[efsavage]

If you go into a brainstorming session with the expectation that you will choose anything, you're right, they are usually going to fail. If you go into them with the sole purpose of exploring a problem space and gathering a diversity of viewpoints, but not evaluating ideas, they can be amazingly valuable. My other two points of advice are to:

1) Keep them short (30-45 minutes). Do them just before lunch so people have a hard stop but also an opportunity to keep the discussion going in an even less structured way. 2) Keep them small (3-5 people). If you've got a dozen people you want to include, break it up, and then if you need/want, do some follow ups where you can group people differently and iterate on some earlier ideas.