From Text to Model and Back: Lessons from 4 Years of Applied AI in Requirements Engineering
What can Generative AI really do for Requirements Engineering today? This session will deliver practical answers based on four years of building and applying LLM-driven systems in real-world projects. Drawing from hands-on experience and insights from the IREB special interest group on AI, we will explore concrete, battle-tested use cases.
We will demonstrate how AI can enhance established RE methods and help with testing. You will see how to use AI to set up automated testing; to work with structured formats like sentence templates and how to achieve bidirectional transformation: automatically generating BPMN diagrams from user stories and back; See how AI can act as a assistant, detecting inconsistencies, duplicates, and gaps across your textual and model-based requirements.
Value for the audience:
The primary value for the audience, especially testers, is learning how to leverage AI to "shift left" and build quality into the requirements themselves. Attendees will leave with practical techniques to:
Automatically generate testable models (like BPMN) from user stories. This allows them to instantly visualize user flows, identify happy paths, discover edge cases, and derive comprehensive test scenarios far more efficiently than by reading text alone.
Use AI as a proactive testing partner. They will see how AI can automatically detect inconsistencies and contradictions between textual requirements and diagrams, finding requirement bugs before a single line of code is written.
Lay a better foundation for test automation. The session will demonstrate how AI can help transform prose requirements into the structured, logical formats that are essential for writing clear, robust, and automated acceptance tests, bridging the gap between a user story and an executable test script.
Problems addressed:
Stakeholders typically express requirements in natural language, while development teams often rely on formal models like BPMN or UML. Manually creating and synchronizing these different representations is time-consuming and error-prone.
Ensuring quality and consistency in complex requirements specifications. As documentation grows, it becomes nearly impossible for a human to manually detect all duplicates, logical contradictions, or gaps between textual requirements and diagrams.
The disconnect between high-level requirements and the specific, structured formats needed for test automation. Translating a natural-language user story into a format like Gherkin (Given/When/Then) could be done in the RE step.
Talk language: English
Level: Advanced
Target group: Testers, Requirement Engineers, Product Owners, Project Managers
Company:
ireo GmbH
DI, MA Simon Jimenez