Two new reports suggest the Pentagon may have finally learned how to acquire technology quickly and inexpensively, at least when it comes to AI-powered capabilities.
One report looks at the Maven Smart System (MSS), part of Maven, a system for rapidly analyzing and sharing information. Palantir currently has the main contract and wants to increase the number of users from hundreds to thousands. However, the Army’s 18th Airborne Corps, which works with up to 70 companies in a DevSecOps environment, played a key role in developing the system through a series of exercises called Scarlet Dragon. Emelia Probasco, a senior fellow at Georgetown University’s Center for Security and Emerging Technology, describes how MSS works in a report published this week.
“It can access sensor data from various sources, apply computer vision algorithms to help soldiers identify and select military targets, and then provide workflow support that allows a request to be approved by the chain of command to attack a target. It can also serve as an archive where battle damage assessments can be stored, as well as provide a map of the locations of friendly forces and targets,” the report said.
Without this software, the Army’s fire support elements – the organizations that coordinate Army fires and joint operations – do not have easy access to sensor and imagery data from commercial and military satellites, despite long-standing requests from the Army.
This was a major organizational achievement, Probasco said, more than a technical one. MSS had senior leaders supporting the program, mature technology, and direct access to developers and testers. And perhaps most importantly, officials took flexible approaches to risk management and funding. But most important was the involvement of leaders who understood three things: the power of artificial intelligence, the needs of the U.S. military, and the bureaucratic intricacies of contracts—particularly the kind of contracts that technology companies need, as opposed to those that the Defense Department prefers.
“Pentagon leadership has praised the importance of having bilingual leaders who are knowledgeable about new technologies and the military to bridge the gap between technical capabilities and operational requirements,” the report said. “Our case study highlighted a third ‘language’ that was critical to the successful development of Scarlet Dragon: contracting and acquisition.”
A second report released last week by CSIS highlights the U.S. Air Force’s Collaborative Combat Aircraft program — essentially an armed drone combat companion for fighter jets — as a possible path for future aircraft procurement. The Pentagon has requested $8.9 billion in funding for the program over the next five years, “a massive increase over what the Department of Defense has spent on CCA’s precursor efforts to date,” Greg Allen CSIS points out in the recently released report, which he co-authored with Isaac Goldston. By making CCA a program so early, the Department of Defense was able to avoid years of delays and potential cancellations.
As part of this effort, the Air Force also fundamentally changed its approach to procurement, abandoning “requirements that were so narrowly worded that innovative proposals from industry were limited and that competition was ended too early in the program life cycle to continue to aggressively push industry forward,” they write.
More importantly, the Air Force took a counterintuitive approach to developing the hardware and software: They were built separately, allowing companies with strong aerospace design skills to participate in building the aircraft, while companies with AI and autonomy technology could develop the software.
“In the past, large procurement programs from hardware-focused defense contractors have frequently been delayed due to software issues,” they write. “Separating hardware and software procurement has several advantages for the government… It allows the national security customer to pursue the most attractive options in both aspects of the program. Were it not so, the government might select a suboptimal software vendor because it finds the hardware offering particularly attractive – or vice versa.”
The CCA program is structured to allow more new entrants to join over time, rather than just a single vendor having a major advantage over the incumbent. The program focuses on “maintaining ongoing competition and opportunities for new entrants. While Anduril and General Atomics were selected to continue with government funding and support, companies that lost this final phase of competition can still compete for the upcoming production contract.”
But the program has not been a runaway success. The Defense Department will only put about 100 of these CCAs into service by 2029, showing that the Pentagon still has a lot of work to do if it wants to increase production of new weapons and reduce costs – the department’s top priority.
“The U.S. lead over China in commercial manufacturing—as in the military—is strongest in the production of small batches of expensive, exquisite systems, such as the high-precision machines that power semiconductor factories around the world,” Allen and Goldston write. “The CCA program and similar programs like the Replicator Initiative demonstrate that the Department of Defense is trying to transform and maintain its competitiveness in a new military technology equilibrium in which mass once again matters.”
