SES/CONT and UCI/SEDET — A Structured AI-Training Method Cuts Case-Processing Time Up to 50% in Two Brasília Government Control Units

In 2024 and 2025, two internal-control units in Brazil's Federal District government — the Sectoral Internal Control Office of the Department of Health (SES/CONT) and the Internal Control Unit of the Department of Economic Development, Labour and Income (UCI/SEDET) — piloted a structured, four-layer training method for using free generative-AI tools in day-to-day audit work, rather than procuring new AI software.

The method, developed by AI trainer Vinicius Santana Gomes at the Federal District's Escola de Governo, focused on teaching auditors how to prompt, verify and document AI-assisted analysis within existing data-protection rules and public-administration principles.

SES/CONT (2024) cut average case-processing time by 18.2%. UCI/SEDET (2025) cut processing time by 50% (from 34 to 17 days), increased the number of technical reports issued by 85%, issued 286 formal recommendations, and reviewed cases with a combined value of US$94.8 million.

The results are self-reported in a single-author preprint and corroborated only by the units' own internal management reports (SEI Doc. 197403428 and SEI Doc. 193384267); they have not yet been independently replicated or peer-reviewed, and the training model has not been reported as rolled out beyond these two units.

Read the full analysis: https://egov.df.gov.br/inteligencia-artificial-e-tema-de-capacitacao-na-escola-de-governo/