Policy Brief

Building the

Tech Coalition
How Project Maven and the
U.S. 18th Airborne Corps

Operationalized Software and

Artificial Intelligence for the

Department of Defense

Author

Emelia S. Probasco

August 2024
Introduction

There are frequent discussions about how the U.S. military should draw from the
country’s commercial innovation base to gain an advantage, especially when it comes
to the application of artificial intelligence (AI). Too often, conversations lament missed
opportunities, valleys of death, painful contracting, or other U.S. Department of
Defense (DOD) processes. While these hurdles are real, there are also positive stories
of instances when commercial tech companies, military leadership, and warfighters
came together to create a meaningful advantage on the battlefield. One of these is the
story of how the 18th Airborne Corps used the Scarlet Dragon Exercise series to
develop the Maven Smart System (MSS), an instance where frontline army users and a
coalition of technology companies—enabled by DOD leadership and policies—pursued
and developed a new technology that is having a meaningful impact on operations.

This paper is about how MSS was developed. We examine leadership actions,
commercial partnerships, and contractual approaches that enabled success, as well as
the organizational and cultural hurdles that had to be overcome. The details of MSS
and its application are not our focus, nor are they appropriate to publish here—this is
still an operational system for the U.S. military. Rather, this examination seeks to
answer the question: What did it take to build MSS for Scarlet Dragon? The goal is to
collect the lessons learned that might enable future DOD innovations with software
and AI.

The example of how the 18th Airborne used Scarlet Dragon as a means of developing
MSS is useful to study for several reasons. As will be discussed in this report, it is a
user/warfighter-driven innovation that bridges intelligence and operations functions to
the benefit of joint fires.* It is also uniquely interesting because of how its development
was managed with flexibility and speed, as well as the participation of numerous
software and AI service providers in a development-security-operations (DevSecOps)
cycle that relied first on commercial service providers. Moreover, the 18th Airborne
case study is interesting because it is not a postmortem analysis: the DevSecOps
development of the Maven Smart System (MSS) though the Scarlet Dragon exercise
series continues today, though its long-term prospects are unclear. The questions that
this prompts for the Pentagon are: How can the 18th Airborne’s successful process for

*
“Joint fires” is the official term encompassing the employment of all types of fires, including artillery,
missiles, and weapons dropped from aircraft.

Center for Security and Emerging Technology | 1

developing MSS be institutionalized, and how can this sort of innovation be
established as a norm for innovation within the DOD?

For all its uniqueness, however, the story of Scarlet Dragon and MSS is also familiar.
Previous studies of Project Maven and other successful quick-reaction units in the U.S.
military highlight similar themes about senior leaders being willing to champion a
program; flexible contracting, funding, and risk management approaches; visionary
front-line leaders; direct access to the operational environment; and the
implementation of mature technology. 1 The nuances of these lessons for this particular
case highlight how the application of software and AI may be different from past rapid
technology adoption efforts.

Background and Methodology

The military has a long history of gathering lessons learned using case studies, and
these include the successes and failures of innovation during the wars in Iraq and
Afghanistan. These studies range from general examinations of organizational
adaptation, to studies of the role of senior leaders in innovation, to detailed looks at the
successes and failures of quick-reaction units, and even specifically at lessons from
army rapid acquisition efforts for command and control systems.2 Prior lessons have
identified success factors that are echoed in the development of MSS for the army’s
Scarlet Dragon exercises, namely: senior military and/or civilian leaders willing to
champion the program; the need for flexibility in contracting, funding, and risk
management; visionary front-line leaders; direct access to the operational environment
with ongoing feedback; and generally relying on the implementation of mature (rather
than developmental) technology.

To gather lessons learned from the 18th Airborne’s experience, our research team
conducted extensive interviews with current and former members of the 18th
Airborne, the Office of the Under Secretary of Defense for Intelligence and Security
(OUSD(I&S)), and various contractors involved with the development of MSS and
Scarlet Dragon between June 2023 and February 2024. We also attended an interim
development exercise held in June 2023 to observe the operators and developers
working together to create the next iteration of the system. Our report includes our
observations from this event and interviews, however, it does not attribute quotations
to protect the privacy of our interviewees. Beyond our in-person meetings, we
gathered information from media reports about Scarlet Dragon and MSS, as well as

Center for Security and Emerging Technology | 2

public contracting announcements, and data from Crunchbase about the companies
that have participated in developing MSS.

What is MSS, What is Scarlet Dragon, and How Did the 18th Airborne Use Them?

While this paper is focused on how the DOD successfully developed and adopted
software and AI for military advantage, it helps to briefly understand what they built in
this case study. Put simply, MSS is an AI-based decision-support system. The MSS
system was refined over time by the 18th Airborne through the Scarlet Dragon
exercise series.* These Scarlet Dragon exercises bring together warfighters—as well as
developers, technicians, testers, and evaluators—to practice the process of finding
potential targets, going through the process of identifying them, locating them, filtering
down to the lawful valid targets, prioritizing them, assigning them to firing units, and
firing against them. This is done in a crawl-walk-run fashion that starts with basic
simulations and ends with actual units firing live rounds against practice targets. These
exercises served to focus the 18th Airborne’s development efforts, and as operational
tests within the DevSecOps process they used for iteratively developing MSS.

The Scarlet Dragon version of MSS can access sensor data from diverse sources, apply
computer vision algorithms to help soldiers identify and choose military targets, and
then provide workflow support that enables a request to be approved by the chain of
command in order to strike a target. It can also serve as a repository where battle
damage assessments can be stored, as well as provide a map of the location of
friendly forces and targets (see image 1).3 With regard to current laws and responsible
AI policies, the system organizes and integrates data, allows users to select and
leverage algorithms to process that data, and gives operators the ability to more
quickly make decisions under accepted army doctrine and decision workflow.†

*
Exercises help forces train as they might operate in the face of potential conflict. They are important for
preparing military forces for combat, but they also allow warfighters to test new technologies and ways
of operating them.

