The inability of existing online learning platforms to adapt to individual learner needs remains a fundamental and unresolved challenge in educational technology, contributing to persistently high dropout rates and poor knowledge retention across self-paced digital learning environments. This research proposes and evaluates a conversational AI assessment framework combined with dynamic personalised learning plan generation as a viable solution to this challenge. The study investigates whether natural, dialogue-based learner profiling yields more meaningful personalisation than conventional form-based or performance-data-driven approaches, and whether adaptive, quiz-based feedback integrated with multimodal resource matching improves learner engagement compared to uniform content delivery. A prototype platform named Flint was developed to implement and evaluate these research propositions, employing a dual-AI-engine architecture that addresses reliability and hallucination concerns identified in prior literature. Results demonstrate that conversational profiling successfully captures richer individual learner profiles, that dynamically generated plans align more closely with individual needs than static curricula, and that integrated gamification sustains motivation across extended learning engagements. The findings provide practical evidence to the growing body of research on AI-driven personalised education and demonstrate the feasibility of deploying large language model-powered individualised learning experiences at scale.