Customer Need and Value Proposition
Humanitarian emergency response organisations face persistent challenges delivering foundational emergency management training efficiently whilst maintaining pedagogical effectiveness and organisational knowledge consistency. Traditional classroom delivery requires expensive facilitator time coordinating schedules, travel logistics, and venue arrangements for participants distributed across global operations, creating barriers to universal training access and frequent knowledge reinforcement. Conventional e-learning platforms address logistical constraints but sacrifice engagement and fail to provide adaptive responses addressing individual knowledge gaps or contextual questions emerging during learning progression. The XRisis Pilot 1 investigation explored whether AI-powered conversational avatars embedded within immersive virtual environments could minimise time and cognitive load required to absorb emergency management theoretical knowledge whilst providing interactive question-answering capabilities replicating human facilitator responsiveness without incurring equivalent resource commitments.
Operational Challenge Context
Action Contre la Faim's Emergency Readiness and Response Unit operates across 56 countries requiring emergency roster personnel to maintain current knowledge of emergency management concepts including disaster typology (rapid-onset versus slow-onset versus protracted emergencies), emergency management cycle phases (preparation, alert, response, recovery, mitigation), organisational coordination mechanisms (Emergency Cells, Rapid Emergency Action Teams, Multi-Sectorial Assessments), and Emergency Preparedness and Response Plan development processes. Conventional training delivery concentrates opportunities at regional workshops requiring international travel, creating geographic and institutional inequities where field staff receive substantially less training investment than headquarters personnel despite bearing primary responsibility for implementing emergency responses in affected communities. E-learning modules addressing theoretical content prove logistically efficient but generate completion rates below target thresholds, with participants reporting difficulty maintaining engagement with passive content and frustration when encountering confusion without access to clarification mechanisms beyond static FAQ repositories requiring navigation interrupting learning flow.
Technical Solution Architecture
Pilot 1 implemented an individual AI-powered briefing experience within Unity-based Alaris headquarters virtual environment, integrating CEA's Conversational Virtual Agent platform with Alcatel Lucent Enterprise Rainbow CPaaS communication infrastructure and DFKI Video Call Alternative Appearance avatar representation. Participants entered the virtual headquarters alone following group facilitator induction, engaging in one-to-one dialogue with Mentor Maud, an AI avatar powered by CEA's large language model capabilities grounded in Action Contre la Faim's emergency management documentation, Standard Operating Procedures, and training materials indexed through Retrieval Augmented Generation knowledge base architecture. The virtual environment presented interactive visualisations of crisis scenarios enabling participants to explore emergency contexts ranging from sudden flooding events to gradual drought-induced food insecurity whilst Mentor Maud explained conceptual frameworks contextualised through scenario examples. Natural language processing enabled participants to ask questions in unrestricted phrasing without memorising command vocabulary, with the AI system retrieving relevant documentation passages and generating explanatory responses maintaining organisational terminology consistency whilst adapting explanation depth responsive to participant query complexity.
Validation Methodology
Paris validation workshop (14 May 2025) engaged 8 participants from Action Contre la Faim's internal emergency roster representing Programme Lead, Logistics, and Finance roles with relevant emergency deployment experience (5 of 8 completed deployments) and simulation exercise familiarity (6 of 8 participated in three or more previous exercises). Participants completed Pilot 1 individually following brief technical orientation, interacting with Mentor Maud through natural language dialogue covering emergency management concepts with session duration approximately 25 minutes. Evaluation instruments included added value rating (1-5 scale) specifically assessing contribution to emergency management competency development, System Usability Scale ten-item questionnaire measuring perceived usability, and qualitative feedback through group debrief capturing participant perspectives on interaction quality, knowledge comprehension effectiveness, and comparative value versus conventional e-learning alternatives. Facilitators observed participant behaviour documenting engagement patterns, question types, and technical issues whilst Linagora summarisation agent generated automatic transcripts enabling post-workshop analysis of dialogue patterns and AI response accuracy.
