Leveraging LLMs to Streamline the Review of Public Funding Applications

TL;DR

The paper demonstrates real-world deployment of AI-assisted evaluation for public funding applications in two EU-funded Portuguese initiatives (IExp and ReClaim), combining GPT-4o–driven processes with human oversight to accelerate reviews while maintaining accountability. It shows significant efficiency gains (e.g., ~20% productivity increase in ReClaim; multi-month reductions in total evaluation time) and enhanced output consistency, supported by quantitative metrics and user feedback. The study analyzes design choices, safety, cost-performance tradeoffs, and integration challenges, offering practical lessons for scaling AI-assisted public-sector review. The findings highlight the potential for large-scale automation to transform administrative workflows, provided governance, data quality, and change management are carefully addressed.

Abstract

Every year, the European Union and its member states allocate millions of euros to fund various development initiatives. However, the increasing number of applications received for these programs often creates significant bottlenecks in evaluation processes, due to limited human capacity. In this work, we detail the real-world deployment of AI-assisted evaluation within the pipeline of two government initiatives: (i) corporate applications aimed at international business expansion, and (ii) citizen reimbursement claims for investments in energy-efficient home improvements. While these two cases involve distinct evaluation procedures, our findings confirm that AI effectively enhanced processing efficiency and reduced workload across both types of applications. Specifically, in the citizen reimbursement claims initiative, our solution increased reviewer productivity by 20.1%, while keeping a negligible false-positive rate based on our test set observations. These improvements resulted in an overall reduction of more than 2 months in the total evaluation time, illustrating the impact of AI-driven automation in large-scale evaluation workflows.

Figures (14)

• Figure 1: Average application evaluation time within the ReClaim initiative, demonstrating a reduction of over 20% following the deployment of our solution.
• Figure 2: The IExp review system leverages GPT-4o to automate the six most time-consuming tasks. Before the analysis, reviewers segment and filter the proposal to avoid overloading the LLM with irrelevant context.
• Figure 3: The ReClaim evaluation system processes a large volume of supporting documents submitted by citizens in various formats. These documents are automatically parsed using GPT-4o, which extracts critical details and performs automated consistency checks, flagging discrepancies for manual reviewer verification.
• Figure 4: Agreement between application summary during the POC and the current application call.
• Figure 5: Agreement on classifying activities as either marketing or organizational, between applicants, reviewers, and the LLM.