Customer Need and Value Proposition
Open-air archaeological museums across Europe face persistent operational challenges balancing visitor experience quality against constrained staff resources, particularly during summer high-season when tourism demand concentrates routine information queries into brief periods overwhelming reception capacity. Āraiši Ezerpils Archaeological Park in Latvia, featuring Europe's only reconstructed 9th-10th century fortified lake settlement, exemplifies these pressures with small permanent staff struggling to efficiently address visitor facility location queries, ticket pricing questions, event schedule requests, and directions to distributed archaeological attractions across extensive outdoor terrain. Conventional solutions including digital information kiosks, multilingual signage systems, or expanded staffing prove economically infeasible within heritage sector budget constraints. The VAARHeT Pilot 1 investigation explored whether voice-activated mobile AR welcome avatars could minimise staff resource burden and visitor wait times whilst maintaining cultural heritage institutional quality standards through AI-powered natural language information retrieval accessible on personal smartphones.
Operational Challenge Context
Āraiši Ezerpils faces typical open-air museum challenges including seasonal tourism concentration (June-August), limited permanent staff capacity (3-5 employees), extensive site geography (40 hectares with multiple reconstruction areas), multilingual visitor populations (Latvian, English, German, Russian speakers), and institutional mission requiring educational interpretation quality exceeding simple directional assistance. Reception bottlenecks emerge when multiple visitors simultaneously require basic information (bathroom locations, ticket prices, event schedules) preventing staff from providing higher-value interpretive guidance about archaeological significance, reconstruction methodologies, or historical context. Shoulder seasons and off-peak periods suffer minimal interpretive support despite institutional desire to extend visitor engagement throughout operational calendar. Budget constraints prevent investment in digital infrastructure requiring ongoing maintenance or seasonal staff expansion creating fixed cost commitments unsustainable beyond high-season revenue periods.
Technical Solution Architecture
Pilot 1 implemented Unity-based mobile AR application deployed on Samsung Galaxy Note10+ 5G Android devices with server-side VOXReality component inference processing on NVIDIA A10G GPU infrastructure hosted in EU-compliant German cloud facilities. Visitor interaction began with ARCore plane detection enabling ground surface scanning and 3D avatar placement at selected position, anchoring virtual guide within physical museum space. The avatar interface combined 3D character model with contemporary styling, text display panels rendering AI-generated responses, and push-to-talk button activating microphone capture for voice input. VOXReality Automatic Speech Recognition operated locally converting spoken questions into text transcriptions with processing contributing to median 1766 millisecond end-to-end latency from voice capture to response display. Transcribed text transmitted to cloud servers for Intent Classification analysing queries against eight authorised intents (facility locations, ticket pricing, event schedules, attraction listings, directions, opening hours, safety restrictions, general museum context), triggering VOXReality Dialogue System implementing Retrieval Augmented Generation querying curated knowledge base comprising museum documentation, visitor FAQs, facility information, and historical context provided by Āraiši staff.
Validation Methodology
Comprehensive usability testing at Āraiši Ezerpils (14-16 July 2025) engaged 39 participants recruited from museum network representing validated visitor persona: predominantly female aged 30-50 with professional occupations and higher education backgrounds, Latvian native speakers with good English comprehension, daily mobile phone users with approximately half having prior AR application exposure, over 75% having used conversational chatbots, and classified as regular or frequent museum visitors providing informed perspectives on heritage experience quality expectations. Cordula Hansen from Technical Art Services designed validation methodology following industry UX research standards with Maynooth University ethical approval. Test sessions followed moderated in-person protocols measuring task completion across eight discrete tasks (application launch, language selection, avatar placement, information retrieval for ticket prices, bathroom locations, event schedules, attraction listings, directions) with completion categorised as successful, requiring assistance, user abandonment, or technical failure. Post-test questionnaires measured Net Promoter Score, user experience dimensions (first impressions, cultural appropriateness, voice interaction naturalness, information efficiency, relevance, accuracy), and qualitative free-text feedback triangulating observed behaviour with self-reported experience.
