AUTOMAÇÃO INTELIGENTE NO E-COMMERCE: VIABILIDADE DA INTEGRAÇÃO DO CHATGPT COM API E TOKENS DE SEGURANÇA

Customer service automation through Artificial Intelligence (AI) has gained global prominence, but its implementation in Brazil faces challenges such as linguistic diversity and compliance with the LGPD (General Data Protection Law). This study examines the technical integration of ChatGPT via API with tokens, evaluating its feasibility and impacts on operational efficiency and customer experience. Using an applied methodology, a prototype was developed in Python with FastAPI and JWT tokens, tested in three Brazilian e-commerce companies (fashion, electronics, and food) with 1,200 simulated and real interactions. Results showed a 40% reduction in average response time (from 12 to 7 seconds), a 25% increase in customer satisfaction (NPS from 6.2 to 8.1), and an 85% automated resolution rate, compared to 60% in traditional systems. Critical challenges were identified, such as the need to adapt the model to regional dialects (e.g., lexical variations between "bolacha" and "biscoito") and additional costs of 15% for security infrastructure to manage tokens. It is concluded that the integration is viable and scalable but requires investments in local data training and encryption protocols. The study provides a practical roadmap for SMEs, balancing innovation and regulatory compliance, and reinforces the role of IA serves to complement human capabilities, not to substitute them.