‘Malha Fina de Convênios’: an AI application for auditing in the Brazilian federal government
Bernardo de Abreu Guelber Fajardo, Sergio Neiva · Public Money & Management · 2025
IMPACTThis article offers valuable insights for public sector auditors, financial managers, and oversight professionals—particularly those operating in federal or decentralized government structures. By analysing Brazil’s ‘Malha Fina de Convênios’ algorithm, the authors show how supervised machine learning, especially Random Forest models, significantly improves audit efficiency, reduces processing costs and mitigates backlogs in discretionary financial transfers. Generalized artificial intelligence (AI) models outperformed agency-specific ones due to their broader data coverage and superior recognition of bureaucratic patterns. These findings underscore the importance of balanced, comprehensive datasets and the need for well-designed governance frameworks to ensure ethical and accountable AI adoption. Crucially, the article reinforces the evolving role of auditors—not as passive reviewers, but as strategic actors in performance auditing and preventive control. Practitioners will benefit from a replicable model that aligns automation with human oversight, enhancing transparency, resource allocation and the overall quality of public financial management.