technology with strong governance, workforce development, citizen engagement and human-centred design.
Mark Wittenburg, Raleigh’s CIO, participated in the study and was part of the report advisory board. He thinks AI may be changing how government functions, but the core purpose of government hasn’t changed at all. “Our job is to serve the people and put community first,” he says. “Whatever technology we bring in should enable the needs of the community, whether that’s public safety, transportation, or delivering clean water.”
Heart of the matter
Raleigh’s early use of AI was to advance its Vision Zero goals, a global initiative that aims to eliminate road deaths and serious traffic injuries, rather than simply reduce them.
“At the time, our transportation team was counting cars at intersections, but the data they gathered only captured how many cars moved through the intersection,” Wittenburg says. “It didn’t reveal any insights we could use to help reduce fatalities.”
Working with NVIDIA, Raleigh built ML models that analyze video footage from intersections, and detect things like stalled vehicles or water on the roadway.
“Instead of reacting to a call from the community, we’re being proactive,” he adds. “Moreover, we can now look at lane utilization and see the left turn lane is 95% subscribed. With this information, we can change the signal timing and test the impact of these adjustments.”
The next step is to go further, looking at near misses. Using advanced analytics and building a vision learning model, the city can hopefully prevent accidents. While near-miss data isn’t a reliable predictor of exactly where the next accident will happen, when combined with other transport data like crash records, traffic volumes, and vehicle speeds, it can help identify trends and address risks before they result in fatalities.
“Accidents don’t actually happen very often but near misses happen more often than you’d expect,” he says. “By analyzing actual accidents and near misses, we have more data points to work with and can make more accurate decisions.”
Sky’s the limit
Raleigh is also using AI for microclimate modelling. As part of a citizen science project, data was collected at different times of the day and in different places. This was compiled into a map to show where hotter zones in the city, called Urban Heat Islands (UHI), are located.
For example, on the hottest days, temperatures in neighborhoods with less green space can be up to 20 degrees higher than in shadier areas with more vegetation. These maps can help Raleigh prepare for and respond to extreme heat.
Band of agents
Additionally, Raleigh is using AI to streamline three areas of its municipal operations. Its gen-AI chatbot, Raleigh, helps employees easily find information across different departments. The city is now pairing this capability with Ral-E, an AI agent that handles IT help-desk tickets, communicates with employees through Teams or email, and resolves defined requests.
It’s also deploying Alli, a separate agent that autonomously handles simple help-desk tasks like password resets. “We’re getting positive results because the agent works 24/7, so if you need help after hours, you can get it,” he says. When these technologies are interacting with people, trust is critical. “Being transparent about when someone is interacting with an agent rather than a person is important,” Wittenburg says. “We also must be clear about what data we’re collecting and how we’re using it.”
Common purpose
Across all these examples, the focus may be different, but the desired outcome of making Raleigh a better place to live is the same. For Wittenburg, a people-first AI strategy is about starting small, testing, and then scaling. “We approach AI with a crawl, walk, run mentality,” he says “And maybe we stub our toe a few times along the way. This process sees us experimenting with low risk, internal stuff before we use it in ways that directly impact or directly interact with the community.”
While private-sector businesses prioritize ROI, the public sector has a different measure of success, he says. “As a city, it’s more of a value discussion,” he continues. “We’re always asking if a project will add value by making staff more efficient and effective, or if it’ll have a positive community impact. The technology should never be in the driver’s seat. It’s just an enabler that helps us add value to the people.”