Researchers have developed a risk assessment platform called SENTRY to help large financial institutions manage technology changes. Current methods rely on subjective questionnaires that can be easily manipulated and don't accurately identify high-risk changes. SENTRY uses machine learning to analyze structured data, application dependencies, and historical incident records to provide more accurate and consistent risk assessments. The system combines gradient-boosted decision trees and hybrid retrieval-augmented generation to capture the risk signal in unstructured change request text. In testing, SENTRY achieved a 0.87 ROC AUC score and 85% overall accuracy, detecting high-risk changes at a rate 3.25 times higher than existing methods. --- Why it matters: This matters because current risk assessment methods are often flawed and can lead to major incidents. SENTRY's deterministic approach provides more accurate and consistent results, which is crucial for regulated industries like finance where the consequences of errors can be severe. Source: https://arxiv.org/abs/2608.21203