Here's Fred Kaplan's take from Slate. IMO, the one really interesting part of the article is his resurrection of the Flynt Leverett proposal:
One good proposal was laid out in a 2004 New York Times op-ed by Flynt Leverett, a former CIA analyst now at the New America Foundation. Leverett described the bricklike wall that has long divided the agency's analytical branch and its clandestine branch, making it nearly impossible for either to share information with the other, much less with competing departments such as the FBI.

To maneuver around this wall, Leverett suggested setting up joint intelligence commands for specific "targets." There might be specific commands to provide intelligence on, say, al-Qaida, nuclear proliferation, Middle Eastern stability, and so forth. The national intelligence director would have the power to draw on personnel and resources from all the intelligence agencies to work together within each of those commands—the heads of which would report directly to him.
Kaplan is much more optimistic than I about the post G-N COCOMs' competence at handling a conflict successfully. He also has a few things just factually wrong as far as I can tell. Still, a matrixed/JTF-like approach to intel might be an interesting alternative for trying to break down the traditional stovepipes.