North Carolina Boosts Treasury Efficiency With AI

In North Carolina, treasury employees are tasked with crucial responsibilities, from reuniting people with their lost money to ensuring local government budgets are balanced. Thanks to a new pilot program with ChatGPT, they are now performing these duties more efficiently than ever.

A recent report on the state's collaboration with OpenAI revealed impressive results: the AI tool saved workers an average of 30 to 60 minutes each day, and 85 percent of those who participated reported a positive experience.

Inside North Carolina's ChatGPT Experiment

The Department of State Treasurer conducted a three-month pilot program, offering an enterprise version of ChatGPT to employees in its Unclaimed Property and State and Local Government Finance offices. An independent study by North Carolina Central University's Institute for Artificial Intelligence and Emerging Research found that the time savings grew as users became more familiar with the technology. Initially, employees saved 15 to 30 minutes daily, but this figure increased to over 30 and even 60 minutes by the end of the pilot.

"Adopting this innovative technology has helped us deliver improved results to our constituents and to taxpayers," Treasurer Brad Briner stated in a press release. "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.”

How State Workers Used ChatGPT

Employees leveraged ChatGPT for a wide range of tasks to boost their productivity. Key applications included:

• Drafting professional communications, reports, and memos.
• Translating complex technical documentation into plain, easy-to-understand language.
• Brainstorming ideas for policy documents and training materials.
• Summarizing lengthy legal texts, multipage reports, and public submissions.
• Serving as a research assistant for clarifying questions on unfamiliar topics.

Participants received prompt engineering training from OpenAI, but survey feedback indicated that more role-specific prompts and guided scenarios could have further boosted their confidence and effectiveness. The study highlighted that the best outcomes were achieved when users treated the AI as a collaborator, working with it to refine and improve the generated outputs.

The Challenges and Roadblocks of AI Integration

The pilot also underscored that AI is a tool to augment, not replace, human expertise. Users had to apply their own judgment and knowledge to achieve the best results. Several roadblocks emerged during the experiment. The most significant functionality issue was occasional inaccuracies, where ChatGPT would generate incorrect statements or false citations. The tool was also found to underperform in highly specialized tasks that required coding, precise legal analysis, or complex mathematical calculations.

The primary factors that limited more frequent use of the AI were concerns about accuracy, the learning curve required to use the tool effectively, and lingering questions about data privacy and security.

A Growing Trend: Other States Embrace AI

North Carolina is not alone in exploring the potential of AI in the public sector. Pennsylvania recently finished its own pilot study with OpenAI. Their yearlong pilot involved over 175 employees in the Office of Administration and yielded similarly positive feedback. In Pennsylvania, 85 percent of participants also reported a positive experience and noted even greater time savings, averaging 95 minutes per day.