†
The DOD has multiple policies and memoranda guiding the responsible deployment of artificial
intelligence, primary among them is DOD Directive 3000.09, “Autonomy in Weapons Systems,” as well
as the 2021 memorandum from the deputy secretary of defense, “Implementing Responsible Artificial
Intelligence in the Department of Defense.”

Center for Security and Emerging Technology | 3

 While the description of the
system may seem simple, its
impact is significant. Without
this software, army fire support
elements (the organizations
that coordinate army and joint
fires) do not have easy access
to sensor and image data from
commercial and military
satellites. To add further
context, one news report called
this feature of direct access to
satellite data for targeting “a
Holy Grail for the Army.”4 Even
where the data is available,
MSS offers a significant
productivity gain because it
automatically highlights
images for intelligence
analysts to review and then
digitally supports the workflow
from intelligence analysts, to
Image 1. Screenshot of Scarlet Dragon System, decision-makers choosing
Source: U.S. Army Dragon Innovation Newsletter targets, to firing units, and back
to analysts examining battle
damage. In the absence of MSS, the firing process is manual and riddled with
inefficiencies and the potential for errors—from collecting the data to processing it,
seeking permission, matching munitions, and granting permission.

Using MSS, the 18th Airborne has demonstrated an ability to match the performance
of the time-critical targeting cell in Operation Iraqi Freedom, a targeting cell that is
widely viewed as the most efficient in U.S. military history. What is even more
impressive, however, is that the 18th Airborne achieved this milestone with roughly 20
people in its targeting cell, whereas the OIF cell benefitted from more than two
thousand staff members. With these successes, army leaders hope to leverage MSS to
meet a new vision for firing units to make one thousand high-quality decisions––
choosing and dismissing targets—on the battlefield in one hour.

Center for Security and Emerging Technology | 4

In addition to its great efficiency for joint “From computer vision, full-
fires, MSS has two other important features.
motion video, and synthetic
First is the flexibility to integrate data from
sensors and platforms that are not part of aperture radar algorithms
the software’s original design. In October identifying targets, to digital
2021, for example, services and combatant workflow tools improving
commands joined a Scarlet Dragon exercise speed and precision of
and integrated their software and hardware targeting teams, to optimizing
tools into the MSS platform.5 In February
machine-to-machine
2023, the same system integrated U.S.
Space Force operations.6 This flexibility gives communication flow, the
the system greater longevity as both military [Scarlet Dragon] exercise
and civilian sensors proliferate and change. marked a critical step toward
digital warfighting.”
The second notable feature is the ability of
MSS to restrict access to data and shape the – Gen. Michael “Erik” Kurilla,
user interface according to the user’s role or
CENTCOM
clearance level, which allows it to serve as a
central command and control platform for
multi-domain operations in both joint (i.e., across military services) and combined (i.e.,
in collaboration with allies) settings. The feature helped to enable members of the
North Atlantic Treaty Organization to participate in a 2023 Scarlet Dragon exercise.7
With this, U.S. leaders can efficiently share information with allies at the proper
classification level and leverage a common operating picture.

The U.S. Navy, Army, Air Force, and Space Force, have all joined in Scarlet Dragon
exercises to date, which is notable given their differing histories, operations, legacy
technologies, and communications networks. A single platform that can ingest data
from a diversity of sensors, as well as process it, provide it to diverse users, and then
create an interface appropriate to a particular type of user, is technically and
organizationally complex. But this platform is also a longstanding vision of the DOD
aligned with the Combined Joint All-Domain Command and Control (CJADC2) concept.
That this instantiation of CJADC2 was created at the tactical edge with operators is
impressive, though the operators could not have done it without the support of the
Pentagon and a uniquely productive relationship with the private sector.

Center for Security and Emerging Technology | 5

What it Took to Build MSS for Scarlet Dragon

Building MSS and its Scarlet Dragon instantiation was perhaps more of an
organizational feat than a technical one, and one of its most notable features was the
participation of at least 21 private-sector software and AI companies (and potentially
up to 70, according to interviews) in an operational DevSecOps environment.8 Before
exploring the tech coalition aspect of development, it is important to understand the
preconditions for MSS, which included senior leaders championing the program and
clearing a path for visionary frontline leaders; flexible contracting, funding, and risk
management approaches; direct access by developers, designers, testers, and
program/projector managers to the operational environment; and the implementation
of mature technology. While these preconditions will be familiar to those who have
studied past successes for rapid innovation, the nuances in this case are especially
interesting for those studying the application and operationalization of AI for national
security. Within boxes in each section that follows, we document traditional lessons
learned that apply, but which have some nuances for the MSS and Scarlet Dragon
case.

Senior Leaders Willing to Champion the Program and Visionary Frontline Leaders

MSS and Scarlet Dragon would not have come to fruition had it not been for a key set
of individuals who could identify the opportunity, make space for experimentation, and
then take ownership of delivering the solution. Those leaders happen to share three
things in common: 1. significant operational experience; 2. nuanced understanding of
AI, networks, and/or data science; and 3. experience and expertise with government
acquisition and contracting strategies.

Among these leaders is U.S. Army Colonel Joe O’Callaghan, the AI fires officer for the
18th Airborne Corps, and the leader responsible for the development and
operationalization of MSS for the artillery fires, specifically. O’Callaghan has a unique
operational background, having enlisted in the navy before commissioning in the army
as a field artilleryman, as well as having led fire support coordination at the battalion
and the combined joint task force level. He has also attended numerous army
educational courses, to include the Army Space Cadre program. In interviews,
subordinates and colleagues conveyed that he is one of the nation’s best artillerymen,
has a broad knowledge of the capabilities and limitations of AI, and maintains a
detailed understanding of army contracting and ways to partner and move quickly
within the bureaucracy. Finally, his ability to communicate openly across leadership,

Center for Security and Emerging Technology | 6

operators, and technical representatives visibly galvanized the Scarlet Dragon team
during exercises and set a clear direction.

