Purpose: Following a breast cancer diagnosis, patients face complex, time‑sensitive decisions guided by multidisciplinary team (MDT) input, which can cause confusion and decisional conflict. Standalone large language models (LLMs) are unreliable for clinical decision support due to hallucinations and “black‑box” reasoning. This study aimed to develop and validate a hybrid AI chatbot that provides personalised, guideline‑based treatment summaries in plain language. Methodology: A bespoke web‑based hybrid AI chatbot was developed (in collaboration with IT and Computer Scientists), combining deterministic NCCN‑aligned decision logic—implemented through a structured decision matrix mapping clinical variables to recommended options—with a conversational LLM interface (Claude Sonnet 4.5). This architecture ensures guideline‑compliant outputs delivered with a patient‑friendly tone. Validation is underway using 150 retrospective cases through a two‑stage concordance analysis comparing chatbot outputs with (1) initial clinical decisions and (2) postoperative histopathology‑informed MDT plans (concordance rates and Cohen’s kappa). After validation, supervised machine learning (ML) optimisation using 1000 retrospective MDT cases will enhance classifier accuracy, triage performance, and alignment with real‑world decision patterns. Results: Development of the functional chatbot is complete. Early testing shows consistent generation of personalised and guideline‑concordant summaries. Ongoing validation is assessing concordance, sensitivity, specificity, and factors contributing to discrepancies. Conclusion: A novel, explainable hybrid AI chatbot for breast cancer decision support has been developed. Validation and ML optimisation are in progress, with final results and optimised model planned for presentation by April 2026. This approach aims to deliver a transparent, reproducible, and patient‑centred AI tool with future plans for real-world pilot trial.