AI User Story Agent
Situation
At Marriott Vacations Worldwide, authoring ~350 user stories with detailed acceptance criteria, dev notes, Figma references, and Aha! requirement links was time-intensive. Each story required a consistent structure: standardized naming, Gherkin acceptance criteria, screenshots, and cross-references. Maintaining quality and consistency across a large enterprise backlog was a manual bottleneck.
Task
Build a tool to accelerate and standardize user story creation from existing design and requirements artifacts, reducing repetitive formatting work while maintaining quality standards.
Action
- Identified the repetitive pattern in story creation and proposed building a custom AI agent to automate the structured drafting
- Built a custom AI agent that ingested Figma mockup screens, Aha! functional requirements, and initiative specs
- Configured the agent to generate structured drafts including standardized naming, 'As a / I want / So that' descriptions, Gherkin acceptance criteria, dev notes, screenshots, and linked references to Figma designs and Aha! requirements
- Established a refinement workflow where AI-generated drafts were reviewed, adjusted, and created in Jira
Result
Improved delivery speed, consistency, and quality of user story output across the enterprise backlog. The agent handled the repetitive formatting and cross-referencing, freeing time for requirements analysis and stakeholder collaboration.
Key Learnings
- Automating the structured, repetitive parts of story writing lets PMs focus on the analysis and collaboration that actually requires judgment
- Custom AI agents are most effective when given a narrow, well-defined task with clear input/output formats
- AI-assisted workflows still need a human refinement step to catch context that the tool misses