An Intelligent Conversational Agent for Automated Training and Assessment in the Hospitality Industry

The hospitality industry faces ongoing challenges when it comes to training and evaluating staff on critical operational procedures in an efficient and scalable way. Traditional training approaches are often expensive, time-consuming, and difficult to personalize for large teams. This paper presents the design, development, and validation of an intelligent conversational agent powered by Large Language Models (LLMs) aimed at automating training and assessment processes for hotel personnel. The proposed system leverages Google’s Gemini 2.0 Flash model, integrated into a conversational interface built with Chainlit, to support natural language interactions across four core modalities: general inquiries, structured training sessions, practice exercises, and formal assessments. The agent is capable of dynamically generating four types of questions—true/false, multiple choice, open-ended, and scenario-based—using internal hotel documentation as its knowledge base. It automatically evaluates user responses, delivers personalized feedback, and produces detailed performance reports enriched with data visualizations. A Technology Readiness Level 4 (TRL-4) validation was conducted in a controlled laboratory setting, where nine comprehensive functional test cases were executed. The results showed a 95% success rate across all validation criteria, demonstrating the system’s ability to accurately response to general queries, provide targeted training content, generate diverse assessment questions, perform objective evaluations with constructive feedback, etc. This proof of concept highlights the potential of LLM-based conversational agents to transform corporate training in the service industry by offering scalable, personalized, and cost-effective learning solutions. Future work will focus on advancing to TRL-5 validation through deployment with real users in operational hotel environments.