AI Tour Meeting uses LLM agents to plan group travel collaboratively

The paper proposes AI Tour Meeting, a framework by Daisuke Kikuta that leverages multiple Large Language Model (LLM)-based agents to solve the complex problem of group travel planning. Each agent is instantiated with a distinct persona, enabling them to represent different travelers with unique preferences and constraints. These agents then engage in natural language discussions to collaboratively find an itinerary that satisfies all parties. The framework offers flexible orchestration, allowing users to configure agent personas, discussion workflows, monitoring tools, and LLM deployment. Its primary design is as a simulation environment for studying how multiple LLM agents behave and negotiate during collaborative tasks, specifically group tour planning. The paper includes system validation and analytical results that demonstrate the framework's utility in generating feasible travel plans and revealing emergent behaviors in multi-agent discussions.

The implications of this work extend beyond travel planning. By simulating group decision-making with LLM agents, researchers and developers can better understand how AI systems with diverse goals can collaborate, negotiate, and reach consensus—a capability critical for applications in robotics, virtual assistants, and automated project management. The code availability encourages further experimentation and practical adoption. AI Tour Meeting highlights a growing trend: using LLMs not just as single chatbots but as interacting entities that can model human-like social dynamics. This could lead to more natural and effective AI-assisted group coordination in both personal and enterprise contexts.