Mentor Maud Concept and Implementation
Nuwa has demonstrated Mentor Maud, an AI-powered training avatar integrating CEA's Conversational Virtual Agent platform with XRisis immersive training environments. The character serves as emergency management mentor capable of engaging participants in natural language dialogue about organisational procedures, emergency management concepts, and humanitarian coordination mechanisms, drawing responses from indexed knowledge bases including Action Contre la Faim's Standard Operating Procedures, emergency management documentation, and training materials rather than relying solely on general language model training. Mentor Maud represents the Deputy Country Director role providing arrival briefings to participants newly deployed to fictional Alaris country office, explaining key emergency concepts, walking through emergency management cycle phases, and answering participant questions about organisational structures and response procedures in contextually appropriate conversational style reflecting the character's organisational position and expertise domains.
Technical Integration Achievement
The demonstration validates successful integration of CEA's Conversational Virtual Agent with Alcatel Lucent Enterprise's Rainbow CPaaS communication infrastructure, with the AI character appearing as call participant that trainees can interrupt, question, and engage through natural dialogue rather than predetermined branching conversation trees characteristic of conventional training chatbots. The integration required developing knowledge grounding pipelines processing Action Contre la Faim documentation into vector embeddings enabling semantic retrieval, prompt engineering establishing appropriate character personality and conversational boundaries, and response generation optimisation reducing latency between participant questions and AI replies to acceptable durations preventing awkward silences that would break conversation flow. The system maintains character perspective throughout extended interactions, remembering previous dialogue exchanges and incorporating them into subsequent responses rather than treating each utterance independently, creating coherent conversation arcs that feel authentic rather than disconnected question-answer pairs.
Stakeholder Response and Value Recognition
Action Contre la Faim stakeholders responded with substantial enthusiasm to the Mentor Maud demonstration, recognising potential for AI-powered training assistance that could provide consistent knowledge transfer without requiring live facilitator availability for basic conceptual briefings freeing human facilitators to focus on complex facilitation tasks requiring human judgement and interpersonal sensitivity. The demonstration transformed abstract technical descriptions ("we could integrate conversational AI") into concrete experiential understanding through direct interaction with working implementation, illustrating how tangible prototypes accelerate stakeholder buy-in compared to specification documents or architectural diagrams that remain difficult to evaluate without hands-on experience. Stakeholders identified multiple potential applications beyond initial Pilot 1 arrival briefing including scenario briefing characters explaining context before exercises commence, AI team members participating in collaborative planning scenarios to increase team size without requiring additional human participants, and post-exercise debrief assistants helping participants reflect on decisions and communication approaches through guided dialogue.
Development Implications and Next Steps
The positive response to Mentor Maud has elevated conversational AI integration as development priority, encouraging investment in dialogue quality refinement, knowledge base expansion, and personality configuration capabilities enabling scenario authors to define diverse AI character types beyond emergency management mentors. The team is exploring how AI avatars might serve multiple training roles including stakeholder characters requiring negotiation (local government officials, partner organisation representatives, community leaders), subject matter experts providing specialised knowledge about specific technical domains (water system engineering, nutritional assessment, logistics planning), and facilitator assistants helping manage participant questions during complex scenarios. Technical priorities include improving speech recognition accuracy for non-native English speakers whose accents current systems struggle to understand, reducing response generation latency enabling faster conversational turn-taking, and expanding supported languages beyond English to French, Spanish, and Arabic prevalent in humanitarian operations. The Mentor Maud demonstration illustrates the value of rapid prototyping for capability validation: building working implementations that stakeholders can experience directly generates far more actionable feedback about feature value than specification reviews or capability discussions conducted abstractly without tangible reference points.