A retail bank wants to reduce fraudulent transactions by detecting unusual card activity in near...

PMI’s Seven Patterns of AI describes Predictive analytics & decision support as using data-driven learning to anticipate outcomes and support decisions under uncertainty. Fraud detection is a classic predictive use case: the system analyzes historical and current transaction behaviors to estimate the probability of fraud and recommend actions (approve, decline, escalate). In CPMAI-aligned delivery, the project manager ensures the AI capability matches the business objective and defines measurable performance metrics and thresholds (e.g., false positives, fraud loss reduction, detection latency). PMI-CPMAI also emphasizes responsible and trustworthy AI practices—particularly around privacy, governance, and monitoring—because fraud models can affect customers’ access to funds and may introduce bias if training data is skewed. Predictive analytics best fits because it supports classification/risk scoring decisions; the other options focus on interaction (conversational), tailored experiences (hyperpersonalization), or self-directed control (autonomous systems).