A practical, vendor-neutral toolkit for public institutions that want employees to use generative AI tools in everyday written work while protecting institutional information and keeping a person accountable for every official product. It is designed for document-intensive teams such as internal control, internal audit, oversight, legal support, procurement, and program administration. The toolkit has five parts that can be used separately or together: Risk Classification Worksheet: four classes of AI use, from low risk to prohibited, with a step-by-step classification guide and a use register. Human-Review Protocol: de-identification and redaction before AI use, structured instructions with the PACTO method, a technical review checklist, validation and signature, and specialized assistants for repetitive processes. Role-Based Curriculum Outline: a twelve-hour course in three sessions with step-by-step hands-on practice. Pilot Measurement Template: baselines defined in advance and reports that keep observations separate from causal conclusions. Facilitator Guide: preparing internal staff to deliver later cohorts. The toolkit systematizes practices the author developed in Brazilian public administration from 2024 to 2026, described in Gomes (2026), arXiv:2606.01517.
No takes yet. Share an insight, caveat, or question.
Vinicius Santana Gomes (2026) studied this question.