AI-Powered Requirement Generator Software
Read time:
10 min
Client:
Collins Aerospace
Industry:
Aviation & NLP Engineering
Start:
End:
Duration:
32 Weeks
Project Summary:
Qubric is a Visual Studio Code extension that automates the creation of Low-Level Requirements (LLRs) based on High-Level Requirements (HLRs) and source code. The software ensures the traceability of generated LLRs, provides actionable metrics and allows engineers to export the LLRs to their requirement management environments such as Jama, Polarion, and IBM Doors.
My Role:
I designed the end-to-end UX/UI for Qubric’s AI-powered requirement generation workflow inside VS Code. I created user flows, interactive prototypes, and high-fidelity designs, and conducted interviews and usability tests with engineers from Collins Aerospace to validate the experience against real engineering workflows.
Team:
Collaborated with the product team, engineers and Collins Aerospace engineers
The Challenge &
My Approach
The Challenge
System engineers face significant challenges when it comes to writing requirements, with the most time-consuming and difficult part being the initial drafting process. While editing existing requirements is relatively straightforward, starting from scratch is a daunting task. Additionally, engineers need to maintain clear traceability between low-level requirements (LLRs) and high-level requirements (HLRs), which can become cumbersome without the right tools. Furthermore, leaving their natural work environment, such as VS Code, to write and edit requirements disrupts their workflow and reduces productivity. There is a need for a solution that simplifies requirement generation, maintains traceability, and integrates seamlessly into their existing tools.
My Approach
My approach was to design Qubric as an AI-assisted workflow, not a black-box generator. Since the product used LLM and NLP capabilities to create low-level requirements from high-level requirements and source code, the experience needed to make the output understandable, traceable, and easy to validate.
I focused on helping engineers review generated requirements, understand how they mapped back to HLRs and code, compare outputs against existing requirements, and stay in control through editing, regeneration, and validation inside VS Code.
Research & Key Insights
1- User Interviews
Through interviews with Collins Aerospace engineers, I uncovered the following key insights:
2- Contextual Inquiry + Data Analysis
I conducted contextual inquiries by observing system engineers in their natural work environment. During these sessions, I captured:
- Key moments in their workflows,
- How they write, edit, and manage requirements in VS Code and Jama,
- Additionally, I analyzed sample requirement documents to identify trends, errors, and inefficiencies in their current processes.
Key Findings:
3- Customer Journey Map
4- User Goals and Need Statements
Strategy & Design Process
1- Brainstorming Ideas
2- Prioritization Matrix
3- Selected Ideas
4- User Flow
5- Low-Fidelity Wireframing & Prototyping
6- User Testing
Objective: Validate the usability of Qubric and identify areas for improvement.
Participants: 5 Collins Engineers
Tasks: Set up Qubric, Import input files and generate requirements, Engage with generated outputs
Methodology: Remote usability testing.
Key Findings:
Final Solution
Results and Impact
Results and Impact
Metrics
Setup Completion Rate: 100% of engineers successfully set up Qubric after the first version was released.
Time Efficiency: Qubric reduced the time spent in the requirement writing process by 40%.
Requirement Quality: 85% of generated requirements scored 4 over 5 (Good) or higher in QVscribe's quality evaluation.
Error Reduction: Similarity and consistency errors were decreased by 70%.
Export Success: 90% of the time, the requirements generated were exported to Jama without additional errors.
Impact
- Engineers experienced streamlined onboarding, with a smooth setup process and no initial barriers to use.
- Increased efficiency in generating and managing requirements, allowing engineers to save time for more critical tasks.
- Enhanced requirement quality, with a significant majority meeting high-quality standards.
- Greater team collaboration, supported by features like the Manifest file for reusability and alignment with project standards.
- Improved workflow integration, ensuring engineers could focus on refining outputs without unnecessary disruptions.
Reflections
The Qubric project has been a significant milestone in enhancing the efficiency and accuracy of requirement generation for Collins engineers. Through usability testing and iterations, we were able to design a tool that aligns with engineers’ workflows while addressing their key pain points.
What Worked Well
1- The successful integration of Qubric into VS Code ensured minimal tool-switching and streamlined the engineers’ existing processes.
2- Iterative usability testing allowed us to identify critical user needs, such as traceability, error reduction, and the ability to undo changes, ensuring the final design met their expectations.
3- Metrics demonstrate the tangible benefits of Qubric, including time savings, improved quality of requirements, and reduced errors.
Future Opportunities
1- Introducing features like saving project-specific manifest files and reusing settings will further optimize workflows.
2- Providing engineers with deeper insights into their requirements, such as error trends and team performance, could add additional value.
3- Expanding Qubric's compatibility with other requirement management tools and enhancing API support could broaden its adoption.
