I described an architecture out loud on a walk last week and read the diagram back in a chat window a few minutes later. Wrong in two places. I said so, and it fixed them.
I used to spend an afternoon on that. Dragging boxes, straightening arrows, nudging a label two pixels left because it sat badly against a line.
None of that afternoon was architecture. It was assembly, and I'd quietly accepted it as part of the job for most of my career.
The interesting bit isn't the time back. It's what a cheap diagram does to how you work. When a version costs an afternoon, you defend the first one. You present it, you get a comment you know is right, and you find yourself explaining why it's fine rather than redrawing it, because redrawing it is your evening. When a version costs four minutes, you throw it away without a thought. You end up arguing with the shape much earlier, when changing it is still free.
I'm not drawing faster. I'm willing to be wrong more often, which is a different thing and worth considerably more.
The diagram was never the work. It was the toll you paid to have the conversation, and the toll just dropped.