Conversational AI for SEQ Forms

Every semester, students fill in the same Likert-scale SEQ forms. The questions are dry, the format is rigid, and the responses are shallow. We built a conversational AI agent on the Grexx low-code platform to replace that static form with a natural-language dialogue — one where students answer in their own words, and the system maps their response to a Likert score silently in the background, via emoji sentiment detection.

The biggest design challenge wasn't technical — it was tone. Early iterations produced responses that were too brief or off-topic. The breakthrough was a tightly engineered agent prompt: neutral, encouraging elaboration without leading, and strictly topic-scoped to prevent deviations. We also hit an unexpected GDPR wall when the public homepage allowed open registration, and worked directly with the Grexx team to implement proper authentication.

Case owner Yannis Haritos noted the team was "conceptually strong, more focused on the big picture than just the low-code development." The AI choosing an emoji based on sentiment — rather than just mapping a keyword — was called out as a standout idea.