I built this after repeatedly seeing good product or engineering ideas stall before they ever reached the backlog.
Feature requests usually arrive as messy text: emails, forms, Slack messages. By the time they’re rewritten into Jira tickets, a lot of context is lost, or the request arrives too late to matter.
This is an n8n workflow that: - Takes a raw feature request in a structured way - Analyzes internal product documents and architecture - Uses an LLM to generate a concise architectural summary and technical task breakdown - Enforces a strict JSON schema to avoid hallucinated structure - Presents the output in a review UI - Requires explicit human approval before creating Jira tickets
Nothing is auto-shipped. The goal is speed without losing control.
The full workflow and setup notes are here: https://github.com/kavishsekhri/AI-Feature-Intake-Engine
There’s also a short demo video in the README showing the end-to-end flow.
Happy to answer questions or discuss trade-offs.
kavishsekhri•1h ago
Without strict parsing and schema validation, the UI broke and downstream systems behaved unpredictably. Most of the workflow is defensive engineering, not prompting.
If anyone’s done similar human-in-the-loop systems, I’d love to compare notes.