Every architecture workshop ends with the same uncomfortable gap. The room was alive. The sticky notes made sense. The model felt right. And then everyone goes back to their desks — and weeks pass before anyone finds out whether the shared understanding actually held.
LLMs change that moment. Not by replacing the workshop, but by making its output immediately testable.
This talk tells the story of a deceptively simple shift: feeding workshop artifacts — domain stories, visual glossaries, context maps, event catalogs — directly into an LLM to generate a working prototype on the spot. Then, doing what a domain expert would do: clicking through it, questioning it, and using what surfaces to sharpen the model before the room disperses.
The distance between modeling and feedback, which used to be measured in weeks, can now be closed in minutes. That changes what a workshop can be — not just a discovery activity, but the opening move in a rapid validation loop.
But this talk goes beyond the demo. It addresses the uncomfortable truth that not all workshop output is LLM-ready. Vague language produces vague prototypes. Implicit assumptions stay implicit. You get back a mirror — and it shows you exactly how precise your glossary was, how complete your context was, how much your team was actually aligned.
That is the real value. The prototype does not just validate the workshop. It continues it.

Attendees will leave with a concrete understanding of:

– Which workshop artifacts translate most reliably into LLM input — and which structural gaps produce the most revealing failures
– What to capture explicitly during a session to make the generation useful rather than generic
– How to introduce immediate LLM feedback into workshops without disrupting facilitation or requiring participants to touch a keyboard
– A new question worth asking at the close of every modeling session: Shall we see what we modeled?

No prior experience with LLMs or prompt engineering is required. The talk includes a live demonstration with real workshop artifacts and an honest reflection on where the approach breaks down.

Why This Talk, Why Now
Architecture workshops have not fundamentally changed in twenty years. We still rely on the same human interpretation chain: model ? artifact ? handoff ? implementation ? feedback — each step adding delay and losing fidelity.
The question is not whether AI will change how we discover and validate architecture. It already does. The question is how software architects can capture that change inside the workshop itself, before the shared understanding fades.
This talk answers that question with working examples and practical guidance.