O’Callaghan would not have had the opportunity to participate in the Scarlet Dragon
exercises had it not been for the efforts of U.S. Marine Colonel Drew Cukor in
establishing MSS. Upon arriving at the Pentagon in 2016 as a member of the Office of
the Under Secretary of Defense for Intelligence and Security (OUSD(I&S)), Cukor was
assigned to join the “Third Offset” strategy team, a group put together and empowered
by then U.S. Deputy Secretary of Defense Bob Work. Before the Pentagon, Cukor had
been a marine intelligence officer and served as a foreign area officer, joint operational
planner, and acquisition supervisor.9 He had also spent time at the National Defense
University and had written two draft operating concepts about the connection between
intelligence and operations, as well as other technical papers, including one explaining
the complementarity of human- and computer-based intelligence systems.10

A part of the 3rd offset strategy group, Cukor successfully argued that the deputy
secretary should focus his strategy on the use of AI for warfare advantage. As a
consequence, Work signed a memo establishing the Algorithmic Warfare Cross-
Functional Team (AWCFT) in 2017, and put Cukor in charge of pursuing opportunities
that would “integrate artificial intelligence and machine learning more effectively
across operations to maintain advantages over increasingly capable adversaries and
competitors.”11

While Cukor and O’Callaghan were the essential visionaries and implementers of MSS
and Scarlet Dragon, their efforts benefitted from the top cover and budget assigned by
Work in his AWCFT initiative. Similar to Cukor, Work had been a marine colonel earlier
in his career, and, like O’Callaghan, he had also commanded an artillery battery and a
battalion. He has an extensive academic background in the sciences, as well as
considerable knowledge of DOD budgetary and bureaucratic policies and politics,
having previously served as the under secretary of the navy, and held a position at the
Center for Strategic and Budgetary Assessments.

Center for Security and Emerging Technology | 7

 Lesson Learned: Trilingual Leaders

Pentagon leaders have hailed the importance of “dual-lingual” leaders who

understand emerging technologies and the military to help bridge the gap
between technical opportunities and operational needs. Our case study
highlighted a third “language” that was critical to the successful development
of Scarlet Dragon: contracting and acquisition. O’Callaghan and Cukor
leveraged their knowledge of all three to cultivate workable technical solutions,
rapidly experiment with them, and then implement new technologies within an
army artillery unit’s operation. Cukor’s acquisition strategy was frequently
credited in interviews for enabling rapid experimentation, which he, himself,
credited to his time overseeing acquisition officers at Marine Corps Systems
Command.

With Work’s guidance in 2017, the AWCFT began its task by exploring AI applications
within the military services. The army initially assigned a military intelligence brigade
as the primary point of contact. True to his earlier research, however, Cukor was
interested in the potential role for AI to more closely connect intelligence and
operations. He found a similarly interested leader in O’Callaghan and U.S. Army
General Michael “Erik” Kurilla, of the 18th Airborne, a unit frequently conducting
deployments in rapid response to global events. Kurilla came to champion the idea,
and the notion of rapid experimentation, both at the 18th Airborne and later when he
led U.S. Central Command. Importantly for the success of the program, he held
O’Callaghan and Cukor accountable by assessing progress in mission achievement (the
efficiency of targeting) rather than platform development milestones. This gave the
development team the freedom to iterate and change directions as necessary to
achieve the desired goal.

The incentive provided by the deputy secretary’s memo, the efforts of Kurilla and his
successor, U.S. Army General Chris Donahue, and the relationships built by Cukor and
O’Callaghan gave the project broad support. That support spanned the DOD, including
the Office of the Under Secretary of Defense for Intelligence and Security, which was
running the AWCFT; the National Geospatial-Intelligence Agency (NGA), which now
runs the Maven program; the Army Research Lab (ARL), which managed the main

Center for Security and Emerging Technology | 8

contracts; the Defense Innovation Unit (DIU); the 16th Air Force/Air Forces Cyber; and,
the II Marine Expeditionary Force, which was the first interservice partner to the army.

Lesson Learned: Openness to Change and Partnerships

MSS and the Scarlet Dragon exercises would not have launched were it not for the
openness of the army and AWCFT leadership to change directions and seek new
partnerships. That openness was important at the beginning of the project but
remains important as AI technologies evolve. Iterating on the system as it was
developed to achieve the goal, rather than iterating to achieve a plan, enabled new
applications, and correspondingly, new ways of thinking and operating.

DevSecOps with Direct Access to the Operational Environment

Senior leaders at the Pentagon deserve credit for MSS and setting the conditions to get
Scarlet Dragon started, but the program’s current success is due in large part to the
18th Airborne. The 18th Airborne is a special unit with the army that is often called
“America’s First Responders.” It is designed for, and tasked with, rapid deployments in
response to global events, including, for example, a deployment to Europe within 48
hours of Russia’s invasion of Ukraine.12 In a way, this made the 18th Airborne an ideal
unit to experiment with MSS and apply it to the joint fires problem.

Successful quick-reaction initiatives often credit a close connection between operators

and engineers, but in the case of MSS, several features appear unique to the quick
application of impactful software and AI for Scarlet Dragon. First among these was the
DevSecOps approach and its 90-day cycle of development, testing, feedback, and
production with the 18th Airborne. This cycle allowed O’Callaghan to set goals under
the Maven Broad Agency Announcement (BAA) rather than hard-and-fast
requirements. Additionally, the 90-day development cycle was synchronized with
regular military exercises that would leverage the technology and quickly reveal either
the success or failure of developer efforts in simulated operational environments.
Leaders like O’Callaghan could also use the regular cycle of development and
exercises as an opportunity to evolve unit operations and technical capabilities
concurrently.

Center for Security and Emerging Technology | 9

 “You don't buy AI like you buy ammunition.”
– Col. Drew Cukor13

The DevSecOps approach enabled quick reactions to changes in the operational

environment. For example, soldiers recounted an experience where they noticed a
degradation in algorithm performance due to shifting weather conditions. While the
algorithm still worked, the soldiers knew it could be improved and were able to
consult and rapidly iterate with the developers to improve the algorithm’s performance
across different weather conditions.

