Mid-Project Presentation Update!

Since the last audit we've settled the path question, sharpened the problem definition, and moved from concept into prototype planning.

The problem we're solving

Most engineering projects start the same way: stakeholders know what outcome they want, but not the technical detail needed to define it. So requirements arrive incomplete, ambiguous, or still evolving. Alignment is assumed through words but only tested once a design exists. By the time the gap surfaces, it's too late.

The hidden cost of this is everywhere once you look for it. Designers loop back to clarify. Clients wait longer for approvals. Requirements that should have been discussed on day one turn up halfway through the project, and a requirement discovered late will always cost more than one discussed early.

Why does Concept AI matter?

The numbers back this up. According to BCG, generative AI can deliver a 40% improvement in creative product design. Autodesk reports that project teams lose 14.1 hours per week to non-productive activities — searching for project information, resolving conflicts, and reworking mistakes. PMI estimates that 5.1% of every project dollar is wasted due to poor requirements management.

Hidden requirements have visible impacts. The flip side: a tool that surfaces them early can improve creativity and productivity and reduce cost at the same time.

Who benefits

The value of Concept AI isn't tied to one role — it sits in the space between several of them.

Systems engineers get clearer requirement discovery and traceability. Designers get a stronger starting point before detailed design begins. Clients get faster concept review and decision-making. End users, downstream of all this, get a more efficient design process and a product that actually matches what was asked for. The common thread is that a shared visual turns private assumptions into public requirements.

The scope — and what we're deliberately not doing

Concept AI supports requirement generation by producing visual concepts. It does not replace the conceptual or detailed design stages, and we've been careful to keep that line clear.

In scope: early-stage concept design, requirement elicitation, requirement interpretation, visual concept generation, and concept review and rating. As stretch goals — only if time allows — rudimentary blueprint generation, customisable output settings, and auditory outputs. Firmly out of scope: detailed dimensional design, cost and schedule estimation, regulatory compliance assurance, and construction planning. Concept AI is concept-only by design.

How it works

The system itself is straightforward in shape. An intelligent agent takes a set of inputs — requirements, constraints, preferences, and context — and produces a visual concept that stakeholders can discuss.

The workflow is a loop, not a line. Stakeholders provide requirements; the agent interprets the set; a visual concept is generated; stakeholders review it; requirements are updated based on what they saw; and the loop runs again. After enough turns, you arrive at a finalised set of requirements that everyone has actually seen rendered, not just read. The generated images aren't the deliverable — they're the conversation starter that gets the hidden requirements out into the open.

Where we are now

The project has moved out of problem definition and proof-of-concept territory and into prototype planning, requirements analysis, and evaluation design.

Completed so far: 

• defining the project scope

• identifying the key stakeholders

• reviewing existing text-to-design and image generation capabilities

• developing a proof of concept for Concept AI

• drafting design requirements and design criteria

• obtaining some of the resources we'll need for the prototype.

Next: 

• finalising the prototype workflow

• testing/reviewing the system

• securing funding for the remaining project components

• completing a feasibility assessment and SWOT analysis.

The scoping and goal-setting work is largely behind us. From here, it's implementation and refinement.

What's next

A formal update will come soon with the passing of Audit 2 as we move on to the next stage of project work.

For more information please look at our repository. To see Mid-Project Presentation Slides please look here.