Every design doc I write now starts at the board. Not with an outline, not with a blank document, not with a template — with a marker and whatever space is free on the wall.
The first thing I do is try to draw the problem. Not the solution, the problem. What are the boxes? What are the flows? Where are the gaps? I've found that I can't articulate a problem clearly in prose until I've drawn it first, because drawing forces a specificity that prose lets you avoid.
Once the problem is clear, the solution space becomes more obvious. I draw candidate approaches on the same board, sometimes in different colors. I look for where they diverge. The divergence points are where the interesting decisions are, and those decisions are what the design doc actually needs to address.
When I sit down to write the doc after a board session, I'm not staring at a blank page. I'm transcribing something I already understand. The hard thinking is done. The doc is the communication artifact, not the thinking space.
This sounds like it adds a step, but it removes more than it adds. Design docs written this way are shorter, clearer, and get through review faster. They're shorter because I've already discarded the wrong approaches before writing a word. They're clearer because I've already worked out what the doc actually needs to explain. They get through review faster because reviewers can tell when the author understood the problem before they started writing.
The board session is usually an hour. The doc writes itself in another hour. That's faster than the alternative, which is spending four hours writing a doc that reveals its own problems in review.