Lesson Learned: Rapid Iteration and a DevSecOps Approach

MSS must continually evolve to seize new technological advantages and react
to new operational challenges. That continuous evolution affects both the
military unit and the developers, and requires a structured process like
DevSecOps to stay in synchronization. When this process is working, it allows
for the rapid coevolution of technology and processes to achieve measurable
impact.

The DevSecOps approach also helped to normalize the rapid onboarding, offboarding,
and updating of companies providing algorithms and software solutions. Critically, that
process was managed by the defense contractor ECS Federal, which acted as a
systems integrator by leveraging warfighter and program feedback to help identify and
rapidly onboard the best possible developers. ECS Federal successfully onboarded
(and, at various times, offboarded) a variety of companies leading in AI services and
development. This included large companies such as AWS, IBM, Maxar, Microsoft,
Raytheon, Sierra Nevada Corporation, and at one point Palantir (which was later
elevated to a co-prime with ECS as it became the primary vendor for the user interface
and workflow software), as well as smaller companies such as Black Cape, Inc.,
Clarifai, and CrowdAI.14

To help facilitate the rapid onboarding and offboarding of these new companies, ECS
iterated on a network it had previously built to facilitate information sharing in

Center for Security and Emerging Technology | 10

Afghanistan, Iraq, and Libya, called the Secure Unclassified Network (SUNet).15 The
network enabled the creation of virtual enclaves that allowed interagency and coalition
partners to securely collaborate, as well as to share information, analytics, and
software services, and “acquire, develop, and deploy mission-specific datasets and
analytics that can be shared across multiple enclaves or restricted based on mission
requirements.”16 Alongside companies providing the data and development services,
the MORSE Corporation was hired to implement a continuous test and evaluation
process that would support the DevSecOps cycle.17 These private companies acted as
amplifiers and translators for both the warfighters’ needs and the developers’
capabilities. It was also clear in our encounters during Scarlet Dragon exercises that
O’Callaghan and his key leaders were constantly connecting industry participants with
operators seeking solutions, and educating both sides on how best to collaborate.

Finally, part of the successful connection between the developers and the operators
was the embedding of industry “field engineers” in 18th Airborne exercises, as well as
deployments. During Scarlet Dragon exercises, for example, our team observed several
field service engineers working alongside soldiers and marines, teaching them how to
use the system and simultaneously collecting feedback on useful and frustrating
features. We also had the opportunity to interview Palantir field service engineers who
worked with the 18th Airborne at Fort Liberty, NC, and during several of its
deployments abroad. These engineers shared their belief that their close connection to
the operators allowed them to make more rapid and effective iterations to the system.
Moreover, their direct interactions with the soldiers connected the engineers more
strongly to the 18th Airborne’s mission and people.

Leaders in both government and industry frequently mentioned the connection

between operators and developers as a key enabler of success. The exposure of
developers to operators likely led to the features often highlighted by DOD leaders,
such as an intuitive user interface, a workflow that complemented the 18th Airborne’s
standard processes, and information-sharing features that allowed for a more efficient
exchange of sensitive information with allies and international partners.

Center for Security and Emerging Technology | 11

 Lesson Learned: Embedded Engineering

Translating military needs and workflows to technology developers is

challenging, as is explaining new technological opportunities to operators who
are unfamiliar with new technologies. Embedding developers with operators
in exercises as well as within real-world operations helped to close gaps and
miscommunications, leading to fast and impactful technological development.

Flexible Contracting, Funding, and Risk Management Approaches

Contract reporting and interviews with members of the 18th Airborne, the DIU, and the
Office of the Undersecretary of Defense for Information and Security (OUSD(I&S)),
made clear that Maven and Scarlet Dragon benefited from certain contracting
mechanisms, including an ARL BAA, accounting for 70% of the funding.18 In addition to
the BAA, the team used Other Transaction Authorities (OTAs) and other research and
development (R&D) agreements like Cooperative Research and Development
Agreements (CRADAs), Technology Investment Agreements (TIAs), and Partnership
Intermediary Agreements (PIAs).19 The team of vendors contributing to MSS was
further extended through ECS Federal, which, as described above, could rapidly
establish subcontracts with vendors to address specific goals in a 90-day DevSecOps
spiral. Furthermore, we observed that news reports from Scarlet Dragon exercises
named participating vendors that were not directly connected to MSS; we believe
these vendors likely contributed on behalf of one of the military units participating in a
demonstration or exercise. Altogether, the AWCFT and the 18th Airborne used several
contracting tools to cultivate and maintain a team of vendors that could rapidly iterate
and deliver software solutions for operational needs.

The flexibility of the BAAs, TIAs, PIAs, and CRADAs, together with the subcontracts
managed by ECS, also gave the program the ability to run a competitive DevSecOps
process. As described earlier, under DevSecOps, the program iteratively set
development goals, and when advantageous, the program would designate multiple
vendors for the same task to test different approaches. Interviews confirmed that at
various times this included two UI/UX developers, two cloud service providers, as well
as multiple computer vision algorithm vendors. This rapid and competitive cycle was
hailed by DOD managers, vendors, and operators alike as establishing a

Center for Security and Emerging Technology | 12

simultaneously collaborative and competitive environment that realized fast iterations
and results.

“What we saw from the persistent experimentation angle

is that having multiple events gives (the XVIII Airborne
Corps) the opportunity to, with less stress, take on the small
problems and solve them in an iterative fashion. They can
also just try something out, which helps on what’s
functional and what can work.”

