Enhancing strategic roadmapping through the integration of topic modeling and generative AI
| dc.creator | André Magalhães Gomes | |
| dc.date.accessioned | 2025-07-14T16:52:03Z | |
| dc.date.accessioned | 2025-09-09T01:00:08Z | |
| dc.date.available | 2025-07-14T16:52:03Z | |
| dc.date.issued | 2025-05-28 | |
| dc.description.sponsorship | CNPq - Conselho Nacional de Desenvolvimento Científico e Tecnológico | |
| dc.identifier.uri | https://hdl.handle.net/1843/83547 | |
| dc.language | eng | |
| dc.publisher | Universidade Federal de Minas Gerais | |
| dc.rights | Acesso Aberto | |
| dc.subject | Inteligência artificial | |
| dc.subject | Processamento da linguagem natural (Computação) | |
| dc.subject | Administração | |
| dc.subject.other | Strategic Roadmapping | |
| dc.subject.other | Natural Language Processing | |
| dc.subject.other | Large Language Models | |
| dc.subject.other | Neural Topic Modeling | |
| dc.subject.other | Retrieval Augmented Generation | |
| dc.subject.other | Generative AI | |
| dc.title | Enhancing strategic roadmapping through the integration of topic modeling and generative AI | |
| dc.type | Tese de doutorado | |
| local.contributor.advisor-co1 | Maicon Gouvea de Oliveira | |
| local.contributor.advisor1 | Jonathan Simões Freitas | |
| local.contributor.advisor1Lattes | http://lattes.cnpq.br/5394006847919001 | |
| local.contributor.referee1 | Robert Phaal | |
| local.contributor.referee1 | Youngjung Geum | |
| local.contributor.referee1 | Tiago Alves Schieber de Jesus | |
| local.contributor.referee1 | Leydiana de Sousa Pereira | |
| local.creator.Lattes | http://lattes.cnpq.br/4226121165174499 | |
| local.description.resumo | Contemporary strategic roadmapping practices are increasingly influenced by digitalization and artificial intelligence (AI), yet integrating advanced AI techniques into roadmapping processes remains limited. This thesis investigates how AI-augmented approaches, particularly neural topic modeling and generative AI, can enhance strategic roadmapping. The research begins with a comprehensive systematic review of literature dating back to the early 1980s, using bibliometrics and topic modeling to catalog the evolution of AI applications in roadmapping, revealing significant methodological advancements but also significant gaps in practical implementation. Addressing this research-practice gap, we developed and evaluated an innovative artifact that combines neural topic modeling with generative AI through Retrieval Augmented Generation (RAG) to extract strategically relevant insights for the pre-population phase of roadmapping while ensuring reliability through explicit grounding in source documents. The artifact evolved through two distinct applications: first, an initial proof-of-concept in the AgeTech domain that utilized BERTopic for clustering and topic labeling, demonstrating feasibility with 44% of final roadmap topics derived from quantitative analysis; second, an enhanced implementation incorporating RAG capabilities to produce topic-based reports with supporting scientific references. This refined artifact was applied in AgeTech and validated across eight live case studies, demonstrating how AI-generated topics can effectively augment the market, product, and technology layers of strategic roadmaps in real-world settings. Expert evaluations confirmed high reliability (98.7% of topics deemed reliable) and strategic relevance across different roadmapping contexts. The results demonstrate how AI-augmented roadmapping can enhance strategic foresight while maintaining the visual and collaborative strengths that make traditional roadmapping effective, enabling organizations to develop more comprehensive, evidence-based strategic roadmaps. | |
| local.identifier.orcid | https://orcid.org/0000-0002-2087-2071 | |
| local.publisher.country | Brasil | |
| local.publisher.department | FACE - FACULDADE DE CIENCIAS ECONOMICAS | |
| local.publisher.initials | UFMG | |
| local.publisher.program | Programa de Pós-Graduação em Administração |