From a vague request to a clean ticket – with an AI agent
“Make it so you can toggle the view.” “The export doesn’t work.” That is how requirements and bug reports arrive day to day – as a free-form sentence in an email, a chat, a form. Before that becomes a usable ticket, someone has to sit down with it: is this a feature or a bug? What is the user story, what are the acceptance criteria, how do you reproduce the error?
That translation is routine, takes time and is a frequent source of misunderstandings between the business side and development.
What the workflow does
It first detects whether the input is a feature or a bug. Then an AI agent produces the matching ticket:
- Feature: summary, user story, subtasks, acceptance criteria.
- Bug: description, steps to reproduce, expected and actual behavior, priority.
The result lands in Jira, structured. A vague “please do this” becomes a work package the team can pick up right away.
Why the AI needs your knowledge
An AI on its own guesses. It does not know your product, your terms or your processes. That is why the agent queries your own documentation first – this is called RAG: the AI searches stored documents and builds its answer on them. So the ticket fits your product instead of generic assumptions. For an honest take on where AI is worth it at all, see When AI is really worth it.
A draft, not a final sign-off
The agent delivers a clean draft – the professional decision stays with a human. That is exactly how I work with AI in general: the machine takes over the tedious structuring, the human reviews and takes responsibility (more on this in How I build with AI). You save the boring half without giving up control.
Try it yourself
The workflow is available as a free n8n template in my template repository – including the RAG connection that lets the AI use your own documents.
If you are wondering where, on your path from idea to finished ticket, AI could save real time, let’s look at it together.