AI can generate an answer in seconds. The harder challenge is building a product users trust enough to act on. Drawing on her experience as a multi-omics health scientist and founder of iCook, Dr Maria Kardakova explores what it takes to move beyond conversational AI towards intelligent systems that understand real-world context and help users make better decisions. Using nutrition as a case study, Maria will show why an LLM is increasingly infrastructure rather than a product moat, and how effective AI products combine evidence, context, constraints, behavioural feedback and execution. She will explore how systems can learn not only from what users say, but from what they actually do — including the recommendations they repeatedly reject. The talk also examines the “last mile” between recommendation and action: turning advice into something a person can realistically do within their time, budget, preferences and circumstances. For product, AI and growth teams, the message is simple: the chatbot is only the interface. The real competitive advantage is the trusted loop between intelligence and action.