FAUNA-GPT: GENERATIVE ARTIFICIAL INTELLIGENCE WITH RAG AND INTENT ROUTING FOR CORPORATE ENVIRONMENTAL EDUCATION AND SUPPORT FOR WILDLIFE MANAGEMENT IN RENEWABLE ENERGY PROJECTS

Read full text Summarise with VIRAT
Authors: Givanildo Ximenes Santana, Camilo Martins Castelo Branco Camurca, Esaú Aguiar Carvalho, João Guilherme de Oliveira Duarte, Barroso Lutif Filho Raimundo, Rickardo Léo Ramos Gomes, Tadeu Dote Sá
Cited by: 0
References: 10

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The energy transition has expanded the deployment of renewable energy projects and, consequently, the relevance of compliance with environmental conditions, legal requirements, and corporate standards associated with wildlife management. Recent evidence suggests that the integration between artificial intelligence (AI) and ESG metrics can accelerate the expansion of renewables and strengthen environmental and social risk governance; however, gaps persist between strategic guidelines and their operationalization by field teams. Traditional corporate environmental education programs often present limitations in standardization, personalization, and the sustainability of behavioral change. In this context, generative artificial intelligence (GAI) based on large language models (LLMs) can offer on-demand guidance and learning; however, its use in critical domains requires governance to reduce unverifiable responses. This study describes the development and preliminary validation of FAUNA-GPT, a customized conversational assistant for guidance and corporate environmental education applied to wildlife management in renewable energy projects. The solution integrates a customized language model, Retrieval-Augmented Generation (RAG), and intent routing (education, incident, audit, contract, and mitigation) to adjust conservatism and mitigate hallucinations. The knowledge base was built from controlled sources, with corporate-standard semantic chunking and a structured FAQ in JSON format. Validation was conducted through simulated scenarios typical of the operational environment, assessing normative adherence, clarity, and adequacy of the level of detail. Preliminary results indicate consistency with source documents and a conservative stance in critical situations, reinforcing the role of the technical professional without replacing them.

Fields of Study & Subjects

Artificial Intelligence in Education Artificial Intelligence Applications Artificial Intelligence in Healthcare and Education

Publication details

Journal / source
Seven Editora eBooks
Publication date
27 March 2026
Language
EN
Document type
Book chapter