Pre-Project Audit 3 Update
Project Audit 3 Progress against Concept of Operations
Our Concept of Operations had 4 main goals. Of these four, three have been deemed partially completed, and one has been judged fully complete. The specific goals and their current progress are as follows:
ConOps §2.2 Goal | Status | Evidence & Outstanding Work |
1. Investigate capabilities of current text-to-design and AI image generation technologies. | Partial | Completed: Research relating to AI Assisted Design, Concept AI Workflow Research, Requirement and Evaluation Metrics Research, plus paper PDFs and key notes. Outstanding: A comparative analysis of the candidate models, and subsequent selection. |
2. Evaluate the financial and operational feasibility of generating visual concepts in real time to elicit stakeholder requirements. | Partial | Completed: Validated project feasibility and established funding of $150, with our host purchasing OpenRouter credit. Outstanding: There remains no financial feasibility section exists, and the "real time" element is unresolved as the team decided on 04/08 that latency is "not an important metric but record for viewing", which sits in tension with a goal written around real-time generation. |
3. Design an architecture for the Concept AI system capable of visual concept design. | Achieved | Completed: Functional Breakdown decomposes the system into eight functions across input, processing, generation, display and iteration. Supported by the QFD, Requirement Analysis and the Requirements & Design Criteria. A working hosted prototype implements the architecture. Outstanding: The System Architecture document still lacks detail and written context, and which will be required for hand-over. |
4. Demonstrate how the proposed system would ensure stakeholder requirements in the early stages of the design process. | Partial | Completed: Demonstrated qualitatively through the live demonstration at the Mid-Project Presentation, POC endorsed by the project host on 13/05, and two external SME sessions — the McKinsey session produced specific design changes on requirement gating. Outstanding: A quantitative test and analysis is required to support these claims, so they are measured rather than anecdotal. |
Project Audit 3 Progress against Targets
We have met our crash targets for project completion that we set for Project Audit 3. We have made significant progress towards both our Goal and Stretch targets. We have put off the SWOT analysis part of the Goal target due to higher priority project work needing to be done now, as well as awaiting the chance to get further clarification on what our project host wants out of the analysis. For our stretch goals all that remains is to add user aid resources to help with using the system, as well as to complete the formal testing process.
The table below has more detail on all three target levels, as well as their current progress:
Level | Target as written | Status | Assessment |
Crash | Partial system design with key components defined and basic workflows documented. | Met | The system decomposed into eight functions and the workflows are documented and traced in the Prototype Testing evidence. |
Goal | Detailed designs for system elements and beginning of SWOT analysis. | Substantial | Detailed design is in place. However, the SWOT analysis was not completed, as recently the group decided to deliver the SWOT analysis after PA4. |
Stretch | Refined architecture, improved user interaction, style controls or enhanced prompt processing. | Substantial | Completed credential self-request, live requirements database, requirement editing, and improved system reliability. However, user guides are pending such as a "how to" video. |
Stretch (Unwritten In ConOps) | Testing results evaluating output quality, reliability and performance against design requirements. | Uncompleted | No test runs recorded. |
System Technical Progress
The system is currently between the Prototype and Minimum Viable Product (MVP) phases. Currently the system is locally hosted and uses self-service credential requests to access a variety of APIs, which the user can choose between. The system takes requirements as inputs, which the system categorises and edits, and generates a pair of images. These generated images are the prompts for discussion with stakeholders for extracting deeper level requirements than those initially presented.
To fully reach the MVP stage, the system will need higher quality and more consistent output images, expanding the accepted inputs beyond just a requirements list, and having greater user aids such as guides or video demonstrations.
Feedback Response from PA2
The feedback from Project Audit 2 was put into a large spreadsheet in our repository. There is too much to cover here as there were over 90 different suggestions. A majority of these focused on tidying up systems documents and improving consistency and readability across our repository, or adding more traceability and interconnectedness. These have all either been decided against, or added as in progress or completed, depending on whether the team judged them as suitable changes. All the feedback suggestions also got a brief description of either why they were rejected or what steps were being taken to address them.


Comments