AI-Driven Decision Support Systems for Optimizing WorkingCapital and Customer Experience in The U.S.: A TransactionBased Simulation Framework for SMEs

Running an SME often feels like walking a tightrope. You need enough cash to cover day-to-day expenses, but you also want to keep customers happy, and that can be tricky when demand jumps around unexpectedly. Most of the tools out there don’t make this easier. They ‎stick to fixed rules, ignore what your customers are actually doing, and rarely adjust when things change. That means decisions have to be ‎made in real time with little guidance, which can be stressful for managers trying to keep everything balanced. In this study, we explore an ‎AI-based system designed to predict short-term cash-flow stress and guide operational decisions that account for customers, using transaction-level retail data. Weekly financial indicators for each SME are combined with customer behavior signals drawn from purchase patterns, ‎frequency, and inferred payment risk. We test several machine learning models using validation that respects the time order of the data and ‎feed their predictions into a simulation framework that compares simple, risk-aware, and mixed decision strategies. The results show that a ‎straightforward, interpretable classification model can detect cash flow stress almost perfectly, outperforming more complex approaches. ‎Interestingly, while customer behavior features do not make the predictions more accurate, they are crucial when making actual decisions ‎based on those predictions. Simulations of operational policies indicate that hybrid, stress-aware rules outperform naive approaches, both in ‎maintaining revenue and in making balanced approval decisions during stressful periods. In the end, the main contribution of AI here is less ‎about raw predictive power and more about providing structured guidance that incorporates customer behavior to help SMEs manage working capital in uncertain conditions‎.