Six concrete shapes. Each is a workflow with named inputs, a compiled middle, and the artifacts that fall out the other side. We name the documents your team already touches, and where the system maintains state behind them.
The patterns above repeat with different documents and identifiers. If you have a workflow where AI demos pass and production stalls — the answer usually lives in the missing state layer. We map it in a focused first conversation.
A focused first conversation ends with a typed scope and a one-page workflow map showing where AI safely fits, what is review-ready, and what infrastructure has to sit underneath first.