Strengthening project leadership and financial governance in infrastructure: An AI-driven cash flow forecasting model

Effective infrastructure project leadership requires proactive financial governance, yet leaders routinely struggle to anticipate cash flow volatility and deviations that undermine strategic budgeting and resource allocation. These deviations often stem from the structural heterogeneity of project cash flows, which existing forecasting tools fail to capture. To address this challenge, this study proposes an AI-driven framework to explicitly model the diverse expenditure behaviors for individual projects, leveraging the path signature method with long short-term memory (LSTM) networks. Using a dataset of 282 completed infrastructure projects, we demonstrate that this signature-based approach outperforms varying benchmark models (improving R 2 from 0.849 to 0.896). Depth sensitivity analyses reveal that capturing higher-order temporal dependencies enables more effective anticipation of cash flow volatility. Beyond improving predictive performance, the proposed framework embeds AI-generated forecasts within governance process, equipping project leaders with proactive financial oversight and more reliable budgeting planning. The findings contribute to data-driven project management by demonstrating how advanced predictive analytics can extend leaders’ capacity to navigate financial complexity across heterogeneous infrastructure portfolios.