– Maj. Adam Schinder, Scarlet Dragon Observer, U.S. Army

Joint Modernization Command20

The contract management by the AWCFT was so notable that it received kudos from
an unlikely source: the DOD inspector general. In a 2022 report, the IG reported that
“the Algorithmic Warfare Cross-Functional Team successfully monitored and
managed” its contracts and its only criticism was also encouragement to the AWCFT
that it more thoroughly document its processes “because AI and machine learning is an
emerging, complex, and rapidly moving technology that requires close monitoring and
management techniques that are not captured in current procedures and best practices
that are used by the DoD acquisitions community.” 21 The biggest recommendation was
that the program formally document its processes so that others in the acquisition and
sustainment community might benefit from the AWCFT’s lessons learned.

Lesson Learned: Contracting for Change

The common thread among these contracting mechanisms is their flexibility,

which can allow for experimentation. This proved especially consequential in
enabling the AWCFT to change direction away from an army intelligence
application and toward the application of MSS for the 18th Airborne’s Scarlet
Dragon, but it also paid dividends in allowing for new vendors and lightweight
processes as new tasks or opportunities arose.

Center for Security and Emerging Technology | 13

A Motivated Commercial Market Implementing Rapidly Maturing Technologies

Leadership, flexible contracting, and access to the operational environment created

essential conditions for success, but they would have been insufficient without the tech
companies which efficiently translated warfighter needs into software solutions within
the DevSecOps framework. Lessons learned from quick-reaction technology insertion
efforts often focus on the technology itself, but a notable finding from the Scarlet
Dragon case study is the uniquely important role of tech companies as not just
developers but also translators and facilitators for identifying opportunities. The tech
companies were needed to translate operator needs into technical capabilities, and
their task was complicated by rapid changes in the field of AI during the project. How
and why tech companies joined MSS and the Scarlet Dragon effort in particular is
worth understanding given the critical role the companies played.

A brief analysis of the companies mentioned as contractors or subcontractors in the

course of our research reveals that many were relatively new to defense contracting
early in the program. Nearly a quarter of the companies mentioned were founded in
the last decade, and just under half have fewer than five hundred employees. This
stands in contrast to traditional prime defense contractors, which have often been
working with the DOD for between fifty to one hundred years and have tens of
thousands of employees. As indicated earlier, to engage some of these new vendors
on the project, and create a safe space for experimentation, the army had to leverage
SUNet.

Lesson Learned: Continue Lowering the Barriers to Onboarding Growing AI

and Software Vendors

Public network enclaves supported faster onboarding times for new

companies. Streamlined processes also made the business decision to join the
effort more straightforward for vendors who are less accustomed to working
with the DOD.

Joining MSS would not have been an obvious business decision for most companies,
especially startups. It was funded through mostly temporary, experimental
mechanisms, and there was no single point of contact or experienced program office.

Center for Security and Emerging Technology | 14

Businesses likely also understood that while their efforts were funded by the
Pentagon, the Scarlet Dragon version of MSS itself was not a program of record with
guaranteed long-term funding. As of today, a part of the original project under the
AWCFT has transitioned to the NGA as a program of record, but according to
interviews, some of the funding needed to evolve MSS comes from the DOD’s Chief
Digital and AI Office (CDAO).

On the other hand, the 18th Airborne’s operating tempo, its DevSecOps approach,
flexible contracting mechanisms, and even the prospect of avoiding predictable
program office processes seemed to appeal, especially since contractors mentioned
these features as highlights of the program in their interviews. Furthermore, interviews
with company representatives as well as Cukor and O’Callaghan indicated that the
projects were not profitable for the companies in some instances, and in others, the
companies invested significantly to discover and experiment with potential system
improvements without a written promise of payment from the government.

“We are in an AI arms race frankly, and it’s happening in

industry. . . . I see an ecosystem of vendors . . . [where] we
still have large defense industry vendors out there that are
providing the mainstay of the weapon system, and a whole
other ecosystem of very fast software companies . . . who
can bring these algorithms to our platforms.”

– Col. Drew Cukor 22

Any company that joins the MSS development efforts needs to be paid if they are to
survive in a market economy, but interviews with government and company
representatives alike made clear that a sense of mission and an interest in challenging
technical problems also drove company actions. In interviews, employees who helped
to develop MSS clearly conveyed their excitement and pride about working with the
18th Airborne soldiers on Scarlet Dragon exercises, as well as their excitement about
deploying overseas to support the unit on missions abroad. Similarly, employees and
government observers reported that the technical difficulty of the project presented
opportunities for companies to test their capabilities and to learn and improve their
products.

Center for Security and Emerging Technology | 15

While contracts and an interest in the mission and the technical complexity motivated
companies, they also reported drawing on internal resources to work “at-risk” for the
government or when they wanted to experiment with a potentially useful technical
improvements for MSS and the 18th Airborne. That companies made the choice to
work without certainty of payment at times, especially for a program that does not
have guaranteed funding, is a difficult business decision. The need to apply internal
funds is a luxury that not all companies necessarily enjoy and could deter commercial
partners, like startups, who may also opt for more reliable revenue from other
customers.

Lesson Learned: Tech Companies are Willing to be Flexible to Meet the

Mission, but that Flexibility Comes at a Risk that not all Companies can
Afford

The actions of the companies working on MSS demonstrates that they are
willing to sincerely partner and take risks, or even endure temporary losses,
when it comes to urgent missions or especially interesting technical problems.
Presumptions that companies were “only in it for the money” were not only
inaccurate but affected the enthusiasm of tech company employees.
Nevertheless, most of the companies working on MSS were mid-sized and
many had been operating for at least five years. Less well-established
companies may not be in the same financial position as the Scarlet Dragon
contractors to take business risks for the government.

Challenges

Historians have studied the challenges of military innovation since at least the interwar
years and, unfortunately, many of the historical observations still apply in the case of
MSS and the Scarlet Dragon exercises.23 The challenges for MSS and Scarlet Dragon
include institutional conservativism which disincentivizes innovators, organizational
structures that seed parochialism, skepticism about new technologies (and, in this
case, the contractors providing that technology), and, finally and perhaps most
different from historical precedent, revenue risks that suppress commercial business
participation.

