AI as a Budgeting Tool: Panacea or Pandora’s Box?

Abstract

Local government officials are consistently tasked with doing more with less. The combination of fiscal institutions, devolution of fiscal federalism, voter frustration, and economic shocks have been the leading causes of these more recent challenges. However, the reality is that smaller local governments, by their very nature, struggle with service provision. This paper explores the next evolution of change to local governments – artificial intelligence (AI), specifically ChatGPT – and its potential for local governments. We discuss the use of AI in budgeting, with a focus on financial data management and statistical analysis, including forecasting and policy recommendations for a small Nebraska municipality. Our experience with ChatGPT highlights its powerful capabilities for data processing. Transitioning to ChatGPT-4 from ChatGPT-3.5 (with or without Python) improved data processing efficiency but introduced financial costs, and we observed that the use of detailed and precise prompts enhances output quality across all versions. Achieving a balance between time, cost, data preparation, and prompt precision is crucial for optimizing the potential of ChatGPT in financial data analysis.