Reform Bought Speed. It Did Not Buy Understanding.

Pete Newell has earned a place in this newsletter before. His work with Hacking for Defense, alongside Steve Blank, is close to where the argument about staying out of the building originated for this series. His latest piece, cross-posted on Uncommon Leadership after running in The Cipher Brief, deserves the same attention, because it makes the same argument this series has been making, just from inside the Pentagon’s newest acquisition reform.

The setup is straightforward. The Department of War executed what Newell calls the most ambitious acquisition reform in six decades: scrapping the JCIDS requirements process, replacing program offices with portfolio executives, and building a new Warfighting Acquisition System designed for speed. On paper, this is the fix everyone has been asking for. In practice, Newell argues it repeats a mistake made during the post-9/11 wars: chasing evolving threats with rapid fixes while no one takes responsibility for understanding them.

His evidence is the counter-drone fight. Cheap interceptors that have proven themselves in Ukraine still have no fast path into U.S. formations, so soldiers are engaging low-cost drones with missiles worth hundreds of thousands of dollars, and a firmware update that defeats a jammer can be built in hours while the American countermeasure takes months. Newell traces the pattern back to his own time running the Army’s Rapid Equipping Force during the counter-IED fight, a resemblance he has drawn out directly elsewhere, and the structural parallel is exact.

The diagnosis is where this gets interesting for readers of this newsletter. Newell frames successful innovation as a six-phase cycle: detect, define, develop, deploy, assess, distribute. The reforms invested almost entirely in the middle two phases, develop and deploy, while three of the six have no organizational owner at all: nobody persistently monitors how the threat evolves at the tactical edge, nobody measures whether fielded systems actually work against an adaptive adversary, and nobody moves what one unit learns to every other unit facing the same problem. The department, in his words, built a faster engine and left the steering unbuilt. It is the same principle behind understanding the problem before proposing the solution, just applied at the level of an entire acquisition system rather than a single contract.

That is the same distinction this newsletter has drawn between AI layered onto broken infrastructure and AI supported by real architecture, just applied to acquisition policy instead of enterprise software. Speed layered on top of an unreformed structure does not transform the structure. It accelerates whatever the structure already produces, gaps included. A contracting process that moves faster without anyone owning detection, assessment, or diffusion is not a faster version of the old problem. It is the old problem arriving on a shorter clock.

It is also, underneath the acquisition language, a data trust problem. Newell’s “detect” phase is ground truth from the tactical edge, the same discovery gap this series has described from the commercial and government dashboard side, where a solution shaped without input from the people who execute it substitutes opinion for fact. His “assess” phase is the question of whether a fielded solution is actually trusted and used, not just delivered, which is the harder and slower work of closing the gap between what the data says and what people on the ground believe. And operators who route around a system that does not answer their real questions are doing exactly what soldiers did for years when they demanded a commercial tool the Army refused to adopt: quietly building or requesting the tools they actually needed, because nothing above them owned the question of whether the official system worked.

His “distribute” phase belongs in this same bucket, and it is a network problem before it is a policy problem. Moving what one unit learns to every other unit facing the same threat requires knowing which units are actually connected to which, and which relationships in that network would carry a lesson fastest if strengthened. That is the same inversion behind the West Point network mathematics this team has written about, where the logic that identifies the most damaging cut in an adversarial network is the same logic that identifies the most valuable connection to reinforce in a friendly one. Newell is describing an organization that has not yet mapped its own network well enough to know where to route what it learns.

None of this is an argument against the reforms Newell is describing. Killing a requirements process that ossified for a generation is real progress, and he says as much. The argument is narrower and more useful: procedural speed and architectural understanding are different problems, and solving the first does not solve the second. Commercial organizations chasing AI transformation on top of Tier 1 data infrastructure are making the identical mistake, just with a lower cost of failure than a formation in the field.

If you are looking at a reform, a platform rollout, or an AI initiative in your own organization and wondering why the speed isn’t translating into the outcomes it promised, this is usually where to look first. Storm King Analytics would be glad to help you find where the ownership gap sits. Reach out at info@stormkinganalytics.com.