An AI model is helping identify when clinicians should be cautious about predicting cancer treatment success. For metastatic non-small cell lung cancer treated with first-line chemoimmunotherapy, the approach flags situations where the system is uncertain, aiming to reduce the risk of overconfident response predictions. The framing is especially relevant in oncology where errors can affect treatment decisions that carry both toxicity and time-to-benefit tradeoffs. By integrating an uncertainty signal, the model offers a pathway toward more clinician-aligned decision support. If validated across additional cohorts, uncertainty-aware oncology analytics could become a standard feature for AI tools embedded into treatment planning workflows.