A natural, conversational AI interview experience available via voice or web screen. Candidates engage in a realistic, structured interaction that feels human, while validated scores are generated behind the scenes, making assessment both accessible and engaging at scale. This is underpinned by an AI scoring methodology built on structured grids that provide a standardised framework for evaluating responses across behavioural competencies. The approach is validated through human-rater calibration studies, test-retest reliability analysis, and fairness analysis to ensure consistent scoring across diverse candidate groups. Multiple LLMs are benchmarked to identify the best-performing model for each scoring task, with a focus on balancing accuracy, consistency, and robustness. In this presentation, we will share concrete metrics on candidate experience from real applicants who completed the assessment, along with the statistical rigor we used in our scoring methodology.