A Multi-Agent AI System for Human-in-the-Loop Cognitive Walkthrough

Artificial Intelligence (AI) is increasingly used to support usability evaluation, yet fully automated approaches often reduce methodological transparency and expert control. This demo paper presents an AI-assisted prototype that operationalizes the canonical Cognitive Walkthrough (CW) through a structured multi-agent architecture. The system integrates Golden Standard benchmarking and preserves expert oversight through a human-in-the-loop interaction model. We describe the prototype and the interactive experience that will be demonstrated during the conference session.