In a practical test of artificial intelligence’s capacity to alleviate physical-world logistical pressures, Taiwanese technology firm XRSPACE has rolled out its AI avatar platform, ‘Perxona’, to manage front-line tourist inquiries in Okinawa, Japan. The real-world trial, which officially commenced on 9 November 2026, is being conducted in partnership with the Miyakojima Tourism Association and Tokyo-based implementation partner Canvi.

The deployment is a targeted attempt to address the operational strain on regional tourism boards. Operating 24 hours a day and communicating across multiple languages, the Perxona AI concierge is equipped with both vocal capabilities and facial expressions to simulate a more natural interactive experience. It is designed to field routine questions regarding beach conditions, local transport options, and weather forecasts. Project organisers anticipate the system will independently resolve approximately 80 per cent of incoming queries, fundamentally shifting the technology from a passive information repository to a proactive, conversational agent.

For observers in Aotearoa and the broader Asia-Pacific—regions where visitor economies are frequently challenged by seasonal surges, language barriers, and chronic hospitality staff shortages—the Miyakojima pilot serves as a highly relevant blueprint. Rather than aiming for total automation, the stated objective is structural offloading. By delegating high-volume, low-complexity questions to an AI system, human staff are theoretically freed to focus on nuanced local advice, accessibility support, and emergency response. It is a model that could eventually find utility in New Zealand's own i-SITE visitor information networks, provided the technology proves sufficiently reliable.

However, replacing a frontline human interface with a generative avatar introduces distinct risks and failure modes. Tourism relies heavily on accurate, up-to-the-minute local knowledge. If an AI concierge hallucinates a rural bus schedule or misinterprets a critical safety warning about ocean currents during severe weather, the physical consequences for international visitors could be severe. Ensuring that Perxona’s underlying data pipeline remains strictly tethered to verified, real-time civic data—rather than probabilistic guessing—will be the defining test of its safety architecture.

Furthermore, the push towards virtual concierges raises valid concerns about the potential dilution of authentic cultural hospitality, a core economic value proposition for destinations across both Japan and New Zealand. While efficiency is a useful metric, it must be carefully balanced against the loss of genuine human connection that often defines travel.

The trial is scheduled to run until 31 December 2026. As the Asia-Pacific increasingly integrates generative artificial intelligence into its vital economic sectors, the results of this deployment will offer critical empirical data. It will help determine whether multimodal AI avatars can genuinely safeguard the human elements of the visitor economy, or if they merely introduce a new layer of digital friction.

Sources


AI disclosure: Automated systems assisted with research and drafting. Gemini independently peer-reviewed this article before publication. Sources are linked for verification.