CODATA AI: uma experiência com modelos de GPT no setor publico

Abstract

The adoption of arti cial intelligence (AI) in the public sector is aligned with the goal of optimizing processes and increasing e ciency in the provision of services to society. In this context, the Data Processing Company of Paraba (CODATA) seeks innovative solutions to improve the work of its employees, in order to favor increased productivity in all departments of the organization. Despite the advancement of AI tools, many available solutions do not fully meet the speci c demands of the public sector, such as security, reliability, and customization. This scenario highlights a research opportunity for the development of an application that lls these gaps, aligning with the institutional needs of the company. This work aims to register the development process of the CODATA AI application, which uses OpenAI models to support and streamline the document writing process by the companys requirements analysts. The proposal is to o er a secure and adaptable solution, di erentiating itself from the options available on the market. The adopted methodology, based on Design Science Research (DSR), involved gathering requirements from the CODATA team, narrating the application development process, and validating the system. In addition, the main technical challenges, lessons learned, limitations, and future prospects of the tool were discussed. The results demonstrated that the application was successful in assisting the work of requirements analysts, automating routines and, consequently, reducing the time spent writing documents. The rst version of the system in production proved to be stable and e cient for this type of task; however, opportunities for improvement and possible scenarios for expanding functionality were identi ed.