North Carolina Treasury Boosts Efficiency with ChatGPT Pilot

North Carolina's Department of State Treasurer recently concluded a pilot program exploring the use of ChatGPT to enhance efficiency in state operations. The initiative aimed to leverage artificial intelligence to assist employees with tasks such as tracking down unclaimed funds and verifying local government budgets. The pilot demonstrated significant time savings for employees, with a majority reporting positive experiences.

Conducted from March to June 2025, the three-month pilot involved employees from the Department of State Treasurer's Unclaimed Property and State and Local Government Finance offices. An enterprise version of ChatGPT was provided for their use. An independent study, carried out by the Institute for Artificial Intelligence and Emerging Research at North Carolina Central University, evaluated the program's impact.

The study found that participants saved an average of 30 to 60 minutes per day, with 85 percent of users reporting a positive overall experience. Initial surveys indicated daily savings of 15 to 30 minutes, which increased as users became more proficient with the tool.

Treasurer Brad Briner commented on the pilot's success, stating, "Adopting this innovative technology has helped us deliver improved results to our constituents and to taxpayers. This important analysis clearly shows that adding the power and speed of artificial intelligence to the talent, experience and judgment of our state employees is the key to unlocking greater workplace achievements."

State workers utilized ChatGPT in various capacities. Key applications included drafting professional communications such as reports and memos, translating complex technical documentation into simpler language, and brainstorming content for policy documents, training materials, and public outreach. Employees also used the AI to summarize lengthy legal texts, multi-page reports, and public submissions. Furthermore, it served as a research tool, allowing users to ask clarifying questions on intricate or unfamiliar subjects.

Participants received guidelines for ChatGPT usage and underwent prompt engineering training led by OpenAI. The study noted that the most effective results were achieved when users collaborated with the AI, refining its outputs with their own expertise.

Despite the positive outcomes, the pilot also highlighted several challenges and limitations. A primary observation was that ChatGPT cannot replace human judgment; users consistently needed to combine the tool's outputs with their own expertise for optimal results. Functional challenges included occasional inaccuracies, with the AI sometimes generating incorrect statements or citations. Users reported that the tool underperformed in tasks requiring precise coding, legal accuracy, or complex mathematical computations.

Broader concerns that limited more frequent AI adoption included persistent questions about accuracy, the learning curve required to use the tool effectively, and privacy considerations regarding what information could safely be entered.

North Carolina's pilot aligns with similar initiatives in other states. Earlier this year, Pennsylvania concluded a yearlong pilot with OpenAI, where over 175 employees in the state’s Office of Administration used ChatGPT Enterprise. That program also reported positive results, with 85 percent of participants having a favorable experience and reporting an average daily saving of 95 minutes.