Quantified Outcomes and Metrics
Pilot 1 validation generated differentiated results demonstrating technical capability whilst revealing limited value proposition for theoretical knowledge delivery application. Added value rating averaged 3.2 out of 5 (64% value perception), lowest among three XRisis pilot components, indicating participant perception that immersive AI avatar delivery provided moderate but not compelling benefit over conventional alternatives (XRisis Final Validation Report May 2025, Section 4.3). System Usability Scale assessment produced average 59% across participants, below the 68% threshold considered acceptable for consumer applications but within range typical for complex professional tools during initial deployment, validating usability concerns whilst demonstrating motivated users could achieve productive interaction despite interface friction. Task completion metrics showed 85% successful knowledge comprehension based on post-briefing assessment questions, comparable to conventional e-learning completion rates, suggesting equivalent learning outcomes without incremental effectiveness advantage justifying immersive complexity. AI accuracy rate reached 92% for factually correct responses grounded in organisational documentation, demonstrating reliable knowledge base integration whilst highlighting 8% error rate requiring continued refinement before production deployment confidence.
Strategic Insights and Lessons
Pilot 1 validation generated the single most important strategic lesson shaping SimExBuilder Platform development priorities: theoretical knowledge transfer in professional training contexts benefits from conventional digital modalities rather than immersive XR implementation, enabling strategic resource allocation focusing XR investment exclusively on experiential applications where immersive technology provides unique value through spatial awareness, situated practice, or emotional engagement impossible to replicate through conventional media. Participant feedback consistently indicated that elaborate virtual environment with 3D avatar representation created cognitive overhead when the fundamental task involved listening to conceptual explanations and absorbing theoretical frameworks, with several explicitly stating preference for simpler video presentation or illustrated e-learning module eliminating navigation and interaction mechanics unnecessary for passive knowledge transfer. The finding that desktop computer interfaces provided adequate value without VR headset requirement profoundly influenced deployment strategy, confirming platform should support multiple access modalities rather than mandating expensive hardware investments organisations cannot justify for theoretical training applications. AI dialogue capabilities successfully demonstrated value for enabling natural language question-asking addressing individual knowledge gaps without requiring live facilitator availability, suggesting conversational AI integration merit persists even when immersive 3D environment proves superfluous, informing architectural decisions about which capabilities constitute core platform requirements versus optional enhancements for specific application contexts.
Platform Evolution and Commercial Pathway
Pilot 1 validation evidence directly informed SimExBuilder Platform commercial development decisions prioritising implementation simulation and soft skills practice (Pilot 3 highest-rated component) over comprehensive emergency management cycle coverage including theoretical briefing phases. The platform evolution strategy positions theoretical knowledge delivery through conventional e-learning prerequisites integrated with organisational Learning Management Systems, reserving immersive simulation capabilities for application-focused training requiring situated practice with AI stakeholders, collaborative team coordination, or experiential scenario immersion where XR provides demonstrable incremental value over conventional alternatives. Commercial positioning emphasises selective application rather than universal XR adoption, explicitly communicating to potential clients that SimExBuilder delivers maximum value when deployed strategically for high-value training applications rather than attempting to address all emergency management competency development needs through immersive delivery regardless of appropriateness.
Partnership Model and Attribution
CEA's Conversational Virtual Agent platform provided core AI dialogue capabilities enabling natural language interaction grounded in organisational documentation, demonstrating technical maturity sufficient for professional training deployment whilst revealing refinement needs around accent recognition and response latency. Alcatel Lucent Enterprise Rainbow CPaaS delivered reliable communication infrastructure supporting seamless avatar presence and dialogue transmission across deployment infrastructure. DFKI Video Call Alternative Appearance technology enabled privacy-aware participant representation maintaining social presence without exposing personal video feeds. Action Contre la Faim contributed domain expertise validating emergency management content accuracy, recruited emergency roster participants ensuring operational environment validation, and provided critical feedback about organisational training priorities informing development roadmap decisions. The consortium collaboration model demonstrated European research programme value enabling small-medium enterprises to access cutting-edge AI capabilities developed within broader research frameworks whilst validating technologies through rigorous operational environment testing generating evidence-based commercial pathway decisions.