IPMA Research Conference: from experience-based to predictive project management
Adding PM AI competency and structured KPIs raised project outcome prediction accuracy from 73.7% to 86.3%.
Lab Director Byung Jun Kang presented research on systematizing project managers' AI competency at the 14th IPMA Research Conference, held at EAN University in Bogotá, Colombia, on project management in a disruptive era.
The study defines PM AI competency in three areas: data literacy, AI collaboration and tool use, and AI ethics and governance. The measurement tool showed Cronbach's α of 0.91, CFI of 0.95 and RMSEA of 0.051. For KPI design it proposes measurable levels, weights that reflect contribution to performance, and a causal sequence of leading, intermediate and outcome indicators.
Adding AI competency to an existing competency model raised explanatory power (R²) from 0.41 to 0.58. A prediction model with AI competency and structured KPIs reached 86.3% accuracy with XGBoost (F1 0.857, AUC-ROC 0.912), against 73.7% (F1 0.703, AUC-ROC 0.781) for the model with existing variables only.
Text mining of 112 project reports found AI-related keywords 18.7 times per document in successful projects versus 11.3 in failed ones, while traditional PM terms such as experience, intuition and judgment showed no meaningful difference. The mixed-methods study combined 315 survey responses, a comparison of four algorithms (Random Forest, XGBoost, neural networks and SVM) and the text analysis.
Original article: Money Today