Center for Security and Emerging Technology | 16

Institutional Responses to Change and Incentivizing Innovators

Those individuals who led and executed the development of MSS and the Scarlet
Dragon exercises, beyond just the individuals highlighted in this paper, challenged
typical DOD processes and crossed organizational boundaries in ways that created
organizational and interpersonal friction. Interviews indicated that changing directions
in the MSS program from intelligence to fire-support applications put these different
communities in tension. Furthermore, the quick execution of contracts with new
vendors disrupted the deliberative and consultative status quo in the Pentagon in ways
that provoked questions.24 The longstanding challenges of being a person who is a
“disruptor” in the military—an inherently conservative institution—are unsurprising,
though it is worth considering how the Pentagon might support and protect these
disruptors in times of rapid technological change.

Connecting Intelligence and Operations

Another friction apparent in interviews and reminiscent of past observations about

branch or service parochialism was the challenge of integrating intelligence and
operational functions.25 While some of the frictions had to do with the differences in
needs, processes, and deliverables for the two communities, some were also likely
financial. Repurposing the focus of the AWCFT (technology and funding) away from
intelligence analysis and toward the operations of the 18th Airborne’s artillery came at
a cost to the intelligence community’s original plans for the funding.

Skepticism about technology and contractors

Besides the challenges internal to the DOD, MSS contractors identified frictions in
some of their relationships with government personnel. These frictions were
summarized by company leaders as being viewed as “just contractors” or “dirty
contractors,” a theme that has been echoed by companies that voluntarily contributed
to the response in Ukraine.26 This friction may have also coincided with initial
skepticism about the utility of AI for artillery fires, which one soldier bluntly
summarized as: “This s--- don’t work.”27 Those opinions were initial obstacles that
were overcome with experience, and later matured into what could be considered as
justified skepticism when operators purposefully disabled algorithms when they
noticed a degradation in accuracy.

Center for Security and Emerging Technology | 17

Business and Financial risk

Companies acknowledged that while Scarlet Dragon gave them a chance to

demonstrate capability to an important DOD customer, they also took business and
financial risks to participate in the project. Several government representatives also
echoed their perception that some companies performed above the funding level of
their contract. On the lower end of that risk were internal R&D funds dedicated to
developing new technologies or techniques for Scarlet Dragon exercises. In other
instances, we heard of reports of companies working on MSS for 18th Airborne
applications before a contract was finalized. In more extreme cases, we heard of
instances where companies participated in the project at a financial loss, either
because of the intuition that the project was important for national security or because
the technical challenges posed by the project complemented the company’s technical
development aspirations.

Regardless of the reason, the fact that we heard in so many interviews about
companies agreeing to work at risk or at a loss contradicts the notion of the “dirty
contractor” sentiment noted above. Not all contractors are in a financial position where
they could work at risk or at a loss, and no business can function for long without
revenue in a market economy. The inability to fully contract or pay for the technical
development that MSS required may have prevented some companies, particularly
startups, from being in a position to contribute.

Conclusions and Recommendations

MSS and the Scarlet Dragon exercises demonstrate the potential for bottoms-up
innovation to meet top-leader intent. As a partial instantiation of the DOD’s CJADC2
vision, Scarlet Dragon and the 18th Airborne deserve high praise for seizing the
opportunity to apply emerging technology to an important operational challenge. That
praise must recognize, however, the role of strong advocacy from the highest levels of
civilian leadership in the Pentagon, as well as support for flexibility and risk taking by
senior leaders across the DOD and from within the tech sector.

Many of these lessons learned reenforce what the DOD should already know about
leadership, flexibility, as well as the importance of the warfighter’s perspective, and
the value of private sector innovation. But, Scarlet Dragon also offers some unique
insights for operationalizing software and AI:

Center for Security and Emerging Technology | 18

• Senior leaders are vital to clearing administrative and bureaucratic hurdles
for innovators instituting change. Developing and applying new technologies
is already challenging, and institutional resistance to change is an unnecessary
hurdle. Senior leaders are essential to clearing those hurdles, communicating
the vision broadly, and holding innovative teams responsible for measures of
effectiveness rather than measures of performance.

• Invest in training and career pipelines that include operations, acquisition,

and technical skills. There are many advantages to “dual fluency” leaders who
can speak to both technical and operational communities, but trilingual
leaders—those who can communicate across operational, technical, and
acquisition communities—were catalysts for success in paving a new path and
developing MSS for artillery fires. Officer assignment and promotion
considerations should take into account the value of all three skills for the future
of innovative technology development. Leadership should also make use of
personnel policy exemptions when gifted trilingual leaders are making
substantial progress in developing technologies along with operational
concepts.

• Build technical and contractual systems that allow for changes of direction.
Because AI is still evolving, and new applications are emerging, the DOD must
remain open to changing directions and seizing new opportunities. To do so, the
DOD should be ready to establish partnerships across internal DOD boundaries,
as well as through flexible contracting approaches. That openness can be
supported by:
o Agile development processes, like DevSecOps. No static software
system is likely to be successful in modern operations. DevSecOps offers
a way for the Pentagon to continuously adjust to changing circumstances.
Instituting DevSecOps approaches and training operators on how to work
best within a DevSecOps environment can be a long-term strategy for
the DOD. Based on the experience of Scarlet Dragon, the DOD should
also fund operational units to experiment and exercise with new
technical solutions as they are iterated upon in the DevSecOps cycle.

