<![CDATA[The development of Artificial Intelligence (AI)-based computer systems in large-scale data processing environments introduces significant uncertainty regarding operational success rates. AI systems handling thousands to millions of processing requests daily are vulnerable to failures caused by workload variations, limited computational capacity, and model complexity. This study applies a quantitative approach using Monte Carlo Simulation to predict the performance and reliability of AI-based computer systems in data processing by utilizing probability concepts, discrete probability distributions, cumulative probabilities, expected value analysis, and Mean Absolute Percentage Error (MAPE). The research data consist of 250 historical processing observations classified into four categories: optimal success (58%), delayed success (24%), temporary failure with retry (12%), and total failure (6%). The expected processing time was calculated as E(X) = 3.816 seconds. A Monte Carlo Simulation with 10,000 iterations produced an average MAPE of 0.58%, indicating excellent predictive accuracy since it is well below the 5% tolerance threshold. Furthermore, under a workload of 50,000 requests per day, approximately 8,950 requests (17.9%) were estimated to potentially affect service reliability. The findings demonstrate that Monte Carlo Simulation is an effective quantitative tool for capacity planning and operational risk mitigation in AI-based computer systems.]]>
