Local governments across the globe are beginning to use AI for an increasing number of capabilities on the job. Karen Dahut, CEO of Google Public Sector, was even quoted as saying, "AI has the potential to revolutionize the way the public sector operates, serves its missions, and supports its citizens.”

But few municipalities have begun to explore AI’s full potential when it comes to the complex world of budgeting and finance. AI in all its forms — generative, predictive, and machine learning — can be used to automate and enhance various budgeting and finance-related processes within local government operations.

“Embracing AI in the Finance Office: Leveraging the Technology to Help You Leave the Office on Time” was an education session at ICMA’s 2024 Local Government Reimagined (LGR) Conference in Palm Desert, California. Presented by Stuart Langer, senior vice president of product strategy at OpenGov, the session gave attendees a peek at the many ways that local governments can harness AI for tremendously greater efficiency in finance and budgeting.

It all starts with data. Langer recalled the time in his career when spreadsheets were first introduced, which revolutionized the way data could be organized and presented. Then search engines changed how we interacted with data. Now, AI is changing our entire relationship with data.

What does that mean for the finance office? Here are just a few areas where generative AI can help:

Data Analysis

• Improves accuracy and speed in handling data analysis tasks traditionally done manually, such as outlier detection and fraud prevention.
• Generates comparative analytics of provided data (i.e., comparison to benchmarks)
• Simplifies budget to variance analysis, including the automation of listing variances by dollar value and percentage.
• Assists with complex Microsoft Excel functions, formula writing, pivot tables, and macros.
• Develops graphs and visualizations for presentation preparation.

RFP Process

• Researches technical topics.
• Assists in drafting RFPs, RFIs, and solicitations, which can be especially helpful for repetitive projects.

Budget Initiative Narratives

• Assists in writing narratives for public consumption, focusing on strategic initiatives and outcomes rather than just budget-to-actual comparisons.

FOIA Requests and Resident Q&A

• Handles queries from residents and provides automated responses via chatbots.
• Creates specific, context-rich prompts for residents to use for better outcomes.

Writing

• Drafts coherent emails in the context and style of your choice.
• Writes responses to inquiries with high-quality messages quickly.
• Creates checklists (audit prep, month-end close, purchasing compliance).

Productivity

• Helps with time management, such as setting reminders.
• Assists in planning, such as generating to-do lists.

Prompt Creation

Prompts are perhaps the most important part of generative AI. Its output is only as good as the prompts you use. You’re asking your AI tool for specific information, so it’s important to pose your request in the best possible way. Writing an effective prompt is both an art and a science. Langer shared a few tips for successful prompt creation:

• Add structure and specificity.
• Provide context.
• Ask for multiple variations.
• Try multiple revisions.
• Ask for explanations and citations.
• Use Gen AI as an editor.
• Use alternative input formats.
• Save what works.
• Experiment with advanced techniques.
• Keep exploring and learning.

Langer suggests using the RELIC formula to guide your prompt writing in Chat GPT or any other generative AI tool:

R — Role (For example, “You are an assistant city manager with over 20 years of experience…”)

E — Exclusion (For example, “Do not include…”)

L — Length (For example, “Keep your writing between 300 and 400 words.”)

I — Inspiration (For example, “Use this article as a reference for style, tone, and voice: [Add link or paste text.])

C — Context (For example, “You work for the City of X and you’ve been tasked with creating a plan for a multicultural day…)

This may seem like a lot to include in a single prompt, but Langer explained that you don’t have to use all of these at once. He has found that just using role, length, and context can usually produce great results.

Key Considerations and Risks

Data Privacy

Of course, of paramount importance when it comes to AI in the local government space is data governance and privacy. Especially when it comes to sensitive financial information, data must be handled ethically, transparently, and lawfully. Data privacy must be considered every step of the way, from collection to analysis to disposal.

Carefully consider which AI tools your organization chooses to use and how your staff will be using them. Also be mindful of employees using a free AI tool with their own personal login while on the job. When using a free AI account, all the data you feed into it now belongs to the company that made the tool you’re using. It’s crucial to have organization-wide AI policies in place and educate staff.

Data Hallucinations

What those new to generative AI often forget is that AI output requires human verification to ensure accuracy. Data hallucinations, in which AI might generate plausible but incorrect data, can and do occur. Double check the information with other sources.

Conclusion

Generative AI is becoming essential in government finance for a multitude of applications, enhancing efficiency and accuracy, but requiring robust verification and governance frameworks. The technology holds promise for revolutionizing budgeting and finance tasks, though it must be handled with care.

Learn more about AI for local government on ICMA's Generative AI resource page.