Center for Security and Emerging Technology | 19

o Flexible contracting mechanisms and fast business partners. R&D
funds and BAAs gave the MSS team the flexibility they needed to make
fast contracting changes, but even this was not sufficient. A prime
integrator willing to seek out the best commercial partners and rapidly
onboard them (rather than an integrator that distributed work tasks to
only its own units and staff) proved essential to MSS’ speed and
flexibility. If the DOD is unable to reach the contracting speed required,
as seems likely, program managers should find trusted systems
integrators and hold them accountable for recruiting and rapidly
facilitating external partnerships with a diverse ecosystem.
o Software licensing or app store-like services. The military needs a rapid
and repeatable approach to agile development and deployment of
software and algorithms across the force. Solutions developed for one
unit should be accessible to others if the use cases are fundamentally the
same. This is especially important as conditions on the ground change
and as new solutions become available. Having a standardized library or
store of applications that have been vetted and are available for
widespread use, with appropriate risk mitigations, gives warfighters the
ability to quickly access new software, data sources, and algorithms at
the point of need. It also benefits commercial companies with a more
efficient delivery path and a potential revenue model. The CDAO appears
to be taking the lead in establishing this type of approach, starting with
its new Open Data and Applications Government-owned Interoperable
Repositories (Open DAGIR) multi-vendor ecosystem.28 For MSS, this
ecosystem will also be enabled by the NGA-managed Maven program.
o Field Service Engineers. The continuous evolution of a technical solution
affects both the military unit and the developers and requires a structured
process to stay in synchronization. Embedding developers with operators
in exercises and real-world operations has the potential to support the
co-evolution of technology and ops while also bridging gaps and
miscommunications, leading to faster and more impactful development
cycles.

Center for Security and Emerging Technology | 20

 • Support business contributions and continuity by signposting long-term
commitment to projects, faster contracting, and continuing to lower barriers
to participation. Cutting-edge companies are essential to rapidly harnessing the
latest technologies and, if they have the funds, are willing to take risks to fulfill
a mission or tackle a challenging problem. Prompt and fair payment for services,
as well as reasonable accommodations for those without prior experience
working on government networks––solutions akin to SUNet––can help ease the
transition for new vendors and encourage companies by lowering the financial
or operational barrier to entry. Even more, clear budget commitments that
project stable funding for a program going forward can give companies
confidence and the evidence they need to build a compelling business plan for
investors—industry needs to see the potential for a longer-term commitment
downstream and a longer-term path for [corporate] success. As one subject
succinctly put it: “industry is getting experiment/proof of concept fatigue (aka
innovation theater). Lots of champions have great ideas and are ready to spend
money on demos, but few have a build/measure/learn plan to move forward
with criteria to pivot or persevere.”29

MSS and Scarlet Dragon are a case study in adoption and rapid iteration with software
and AI for the DOD. That Scarlet Dragon achieved an efficiency comparable to the best
the U.S. military has achieved in recent history––with two thousand fewer staff––
should serve as a strong incentive for future responsible experimentation and
development. To seize the potential, the DOD should establish the conditions for
visionary leaders in technology companies and military units to build new solutions by
addressing cultural, procedural, training, and contractual impediments. At the same
time, both the military and the private sector will need to incentivize, develop, and
retain more visionary leaders ready to create the next generation of innovations for
national security.

Center for Security and Emerging Technology | 21

Author
Emelia S. Probasco is a senior fellow at CSET.

Acknowledgements
This paper is the result of an ongoing collaboration and a memorandum of
understanding (MOU) between CSET, the Johns Hopkins University Applied Physics
Laboratory (JHU-APL), and the 18th Airborne Corps. The author wishes to
acknowledge the substantial contributions of fellow project members: Christine Fox,
Alan Brown, and Toni Matheny from JHU-APL, as well as Igor Mikolic-Torreria from
CSET.

This project would not have been possible without the support of the 18th Airborne
Corps, which shared data from the MSS system and gave the project team members
access to training events. Most especially, the project is grateful to Joseph
O’Callaghan, Zach Riley, and Brad Colby Jordan, who generously gave their time to
multiple interviews.

Additionally, the author thanks all the individuals who shared their insights and
experiences working with Project Maven, MSS, and Scarlet Dragon. Including, Akash
Jain, Anna Rubinstein, Bob Hagen, C. Anthony Pfaff, Cam Stanley, Charles Stevens,
Colleen Gaydos, Dan Keller, Kurt Campbell, Josh Wellner, Jared Dunnmon, Shannon
Clark, Drew Cukor, and Jason Brown. Thanks also to Jon Askonas, Jason Brown, Owen
Daniels, Shelton Fitch, Matt Mahoney, and Helen Toner for feedback, editorial review,
and editorial support.

© 2024 by the Center for Security and Emerging Technology. This work is licensed
under a Creative Commons Attribution-Non Commercial 4.0 International License.

To view a copy of this license, visit https://creativecommons.org/licenses/by-nc/4.0/.

Document Identifier: doi: 10.51593/20240014

Center for Security and Emerging Technology | 22

Endnotes
1
Chad Heitzenrater, Bradley Wilson, Sydne J. Newberry, Sarah W. Denton, James Ryseff, Jeff
Broughton, “Lessons Learned from the Algorithmic Warfare Cross-Functional Team—Project Maven,”
RAND Corporation, 2023.

2
David Barno and Nora Bensahel, Adaptation Under Fire: How Militaries Change in Wartime, New York:
Oxford University Press, 2020; Liam S. Collins, “Military Innovation in War: The Criticality of the Senior
Military Leader,” Princeton University, 2014, http://arks.princeton.edu/ark:/88435/dsp01br86b3727;
Jonathan Askonas, “Adapt or Die: Lessons Learned for U.S. Army Quick Reaction Capability
Organizations,” Author’s draft. U.S. Army Asymmetric Warfare Group History & Lessons Learned
Project; and Shara Williams, Jeffrey A. Drezner, Megan McKernan, Douglas Shontz, and Jerry M.
Sollinger, “Rapid Acquisition of Army Command and Control Systems,” RAND Arroyo Center, 2014,
https://www.rand.org/pubs/research_reports/RR274.html.

3
Osvaldo Fuentes, “Data-centric exercise showcases joint capabilities, lethality,” U.S. Army, February 6,
2023,
https://www.army.mil/article/263764/data_centric_exercise_showcases_joint_capabilities_lethality.

4
Theresa Hitchens, “Army, NRO Pioneer Direct Sat Imagery Downlink In ‘Scarlet Dragon,’” Breaking
Defense, October 11, 2021, https://breakingdefense.com/2021/10/army-nro-pioneer-direct-sat-
imagery-downlink-in-scarlet-dragon/.