Quantified Outcomes and Metrics
Validation produced differentiated results demonstrating functional technical capability whilst revealing insufficient value proposition and accuracy limitations. Net Promoter Score achieved 16 (16 promoters, 12 passives, 10 detractors from 38 respondents) indicating positive overall experience without compelling recommendation likelihood, positioning avatar in "needs improvement" category rather than "great" or "excellent" range typical for successful museum technology deployments (VAARHeT Pilot 1 Validation Report July 2025). Task completion rates varied 64-90% dependent on query complexity: ticket prices 90%, bathroom locations 79%, event schedules 64%, attraction listings 85%, directions 77%, demonstrating generally functional operation whilst revealing accuracy degradation for time-sensitive information requiring precise current data versus static facility details tolerating approximate responses. AI accuracy rate reached only 75% for factual correctness, substantially insufficient for heritage institutional credibility requirements where approximately 1 in 4 responses containing errors or hallucinations undermines visitor trust and institutional reputation. Response latency achieved 92.1% acceptable perception rating with median 1766 milliseconds end-to-end processing, comfortably meeting subjective quality thresholds despite opportunities for optimisation.
Strategic Insights and Lessons
Pilot 1 validation generated four strategic lessons fundamentally shaping Culturama Platform development priorities. First, theoretical content delivery in cultural heritage contexts benefits from conventional digital modalities (website FAQs, simple chatbots, digital signage) rather than immersive AR implementation, enabling strategic resource allocation focusing XR investment exclusively on experiential applications where immersive technology provides unique value through spatial awareness or temporal transportation impossible to replicate through conventional media. Second, heritage institutions require substantially higher AI accuracy thresholds than commercial applications due to reputational risk from factual misinformation, with 75% accuracy proving absolutely insufficient despite potentially acceptable performance in commercial customer service contexts where occasional errors prove tolerable if overall utility remains high, requiring curator-validated knowledge bases, strict RAG guardrails, and explicit uncertainty communication rather than speculative response generation. Third, minority and regional language support proves essential for European heritage applications serving linguistically diverse visitor populations, with English-only deployment excluding local communities contradicting cultural missions to serve regional populations as primary constituencies. Fourth, voice interaction provides genuine convenience value for hands-free information access and natural question phrasing, but value realisation depends critically on technical reliability where speech recognition failures or response generation inaccuracies undermine user trust more severely than equivalent failures in manual text interfaces.
Platform Evolution and Commercial Pathway
Pilot 1 validation evidence directly informed Culturama Platform development decisions pivoting away from avatar information delivery automation entirely, with strategic analysis concluding simple museum chatbot accessible through website without AR representation would deliver equivalent or superior functional value at substantially lower development cost and technical complexity whilst eliminating avatar placement friction, reducing mobile application size and performance requirements, and avoiding expectations mismatch where 3D avatar representation creates expectation of sophisticated interaction that current AI capabilities cannot reliably satisfy. Culturama commercial positioning focuses immersive XR investment on experiential heritage interpretation where technology provides unique capabilities (virtual site exploration enabling access for mobility-limited visitors, temporal reconstruction showing historical periods impossible to experience through conventional media, interactive archaeological process demonstration revealing construction techniques through 3D spatial manipulation) whilst leveraging conventional digital solutions for theoretical knowledge delivery and routine information access where simpler approaches serve user needs without immersive complexity overhead.
Partnership Model and Attribution
Āraiši Ezerpils Archaeological Park contributed operational environment validation access, visitor participant recruitment from museum network, heritage domain expertise validating historical content accuracy, and strategic feedback about institutional priorities informing development roadmap decisions. Cordula Hansen from Technical Art Services provided critical liaison function translating between technical capability and heritage institutional requirements, bringing user experience research methodology, digital cultural heritage technology expertise, and heritage preservation methodological knowledge bridging XR Ireland technical vocabulary and museum professional operational language. F6S Innovation provided third-party coordination support through VOXReality consortium structure facilitating mentor connections and administrative oversight. Maggioli Group as VOXReality consortium leader provided programme governance, funding coordination, and access to European research partner technologies enabling SME leverage of cutting-edge AI capabilities.