Application and Impact of Large Language Model (LLM) Agents in Financial Operations and Public Policy

The rapid digitalization of finance, accelerated by advances in artificial intelligence, has positioned Large Language Models (LLMs) as transformative tools not only for private-sector financial operations but also for public-sector economic governance. While much of the existing literature has focused on LLM applications in investment and trading, their potential to reshape core financial operations—such as regulatory compliance, fraud detection, contract intelligence, and risk reporting—remains underexplored. Simultaneously, governments and central banks are beginning to grapple with how LLM agents can assist in policy formulation, macroeconomic simulation, fiscal transparency, and real-time economic sentiment monitoring. This research undertakes a systematic examination of the deployment of LLM agents across these dual domains, with a particular emphasis on operational efficiency, decision support quality, and institutional accountability. Preliminary findings suggest that while LLM agents significantly reduce processing times for tasks like regulatory reporting and contract review, their outputs exhibit concerning variability in reasoning consistency and are susceptible to subtle prompt-induced biases. In the policy domain, LLMs demonstrate promise in summarizing vast economic literature and generating scenario simulations, yet they struggle with causal reasoning and fail to adequately quantify uncertainty—raising serious concerns for their use in fiscal and monetary decision-making. The research concludes by proposing a human-in-the-loop governance architecture that balances automation with oversight, alongside a set of regulatory benchmarks for evaluating LLM reliability in financial and policy settings. By bridging the technical, operational, and ethical dimensions of LLM adoption, this study contributes to the emerging field of AI governance in finance and offers practical guidance for institutions navigating the transition toward AI-augmented financial ecosystems.