Much of that work falls to CIO Mark Wittenburg and the city tech team, whom he credits with driving much of the progress over the past year as Raleigh has quietly expanded its use of AI, digital infrastructure and cybersecurity collaboration. Efforts span everything from AI tools at City Hall to systems analyzing traffic camera footage and experimental technology embedded in municipal services.
Wittenburg describes that strategy as a “crawl, walk, run” approach to AI — a philosophy intended to make sure that experimentation does not outpace oversight. Early pilots are intentionally low-risk, designed to test concepts before expanding them across departments or releasing them publicly. The CIO characterized it as, “crawl is okay, let’s pilot something nice and easy; something low-risk that gets our feet wet and then adds value to the organization.” The next “walk” phase involves internal testing and supervision before the technology eventually reaches residents.
That philosophy is now helping Wittenburg and his team guide a growing list of initiatives across Raleigh’s government, particularly as AI begins moving beyond traffic analytics and infrastructure planning into administrative workflows in city departments.
AI MOVES INSIDE CITY HALL
Raleigh has drawn attention in the past two to three years for applying AI to physical systems like traffic monitoring, and for its work in digital twins. But over the past year, officials have begun exploring how the same technology could simplify internal government operations.
One of the earliest steps was launching an internal chatbot called “Raleigh,” developed with ServiceNow. The assistant, deployed in October 2025, was created to help city employees navigate internal processes and locate information across departments.
Leaders expected the chatbot to answer basic administrative questions. Instead, it quickly demonstrated how AI could connect multiple systems across government.
Within an hour of launch, a city employee asked the system how to change their name in internal records. Rather than providing a simple instruction, the chatbot produced a far broader answer. According to Wittenburg, rather than directing the user to a single system, the AI recognized connections across HR and facilities processes, suggesting additional tasks like updating beneficiaries or requesting a new identification badge. For the IT team, the episode illustrated how quickly AI could synthesize institutional knowledge.
TESTING AI THAT CAN ACT, NOT JUST RESPOND
Raleigh’s next step in AI experimentation began last month, when staff started testing a new ServiceNow capability known as a “zero touch agent” — essentially agentic AI. Unlike traditional chatbots that only respond to questions, the system is designed to perform tasks automatically.
The goal is to determine whether AI can help manage IT service requests and internal support queues — a process that currently requires human staff to review and route tickets manually.
“The whole idea is that not only is the agent answering questions, but then we would have an AI help desk agent that would actually be servicing queues, talking to customers and working through our backlog,” Wittenburg said.
If the internal pilot continues to perform well, Raleigh plans to expand the technology to its public service portal, Ask Raleigh, where residents submit requests for city services. Wittenburg said they are considering exploring future deployment of agentic AI too, though only after extensive internal testing and oversight.
AI WATCHES THE STREETS
While some AI tools are focused on internal operations, others are analyzing the physical movement of the city itself.
Raleigh is also currently piloting a project known as “Raleigh in Motion.” The system uses a visual language model to analyze traffic camera video feeds across the city, letting staff query footage directly and generate insights about how people and vehicles move through intersections.
The pilot has been running for about three months, and is slated for expansion in the next several months. Earlier systems focused primarily on counting vehicles or measuring traffic flow, but the new model adds generative AI capabilities that can identify patterns and trends automatically.
City planners could, Wittenburg says, ask the system how many pedestrians crossed a particular intersection, or how movement patterns have changed over time. The AI can then highlight data insights such as shifts in crossing patterns or changes in traffic direction, providing planners with a deeper understanding of urban mobility.
Still, some of Raleigh’s vital technology projects may never be visible to residents at all.
One example is a proposal being developed with the city’s Solid Waste Services department, which is exploring a new system that would equip garbage trucks with cameras and routing technology. A request for proposals is being prepared.
The system would combine video feeds and GPS tracking to create optimized collection routes. Over time, the technology could even provide residents with more precise service windows. Instead of broad collection time ranges, staff would use a “street digital twin” and then interface with an “address database” to optimize routes for collections.
Missed trash cans are the most frequent calls received by the city’s solid waste department; cameras could help determine whether the container was actually missed or simply never placed at the curb. At the same time, the trucks’ cameras could potentially identify infrastructure issues — such as potholes or missing signs — during routine collection, according to Wittenburg.
CYBERSECURITY EXPANDS WITH THE CITY’S DIGITAL FOOTPRINT
As Raleigh’s technology footprint grows, so does the need to secure it.
The city hired a new chief information security officer about two years ago, and the IT department has since begun implementing zero-trust architecture, a cybersecurity model that verifies every device and user before granting systems access.
The effort is still in its early stages, but Wittenburg said the city is focusing heavily on vendor risk assessments as new systems come online. Understanding how secure external partners are, he said, is essential as governments adopt more connected technologies.
Cybersecurity has also become a regional collaboration. Raleigh’s security leadership helped establish a statewide information-sharing group with Wake County and North Carolina officials, allowing agencies to exchange threat intelligence and assist one another during cyber incidents.
The network effectively creates a mutual support system for local governments facing increasingly complex threats.
TRAINING THE WORKFORCE FOR AI
Technology alone is not enough to transform government operations; employees must also understand how to use it responsibly, the CIO said.
To support those efforts, Raleigh released Copilot Chat last year to assist employees, and is currently piloting Copilot for Microsoft Office 365. Officials created a three-part curriculum that includes cybersecurity fundamentals and data privacy, teaching staff the “dos and don’ts of AI.” Later sessions focus on prompt engineering — how to ask AI systems effective questions — and explaining how different models handle data.
The goal is to help employees understand the capabilities and the limitations of generative AI systems before integrating them into daily workflows.
THE LONG GAME OF GOVERNMENT INNOVATION
Regardless of the technology type, Wittenburg emphasized, the city tests it internally, evaluating how it performs in real workflows. Only then would leaders move toward broader deployment.
If the project succeeds, it could impact residents every day — even if much of the technology itself remains out of sight — through faster city services, smarter infrastructure planning and, as Wittenburg noted, perhaps even a notification telling them exactly when the trash truck will arrive.
