UUX aspects of a conversational AI assistant

This dissertation explores usability and use experience (UUX) aspects of a conversational AI assistant designed for food ordering within a messaging environment. A prototype built on Telegram was created to guide users through a step-by-step ordering process (craving → budget → location → shortlist → confirmation → payment). The assistant was developed as an n8n workflow that integrates a large language model with persistent conversation history using PostgreSQL/Supabase, along with tools for food searches, saving orders, and generating payment links. UUX features were integrated into prompts and tool contracts, including fixed three-option suggestions, numbered choices, brief confirmations to maintain flow, explicit checkout steps, trust indicators for payment links, and strategies to handle missing information or no-results scenarios. The system was tested with 10 participants through task-based sessions, post-task surveys, and standardized questionnaires (SUS, UEQ-S, NASA-TLX), supplemented by open-ended feedback. Findings indicate that the staged process enhances clarity and reduces decision-making effort, while transparent checkout procedures and consistent summaries increase perceived control and trust. Remaining challenges mostly relate to edge cases in interpreting user intent and system refinement. This work offers a practical blueprint for creating UUX-centric conversational assistants using a low-code, tool-enhanced architecture, along with insights to inform future improvements.