Find out how Medcast is building trustworthy medical-grade AI in this episode of Talking HealthTech, recorded live at the Digital Health Festival 2026 in Melbourne, Australia's premier gathering for health innovation, technology, and policy. Peter Birch sits down with Dr Stephen Barnett, CEO of Medcast, to explore what it genuinely takes to make AI safe, reliable, and fit for purpose in a clinical setting. With a background spanning general practice, research, and digital health innovation, Barnett brings a grounded and practical perspective to one of the most talked-about topics in healthcare right now. Together, they unpack the concept of knowledge translation, the persistent gap between cutting-edge AI and trustworthy clinical tools, and why large language models alone are not enough. The conversation covers retrieval-augmented generation, the risks of AI hallucinations, the importance of curated and validated data sources, and the orchestration layer that sits between a powerful model and a safe medical answer. They also discuss how clinical workflows are changing, the shift in how doctors use digital tools in front of patients, and the emerging conversation around patient-facing AI. Key Takeaways 🧑⚕️ Knowledge translation in healthcare relies on getting the right information to the right person at the right time, impacting both clinician learning and patient outcomes. 🤖 Large language models require an orchestration layer and data curation processes to be considered "medical grade" and trustworthy for clinical use. 🔎 Retrieval-augmented generation and other governance measures help address hallucinations and bias in AI-powered healthcare solutions, supporting quality, safety, and auditability. 💡 Medluma, Medcast's AI platform, addresses compliance, standardisation, and onboarding within healthcare organisations, while empowering clinicians with trusted knowledge access. 📱 Patient-clinician trust depends on using validated, professional tools for information sourcing, and future AI integrations may also support patient self-service within set guardrails. Timestamps 00:00 - Introduction and guest overview 00:44 - Medcast origins and purpose 02:54 - Obsession with AI and knowledge translation 03:49 - Medical safety in language models 06:13 - Techniques for trustworthy AI 10:01 - Adapting to rapid AI changes 13:32 - Knowledge translation across clinical and non-clinical roles 14:49 - Patient trust and professional boundaries 18:22 - Medcast and Medluma future roadmap Check out the episode and full show notes on the Talking HealthTech website. If you’re enjoying the show and want access to exclusive healthtech discussions, meetups, and member-only content, you can learn more about becoming a THT+ Solo Member here: talkinghealth.tech/solo_shownotes And if this episode was useful, leaving a review or sharing it with someone in the industry always helps. Mentioned in this episode: THT+ Company Partnership Learn more about THT+ Company Partnership options for start-ups, scale-ups and enterprise digital health companies looking for visibility, content, community access and industry connection: talkinghealthtech.com/partners. THT+ Company Partnership Learn more about THT+ Company Partnership options for start-ups, scale-ups and enterprise digital health companies looking for visibility, content, community access and industry connection: talkinghealthtech.com/partners. THT+ Company Partnership Learn more about THT+ Company Partnership options for start-ups, scale-ups and enterprise digital health companies looking for visibility, content, community access and industry connection: talkinghealthtech.com/partners.