An AI feedback agent instead of a survey

In one line: A conversational agent asks follow-up questions, detects language, and returns structured insights — replacing the survey form pattern for the same traffic.

Context

Static forms produced shallow answers and high abandonment. I wanted qualitative depth without manually reading every free-text box.

Why it was worth building

Survey SaaS tools optimize for form completion, not insight quality. A custom n8n + GPT-4 loop trades platform convenience for conversation shape and structured extraction under my control.

What I built

• n8n orchestration — webhook intake, branching, and storage of structured results.
• GPT-4 dialog — adaptive follow-ups instead of fixed question lists; multilingual detection.
• Structured outputs — categories and priorities suitable for product review, not raw prose only.

How it was adopted

Deployed as a production feedback path integrated with existing site infrastructure; responses feed review workflows rather than a third-party survey dashboard.

Effect

Richer signals per session than one-shot forms; no dependency on a standalone survey SKU for the same workflow class.

Where AI fits now

This is an automation pattern: deterministic n8n steps with a model in the middle. Today I would add eval fixtures for prompt drift and route cheaper models for classification — same architecture as the AI automation in production case.