Contents:
Overview
Maintaining a healthy, "Dev-Ready" backlog is often the most time-consuming part of product management. Between vague descriptions, missing estimations, and stale comments, your story map can quickly become cluttered.
By connecting StoriesOnBoard to an AI agent via the Model Context Protocol (MCP), you can transform your AI into an intelligent Backlog Grooming Assistant. This assistant can audit your map, synthesize team discussions into requirements, and ensure every card is ready for the next sprint.
What can the Assistant do
Unlike a standard AI that simply generates text, an MCP-powered assistant has direct access to your story map hierarchy. It can:
Audit for Gaps: Identify cards missing story points, priority, or descriptions.
Contextual Enrichment: Read existing comments and attachments to write detailed requirements.
Automated Refinement: Standardize user story formats and add testable Acceptance Criteria (AC).
Visual Cleanup: Use color-coding to highlight cards that are "Ready" vs. "Blocked."
The product backlog
We are focusing on "Release 2". Backlog items has missing details at this point.
Backlog Refinement & Enrichment Prompt
We use the AI agent to analyze the backlog and find backlog items that has missing details such as user story or acceptance criteria or story points.
We used the following prompt:
Act as a Product Analyst to audit and enrich the story map. Your goal is to identify 'thin' cards and use existing context to make them detailed and actionable. Follow this multi-step workflow:
Audit the Map: Scan the board. Identify any white cards (subtasks) that are missing descriptions, estimations (story points), or priority labels in "Release 2".Deep Dive for Context: For every 'thin' card identified, read its full history. Look specifically at:Existing comments or discussions.Attachment names or linked web resources.The context of its 'parent' task.
Synthesize & Enrich: > * If a card is missing a description, write a concise one in the 'As a user, I want to... so that...' format based on the context you found.If it is missing an estimation, suggest a story point value (1, 2, 3, 5, or 8) based on the task complexity.
Execute Updates: Apply these details to the existing cards.Refine with New Cards (Optional): If a task is too broad (e.g., 'Fix all bugs'), break it down into 3-4 specific, smaller subtasks and create them in the "Release 2", add the "New item" card annotation to these new cards and write clear descriptions. Before applying the changes, provide a list of the cards you intend to enrich and your suggested descriptions for my approval.
With this update the AI assistant identified the large items in "Release 2" and broke them into smaller manageable items and included the user story and added user story points.
Acceptance Criteria Enrichment Prompt
Preparing back items for the development team to reach dev-ready stage.
We used the following prompt:
I need to prepare 'Release 2' for development. Your goal is to generate clear, testable Acceptance Criteria (AC) for every card currently assigned to that release. Execute the following steps:
Identify the Scope: find all cards currently assigned to the release named 'Release 2'.Gather Context: For each card in this release, analyze the existing description, parent Task (Yellow card) context, and any team comments.Draft Acceptance Criteria: For every card, generate a list of 3–5 'Given/When/Then' or bulleted Acceptance Criteria. Ensure these are:Atomic: Focused on a single piece of functionality.Testable: A QA engineer should be able to verify them clearly.Detailed: Include edge cases where appropriate (e.g., error handling or validation).
Update the Cards: append these Acceptance Criteria to the description field of each card.Visual Confirmation: Change the card color of these enriched cards to [Specify Color, e.g., Green] so I can visually identify which ones are 'Dev-Ready.' Before updating, present a summary of the AC you've drafted for the first 3 cards so I can verify the quality and tone."
As a result, the AI assistant added acceptance criteria to each user story and marked dev-ready items with green color.
Prompt examples for backlog refinement
To start grooming your backlog, use these prompts in your MCP-enabled AI tool (like Claude Desktop or Cursor).
1. The "Definition of Ready" Audit
Scan my story map and find all white cards in 'Release 1' that are missing a description or story point estimation. List them for me and suggest a prioritized plan for how we should fill those gaps based on the parent task's importance.
2. Dependency & Edge-Case Discovery
Look at our 'Payment Integration' and 'User Profile' activities. Based on the current subtasks, identify any missing 'edge-case' cards we might have overlooked—such as 'Expired Session during Checkout' or 'International Currency formatting.' If you find gaps, draft them as 'Low' priority subtasks and flag them for my review by highlighting them with red color.
3. The "Epic Splitter"
I have a task called 'User Authentication' that feels too large. Analyze its current subtasks and suggest how we can break it down into smaller, atomic stories. Add these new stories to the map with a 'High' priority label.
4. Estimation & Effort Alignment
Compare all subtasks in the 'MVP' release. Identify cards that have very similar descriptions but different story point estimations (e.g., one is a 2, another is an 8). Add a comment to these cards flagging the inconsistency, and suggest a 'normalized' estimation based on the complexity of other groomed cards on the board.
Learn more about: