Turn ideas into ready-to-refine user stories
As part of my ongoing experiments with AI, I explored how this can support user story creation and backlog management without replacing the essential conversations in Scrum. Think of it as a sparring partner that drafts ideas quickly, so teams can focus on collaboration and effective backlog refinement.
For example, in e-commerce, small improvements add up fast. This simple agent helps by turning a requirement or user journey into user stories with acceptance criteria.
It’s here for inspiration and to learn from: something tangible the team can refine together.
Try writing a requirement, or simple user journey into the box to see it in action. Some examples:
Requrement: Enable customers to filter products by price range so they can quickly find items within their budget.
User Journey:
Persona: Returning shopper
Goal: Reorder their last purchase from the account page
Steps:
-
Log in
-
Go to “Previous Orders”
-
Select last order
-
Add to cart and checkout
Results
User stories and acceptance criteria will be generated here. Note, this could take 10 to 20 seconds to generate.
Suggestions to help
Guidance will be provided here if needed
From Prototype to Working Agent: What Still Needs Solving?
Building this agent has been a learning experience in itself. It already shows how AI can offer alternate views on a requirement, but this is what I’ve learned so far:
-
Context matters. The more detail given (persona, goal, constraints), the stronger the stories and acceptance criteria.
-
Stakeholder and team conversations remain essential. AI can draft, but refinement still needs discussion, negotiation, and shared understanding.
-
Domain knowledge is key. The current agent drafts good stories, but it doesn’t yet “speak” the language of a specific product or industry. A next step could be to link it with a separate AI agent that holds domain knowledge (e.g. common workflows, business rules, compliance standards).
And here’s what could improve in the next iterations:
-
Adding exports to Jira or Azure DevOps.
-
Building a glossary or knowledge base so the agent “remembers” a team’s terminology.
-
Layering in learning features so the agent adapts over time to a specific product or team.