Evaluación de modelos de inteligencia artificial para la automatización de la gestión del cambio en el ciclo de vida del proyecto
Fecha
2025-11-27
Autores
Esquivel Venegas, Freddy
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Nacional (Costa Rica)
Resumen
La presente investigación, titulada "Evaluación comparativa de modelos de inteligencia artificial para la automatización de la gestión del cambio en el ciclo de vida del proyecto", se desarrolló como parte del Trabajo Final de Graduación de la Maestría en Tecnologías de la Información con énfasis en Gestión de Proyectos de la Universidad Nacional de Costa Rica. El estudio tuvo como propósito determinar qué modelos de inteligencia artificial (IA) resultan más adecuados para automatizar funciones de gestión del cambio, y bajo qué marco metodológico de dirección de proyectos se optimiza su aplicación. La investigación adopta un enfoque mixto, con un diseño descriptivo-comparativo y de corte transversal, fundamentado en una revisión documental, un instrumento de evaluación tipo Likert aplicado a expertos y una matriz multicriterio que permitió comparar el desempeño de tres modelos de IA —PMI Infinity, RoBERTa y FLAN-T5— frente a tres marcos metodológicos de gestión de proyectos: PMBOK 7, PRINCE2 7 y SAFe. El análisis evidenció que PRINCE2 7 se consolida como el entorno más favorable para la integración de modelos de IA, mostrando un alto nivel de compatibilidad en dimensiones de gobernanza, trazabilidad y control de cambio. PMBOK 7 y SAFe, aunque sólidos, presentan áreas que requieren mayor madurez organizacional, especialmente en transparencia y facilidad de implementación. Se concluye que ningún modelo ni marco metodológico es universalmente superior, pues la efectividad de la automatización depende del contexto, las capacidades técnicas y el nivel de flexibilidad de cada organización. Como propuesta de implementación y análisis financiero, se simuló el desarrollo de un prototipo funcional basado en el modelo FLAN-T5, orientado a automatizar tareas de gestión del cambio en el caso de estudio de la empresa ficticia SoftDev Solutions, evidenciando mejoras en tiempo de respuesta, consistencia documental y toma de decisiones. Los resultados obtenidos demuestran el potencial de la IA para optimizar la gestión de proyectos, siempre que su integración se acompañe de una adecuada gestión del cambio, fortalecimiento de capacidades digitales y políticas éticas de manejo de datos. Se recomienda a futuras investigaciones explorar la aplicación de modelos generativos avanzados como GPT-4 o Gemini, integrando estrategias de fine-tuning contextualizado con datos reales de proyectos para validar empíricamente los hallazgos de este estudio.
Abstract. This research, entitled ‘Comparative evaluation of artificial intelligence models for the automation of change management in the project life cycle,’ was developed as part of the Final Graduation Project for the Master's Degree in Information Technology with an emphasis on Project Management at the National University of Costa Rica. The purpose of the study was to determine which artificial intelligence (AI) models are most suitable for automating change management functions and under which project management methodological framework their application is optimised. The research adopts a mixed approach, with a descriptive comparative and cross-sectional design, based on a documentary review, a Likert-type assessment tool applied to experts, and a multi-criteria matrix that allowed for the comparison of the performance of three AI models—PMI Infinity, RoBERTa, and FLAN-T5—against three project management methodological frameworks: PMBOK 7, PRINCE2 7, and SAFe. The analysis showed that PRINCE2 7 is consolidating its position as the most favourable environment for the integration of AI models, demonstrating a high level of compatibility in terms of governance, traceability and change control. PMBOK 7 and SAFe, although solid, present areas that require greater organisational maturity, especially in transparency and ease of implementation. It is concluded that no model or methodological framework is universally superior, as the effectiveness of automation depends on the context, technical capabilities and level of flexibility of each organisation. As a proposal for implementation and financial analysis, the development of a functional prototype based on the FLAN-T5 model was simulated, aimed at automating change management tasks in the case study of the fictitious company SoftDev Solutions, showing improvements in response time, document consistency and decision-making. The results obtained demonstrate the potential of AI to optimise project management, provided that its integration is accompanied by adequate change management, strengthening of digital capabilities and ethical data management policies. Future research is recommended to explore the application of advanced generative models such as GPT-4 or Gemini, integrating contextualised fine-tuning strategies with real project data to empirically validate the findings of this study.
Abstract. This research, entitled ‘Comparative evaluation of artificial intelligence models for the automation of change management in the project life cycle,’ was developed as part of the Final Graduation Project for the Master's Degree in Information Technology with an emphasis on Project Management at the National University of Costa Rica. The purpose of the study was to determine which artificial intelligence (AI) models are most suitable for automating change management functions and under which project management methodological framework their application is optimised. The research adopts a mixed approach, with a descriptive comparative and cross-sectional design, based on a documentary review, a Likert-type assessment tool applied to experts, and a multi-criteria matrix that allowed for the comparison of the performance of three AI models—PMI Infinity, RoBERTa, and FLAN-T5—against three project management methodological frameworks: PMBOK 7, PRINCE2 7, and SAFe. The analysis showed that PRINCE2 7 is consolidating its position as the most favourable environment for the integration of AI models, demonstrating a high level of compatibility in terms of governance, traceability and change control. PMBOK 7 and SAFe, although solid, present areas that require greater organisational maturity, especially in transparency and ease of implementation. It is concluded that no model or methodological framework is universally superior, as the effectiveness of automation depends on the context, technical capabilities and level of flexibility of each organisation. As a proposal for implementation and financial analysis, the development of a functional prototype based on the FLAN-T5 model was simulated, aimed at automating change management tasks in the case study of the fictitious company SoftDev Solutions, showing improvements in response time, document consistency and decision-making. The results obtained demonstrate the potential of AI to optimise project management, provided that its integration is accompanied by adequate change management, strengthening of digital capabilities and ethical data management policies. Future research is recommended to explore the application of advanced generative models such as GPT-4 or Gemini, integrating contextualised fine-tuning strategies with real project data to empirically validate the findings of this study.
Descripción
Esquivel Venegas, Freddy (2025). Evaluación de modelos de inteligencia artificial para la automatización de la gestión del cambio en el ciclo de vida del proyecto. [Tesis de Licenciatura]. Universidad Nacional, Heredia, Costa Rica.
Palabras clave
TECNOLOGÍAS DE LA INFORMACIÓN, INTELIGENCIA ARTIFICIAL, AUTOMATIZACIÓN, INFORMATION TECHNOLOGIES, ARTIFICIAL INTELLIGENCE, MODELOS