5
Jane Edwards, “Army Conducts AI-Enabled Target Identification Exercise Under Scarlet Dragon
Program; Lt. Gen. Erik Kurilla Quoted,” ExecutiveGov, October 6, 2021.
https://executivegov.com/2021/10/army-conducts-ai-enabled-target-identification-exercise-under-
scarlet-dragon/.

6
Osvaldo Fuentes, “Data-centric exercise showcases joint capabilities, lethality,” U.S. Army, February 6,
2023,
https://www.army.mil/article/263764/data_centric_exercise_showcases_joint_capabilities_lethality.

7
Jonathan Jay Koester, “Modernization leaders use Scarlet Dragon exercise to continue Project
Convergence campaign of continuous learning,” U.S. Army, Feb 13, 2023,
https://www.army.mil/article/263951/modernization_leaders_use_scarlet_dragon_exercise_to_continue
_project_convergence_campaign_of_continuous_learning.

8
Reviewing the data from 2018-2023 available to the authors and validated via usaspending.gov and
multiple interviews, we confirm the participation of 21 companies. However, some interviews indicated
that as many as 70 companies had a role in creating MSS.

9
“Colonel Drew Cukor, USMC,” GovExec, https://cdn.govexec.com/media/kmc/cukor_bio1.pdf.

10
See Drew Cukor, “Operate to Know,” National Defense University thesis 2014; and John M. Graham,
Kathleen M. Carley, and Drew Cukor, “Intelligence Database Creation and Analysis: Network-Based

Center for Security and Emerging Technology | 23

Text Analysis versus Human Cognition,” Proceedings of the 41st Annual Hawaii International
Conference on System Sciences (HICSS 2008), Waikoloa, HI, USA, 2008, pp. 76-76, doi:
10.1109/HICSS.2008.213.

Robert Work, “Establishment of the Algorithmic Warfare Cross Functional Team (Project Maven),”
11

April 26, 2017.

12
Rachel Riley, “18th Airborne Corps’ Unit Returns to Bragg After Eight-Month Deployment,” The
Fayetteville Observer, October 19, 2022,
https://www.fayobserver.com/story/news/military/2022/10/19/18th-airborne-corps-units-return-to-fort-
bragg-from-europe-deployment/69569777007/.

13
Cheryl Pellerin, “Project Maven to Deploy Computer Algorithms to War Zone by Year’s End,” U.S.
Department of Defense, DOD News, July 21, 2017, https://www.defense.gov/News/News-
Stories/Article/Article/1254719/project-maven-to-deploy-computer-algorithms-to-war-zone-by-years-
end/.

14
Palantir began as a subcontractor to ECS Federal and later won a direct contract for the project. The
companies listed are reflected in contract W911QX21C0022 on USASpending.gov.

15
See contracts W911QX17D0012 and W911QX17D0010 on USASpending.gov

16
Brian Stanton, et al., “US Army Artificial Intelligence Innovation Institute (A2I2) Aiding Multi-Domain
Operations (MDO).” ARL-TR-8992 DEVCOM Army Research Laboratory, July 2020,
https://apps.dtic.mil/sti/citations/AD1104260.

17
Morse Corps., “JAIC Selects MORSE for AI Test and Evaluation Effort With $250 Million Ceiling,”
February 10, 2022, https://www.morsecorp.com/pressrelease.html?prId=pr_02_10_2022.

Government Accountability Office, “Artificial Intelligence: Status of Developing and Acquiring

18

Capabilities for Weapon Systems,” February 2022, https://www.gao.gov/products/gao-22-104765.

19
For more information on these types of agreements, see: “R&D Agreements | Adaptive Acquisition
Framework.” Accessed July 29, 2024, https://aaf.dau.edu/aaf/contracting-cone/rd-agreements/.

20
Jonathan Jay Koester, “JMC answers call for Persistent Experimentation,” U.S. Army,
https://www.army.mil/article/269729/jmc_answers_call_for_persistent_experimentation.

21
U.S. Department of Defense Inspector General, Report No. DODIG-2022-049 “(U) Evaluation of
Contract Monitoring and Management for Project Maven,” January 6, 2022.

22
Jon Harper, “Pentagon Struggling To Take Advantage Of Artificial Intelligence,” National Defense, Vol.
102, No. 766 (September 2017), pp. 24-25.

23
David E. Johnson, Fast Tanks and Heavy Bombers: Innovation in the U.S. Army, 1917–1945. Cornell
University Press, 1998.
Center for Security and Emerging Technology | 24
24
The program attracted so much attention, in fact, that it is the subject of two inspector general reports.
Both largely confirm the value of the AWCFT’s approach to development MSS. See U.S. Department of
Defense Inspector General, Report No. DODIG-2022-049 “(U) Evaluation of Contract Monitoring and
Management for Project Maven,” January 6, 2022. And, U.S. Department of Defense Inspector General,
Report No. DODIG-2023-044 “(U) Evaluation of Cybersecurity Controls on the DoD’s Secure
Unclassified Network,” January 12, 2023.

25
David E. Johnson, Fast Tanks and Heavy Bombers: Innovation in the U.S. Army, 1917–1945, Cornell
University Press, 1998.

26
See Christine H. Fox and Emelia S. Probasco, “Volunteer Force,” Center for Security and Emerging
Technology, May 2023, https://cset.georgetown.edu/publication/volunteer-force/.

27
Katrina Manson, “AI Warfare is Already Here,” Bloomberg Businessweek, February 28, 2024.

28
U.S. Department of Defense, “CDAO Announces New Approach to Scaling Data, Analytics and AI
Capabilities,” May 30, 2024,
https://www.defense.gov/News/Releases/Release/Article/3791829/cdao-announces-new-approach-to-
scaling-data-analytics-and-ai-capabilities/.

29
Jason Brown in an email message to author, June 30, 2024.

Center for Security and Emerging Technology | 25