I've been asking about several specific aspects of game design, and several of the suggestions have been of the form, "You need to figure out what the details of the game are going to be before you do the specifics." This may be a stupid question, but how do I get from a list of vague ideas and semi-working demos to a usable design? I've heard good things about "iterative design" in the sense of making up some game mechanics and seeing if something fun comes of them, then building a game around them. I've also got several general ideas of stories, since writing is my specialty. There are also one or two specific game mechanics I'd like to include, and I'd like to do something with AI that hasn't been done already. But none of that is a game idea! What do I need to do to get organized?

Well.... the only thing I use to get organized is a pad of paper and pencil. Write down all your ideas because things that sound good now, might sound really dumb tomorrow, and when you go back through (read it over) sometimes you think of things that would be really neat, or a cool way to change/addon to that idea. I also really like flow charts, so you can organize, say, the story of a game. So you put where you start in a bubble, then an arrow to where they go next or the next event. If there are multiple paths, it makes it very easy to see where a story may separate and then come together.

With trying to do something with AI that's really new and innovative, you might want to finish the game first, then go back and rework the AI. Just make it so the first time through, you design that part thinking of how it'll work, later, so it's easy to change. Making a game isn't too easy, so I just find it easier/simpler to make something that works, THEN go back and make all the complexities of AI, crazy magic stuff, and any of those kinds of things. Much easier to do if you can playtest it right away (because you already have something that runs/works/can play).

Choosing a Goal and a Topic was really helpful in guiding my first preproduction:
http://www.vancouver.wsu.edu/fac/peabody/game-book/Chapter5.html

What I found helpful also was to vaguely define the Unique Selling Points of the game before full blown design. In your case, it's Innovative AI and Game Mechanics.

What I typically see when people try to jump start a creative project is they throw out a bunch of creative details without any main theme to tie it together. The more it fleshes out, the more tangent the creative ideas. In the end, it's just a bag of tricks. Most captivating pieces of work have a central theme and every detail supports that theme. It's cohesive. Any detail that detracts from your theme should be abandoned. This is true whether you are writing a term paper or designing a web page. And so it probably should be for a game.

You need a coder. If you don't have anybody technical on board, then you can't create a design that is of much use to anybody. It's not 'usable' unless someone can say, "yes, I can implement that" for you. And that requires that you have such a person to talk through your overall doc with you and flesh out what each bit actually requires. I'm surprised at how often designers miss the really important stuff in their designs in favour of high concepts that don't really say very much.

Of course, your original post was a bit like that too - high on abstraction and low on specifics. ;) Being more specific will take you a long way in creating useful designs.