Intelligent Management and Artificial Intelligence: Trends, Challenges, and Opportunities, Vol.2
Proceedings on 28th European Conference on Artificial Intelligence ECAI 2025 – InMan Workshop
SPRINT INTO THE FUTURE: INTELLIGENT MANAGEMENT THROUGH LLM-POWERED COMPUTER SIMULATIONS IN AGILE PROJECTS
DOI: 10.18276/978-83-8419-053-1-32
Cytuj
Pełny tekst jest dostępny
| ISBN (online): | 978-83-8419-053-1 |
| Autorzy: |
Tomasz
Wiśniewski
University of Szczecin, Institute of Management, Poland Rafał Szymański University of Szczecin, Institute of Spatial Management and Socio-Economic Geography Poland Marta Starostka-Patyk University of Technology Częstochowa, Częstochowa, Poland |
| Słowa kluczowe: | Large Language Models Agile Project Management Computer Simulation IT Project Planning AI-Driven Decision Making |
| Data publikacji całości: | 2025-10-02 |
| Liczba stron: | 14 (482-495) |
| OAI-PMH: | Metadane |
| Klasyfikacja JEL: | C63 C88 |
| Cited-by (Crossref) ?: | Włącz JavaScript, aby pobrać dane cytowań. |
| |
Abstrakt
Purpose: This study investigates how Large Language Models (LLMs) integrated with computer simulations can enhance intelligent management in Agile Project Management. It examines generative AI’s ability to create simulation models from natural language, enabling faster and more accurate scenario analysis. The research integrates Agile Earned Value Management (AgileEVM) metrics to support decision-making, optimize resources, and improve adaptability in complex IT projects.
Need for the study: Traditional project management tools require specialized skills and struggle with evolving Agile requirements. As IT projects grow in complexity, solutions that merge technical precision with operational flexibility are essential. LLM-generated simulations provide advanced analytics without coding expertise, addressing gaps in speed, accessibility, and adaptability.
Methodology: A framework using GPT-4 generated Python-based simulations from textual project descriptions, incorporating AgileEVM metrics (CPI, SPI, velocity, defect rates). A case study tested three project scenarios with varying cohesion, leadership, and complexity. GPT-4 outputs were benchmarked against ARENA discrete-event simulation and Asana AI across 20 sprints.
Findings: GPT-4 simulations matched ARENA results with