Loading...
Publication
Assistente para Consulta de Modelos de Produto
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 5.79 MB | Adobe PDF |
Abstract(s)
É inegável que a transformação digital no setor segurador introduz desafios significativos
em diversos processos internos, em particular na gestão da informação de produtos. Estes
desafios decorrem não apenas da elevada complexidade e variabilidade dos modelos de
produto de seguros, frequentemente sujeitos a alterações, mas também das exigências técnicas
das ferramentas atualmente utilizadas para suportar estes processos, que requerem
um conhecimento aprofundado do domínio. Com os recentes avanços tecnológicos e a crescente
adoção da Inteligência Artificial (IA) Generativa na automatização e otimização de
processos em diversos setores, torna-se imperativo que as empresas do setor segurador se
adaptem ao novo contexto tecnológico. A utilização das potencialidades desta tecnologia
possibilita acelerar e simplificar processos internos que, de outra forma, seriam morosos e,
eventualmente, uma desvantagem competitiva no setor.
A fim de dar resposta a estes desafios, a presente dissertação explora o desenvolvimento
do Product Machine Explorer, um sistema que integra as potencialidades da Inteligência
Artificial Generativa, em particular dos Large Language Models (LLMs) e dos agentes de
IA. Este sistema é integrado com a plataforma de modelação Product Machine (PM),
desenvolvida pela msg Life Iberia, a fim de apoiar a consulta de modelos de produto de
seguros. Na perspetiva do utilizador, o sistema materializa-se num chatbot que, apoiado
em estruturas de dados adequadas e técnicas avançadas de consulta, interpreta pedidos em
linguagem natural e devolve respostas fundamentadas na informação estruturada, permitindo
uma exploração mais eficiente e intuitiva dos modelos existentes.
O desenvolvimento e a melhoria contínua do sistema basearam-se na metodologia Evaluation-
Driven Development (EDD), complementada por atividades de engenharia do conhecimento
para a adaptação da estrutura de dados, bem como por boas práticas de engenharia de software,
de prompt e de contexto associadas ao uso de LLMs e à implementação de sistemas
baseados em agentes. O processo de avaliação foi estruturado a partir da definição de
métricas específicas, cujo cumprimento pôde ser verificado por meio dos resultados obtidos.
Os resultados alcançados, aliados ao feedback recolhido junto de profissionais que testaram
a ferramenta na fase final do seu desenvolvimento, evidenciam a eficácia do sistema em
aumentar a produtividade e a eficiência operacional, reduzir a dependência de conhecimento
especializado e simplificar processos tradicionalmente complexos. Assim, o trabalho constitui
uma contribuição relevante para a transformação tecnológica da gestão de informação no
setor segurador.
It is undeniable that digital transformation in the insurance sector introduces significant challenges in various internal processes, particularly in product information management. These challenges stem not only from the high complexity and variability of insurance product models, which are frequently subject to change, but also from the technical demands of the tools currently used to support these processes, which require extensive domain knowledge. With recent technological advances and the growing adoption of Generative AI (GenAI) for the automation and optimization of processes across multiple sectors, it has become a need for organizations in the insurance sector to adapt to the new technological context. Leveraging the potential of this technology allows internal processes to be streamlined and optimized, which would otherwise be time-consuming and could result in a competitive disadvantage in the sector. This dissertation introduces Product Machine Explorer, a system that leverages the capabilities of Generative AI, specifically Large Language Models (LLMs) and Artificial Intelligence (AI) agents, to address these challenges. This system is integrated with Product Machine (PM), a modeling tool developed by msg Life Iberia, with the aim of supporting the exploration of insurance product models. From the user’s perspective, the system is implemented as a chatbot that, supported by appropriate data structures and advanced query techniques, interprets natural language requests and returns answers grounded in structured information, enabling a more efficient and intuitive exploration of the existing models. The system’s development and continuous improvement were based on the Evaluation- Driven Development (EDD) methodology, complemented by knowledge engineering activities for adapting the data structures, along with best practices in software, prompting and context engineering associated with the use of LLMs and the implementation of agentic systems. The evaluation process was supported by the formulation of evaluation metrics, which were which were later validated by the achieved results. The results, together with feedback collected from professionals who tested the tool in the final development phase, demonstrate the system’s effectiveness in increasing productivity and operational efficiency, reducing dependence on specialized knowledge, and simplifying traditionally complex processes. Therefore, this work represents a significant contribution to the technological transformation of information management in the insurance sector.
It is undeniable that digital transformation in the insurance sector introduces significant challenges in various internal processes, particularly in product information management. These challenges stem not only from the high complexity and variability of insurance product models, which are frequently subject to change, but also from the technical demands of the tools currently used to support these processes, which require extensive domain knowledge. With recent technological advances and the growing adoption of Generative AI (GenAI) for the automation and optimization of processes across multiple sectors, it has become a need for organizations in the insurance sector to adapt to the new technological context. Leveraging the potential of this technology allows internal processes to be streamlined and optimized, which would otherwise be time-consuming and could result in a competitive disadvantage in the sector. This dissertation introduces Product Machine Explorer, a system that leverages the capabilities of Generative AI, specifically Large Language Models (LLMs) and Artificial Intelligence (AI) agents, to address these challenges. This system is integrated with Product Machine (PM), a modeling tool developed by msg Life Iberia, with the aim of supporting the exploration of insurance product models. From the user’s perspective, the system is implemented as a chatbot that, supported by appropriate data structures and advanced query techniques, interprets natural language requests and returns answers grounded in structured information, enabling a more efficient and intuitive exploration of the existing models. The system’s development and continuous improvement were based on the Evaluation- Driven Development (EDD) methodology, complemented by knowledge engineering activities for adapting the data structures, along with best practices in software, prompting and context engineering associated with the use of LLMs and the implementation of agentic systems. The evaluation process was supported by the formulation of evaluation metrics, which were which were later validated by the achieved results. The results, together with feedback collected from professionals who tested the tool in the final development phase, demonstrate the system’s effectiveness in increasing productivity and operational efficiency, reducing dependence on specialized knowledge, and simplifying traditionally complex processes. Therefore, this work represents a significant contribution to the technological transformation of information management in the insurance sector.
Description
Keywords
Exploração de Modelos Produtos de Seguros IA Generativa Large Language Models Agentes de IA Agentic Workflows