
Conway's Law and the Pentagon's Audit Crisis
Last updated 06:29 AM EST • Read time: 6m
“Organizations design systems that mirror their own communication structure.” – Melvin Conway, also known as Conway’s law.
In its 2023 audit, the Department of Defense was able to account for only $1.9T in assets, leaving half — $1.9T, including equipment and facilities — unaccounted for. Audits are often a topic du jour, and we involved in the defense ecosystem need to grapple with a few truths:
- We’re not talking about the efficacy of investments or understanding who uses what equipment. The most basic purpose of an audit is to count where dollars are going. We should be able to do that.
- $1.9T is a huge sum. It’s more than the GDP of South Korea.
- This isn’t an aberration, it’s the norm. Since the Pentagon was first required to undergo yearly audits in 2018, the Department has never been successful.
There are three primary problems which, until solved, will be blockers for a successful audit of the Pentagon.
- A Special Forces operator's Army rifle, SOCOM scope, and CENTCOM machine gun exist in separate databases. They’re managed by one person (the operator) on pen and paper, indicative of a zero-integration problem. We fight as a joint force, but the systems that track materiel don’t reflect the Pentagon’s organizational structure — making auditing equipment hard and onerous, if possible at all.
- The second problem is that the military connects people through processes but not data. Take jump pay: service members' actual jumps are tracked on paper manifests filed separately from the payment systems that compensate them, making it take weeks — if possible at all — to verify who received jump pay and whether they actually completed the jumps during audits.
- Finally, the Pentagon builds monolithic, branch-specific "software towers" that don’t talk to each other — despite billions of dollars spent trying. The centralization-distribution problem occurs because few functions are appropriately centralized. The Defense Finance and Accounting Service is one such attempt to centralize payment processing, but we still have redundant positions like individual jump pay trackers across every airborne unit. If we want to centralize payment processing, we should actually do it.
The U.S. military is Conway’s law manifest — siloed teams result in siloed software; teams that do not interact, like HR and special operations, do not have software systems that integrate well; buyers, users, and auditors have little technological overlap; attempting interoperability becomes a herculean effort. Given the realities, the Pentagon’s failure to pass an audit is not surprising.
Ground truth for the military is generated by the individual. The warfighter knows what equipment they have and what their compensation should be. The logistics specialist knows inventory on-hand and in transit. Installations command professionals know what infrastructure is available.
To successfully audit the Pentagon, we need to use that ground truth, which is admittedly easier said than done.
Today, no one in the military is responsible for thinking about the holistic experience of the end user. The people that draw up the functional requirements for an HR system don’t talk to the people that draw up the functional requirements for the logistics system who don’t talk to the people that draw up the functional requirements for the hundreds of other systems across the enterprise. They’re in different buildings on different installations. They solve their own problems using niche tools designed for domain experts who have been functionally trained to execute their role, siloed from all others.
Remember Conway’s law: Organizations design systems that mirror their own communication structure. The breakdown in systems that prevent the Pentagon from completing a successful audit is ultimately a symptom of the disease that is the DOD’s org chart.
In lieu of someone responsible for thinking about how all these different systems and responsibilities fit together, the “common interface” becomes word of mouth to and from the end user. If you really want to know how much fuel a command has, you won’t get an accurate answer from siloed, enterprise systems (obviously — or else, the Pentagon would be able to pass an audit!). Instead, we send an E-4 with a clipboard to go inspect the blivets. It’s inefficient, ineffective, slow, and painful for the warfighter and the Pentagon alike.
End users need common digital interfaces to document their world. If we want human beings who are deployed in INDOPACOM to be able to consume information, whether it comes from a logistics system or an HR system or a readiness system or an air threat system or a counter-drone system, DOD must provide the information in a consistent way to that human being to 1) minimize their cognitive load and 2) align enterprise systems into a set of standardized user interfaces for the individual on the ground to accurately capture information.
If that E-4 could digitally track the blivets after inspection, rather than pen and paper, imagine how much easier their lives would be. Quality of life creates an incentive for end users to use the digital tools. A Special Forces operator tracks his gun on his phone; logistics personnel can update on-hand quantities of spare parts for trucks from their tablet; the installation commander memorializes the state of infrastructure from their laptop.
Then, ground truth is digitally captured at the edge.
In parallel, digital tools must have traceable provenance — secure and reliable documentation and attribution anytime any data is changed. Then, deploy a data transformation and analytics layer. This harmonizes the information — a gun is a gun is a gun, regardless of whether it's managed by an airman or soldier or Marine — and enables different personnel in leadership to analyze joint force ground truth data.
When this happens, the Pentagon will functionally break down the barriers between systems traditionally mapped to inherently siloed communications structures.
What happens next? The Pentagon can finally — finally — pass an audit.
James Boyd, founder and CEO of Adyton, is a former Special Forces Communications Sergeant. Prior to founding Adyton, Boyd spent seven years at Palantir, developing their collaborative targeting platform used by SOF in Iraq and Syria, deploying enterprise Palantir mesh across US SOCOM and TSOCs for intel fusion, and designing and implementing other classified intelligence systems. After graduating from Stanford University, Boyd served in the U.S. Army and First Special Forces Group before he joined Palantir upon conclusion of his service.

In Pursuit of Nuclear Abolition
By Peter Huessy • RealClearDefense
Innovation Alone Won’t Prepare the U.S. for the Future Battlefield
By Joshua Segal • RealClearDefense
The sUAS Tactical Trifecta
By Bryan T. Woody • RealClearDefense
The Revolutionary War: When Strategic Options Share the Same Weakness
By Henry Yep • RealClearDefense
The Houthis Are Testing the Limits of Collective Defense
By Saima Afzal • RealClearDefense

Who Shoots Down the Drone Over the Shipyard?
By John Ferrari & Dillon Prochnicki • RealClearDefense
Northrop Grumman Breaks Out the Big Guns at AUSA
By Daniel Terrill, Defense News
Army Pursuing Diverse Array of New Anti-Air Interceptors
By Trevithick, The WarZone
Who Shoots Down the Drone Over the Shipyard?
By John Ferrari & Dillon Prochnicki, RCD
F-35C Lands for the First Time on a Ford-Class Aircraft Carrier
By The Aviationist
