Scalable AI For Project Portfolio Management: A Mixed-Methods Study Combining Distributed Computing Benchmarks

Abstract

This study examined the effectiveness of scalable artificial intelligence in enhancing project portfolio management through a quantitative, quasiexperimental design integrating distributed computing benchmarks and decision support evaluation. The research analyzed 312 project instances across six portfolio environments alongside 128 professional respondents involved in portfolio decision-making. The findings revealed that scalable AI systems significantly improved portfolio performance, with regression analysis indicating that key predictors such as predictive accuracy (β = 0.46), system throughput (β = 0.39), and resource utilization efficiency (β = 0.29) had strong positive effects on decision quality, while system latency showed a negative relationship (β = -0.34). The model explained approximately 69% of the variance in portfolio decision outcomes, demonstrating substantial explanatory power. Benchmarking results showed that high scalability configurations achieved throughput levels of 289.4 tasks per second and decision accuracy rates of 88.6%, compared to baseline systems with 182.3 tasks per second and 72.5% accuracy. Correlation analysis further confirmed strong associations between AI performance and portfolio success indicators, with predictive accuracy showing the highest correlation with decision accuracy (r = 0.68) and portfolio responsiveness (r = 0.66). Sub-group analysis revealed that large-scale portfolios experienced more pronounced efficiency gains, while experienced users reported higher trust and satisfaction levels, with mean trust scores increasing from 3.21 to 4.26 across experience groups. Effect size analysis indicated moderate to strong impacts, confirming the practical significance of the findings. Overall, the results demonstrated that scalable AI systems substantially enhanced decision-making accuracy, reduced processing delays, and optimized resource allocation in complex portfolio environments. The study provided empirical evidence that integrating AI with distributed computing infrastructure leads to measurable improvements in both computational performance and portfolio-level outcomes, offering a robust quantitative foundation for advancing intelligent project portfolio management systems.