At a roundtable with industry and academia at the Defense Innovation Unit, I met a CEO whose company had developed a language translation app that was providing critical support across the Department of Defense. He shared that the department was paying high prices at the last minute across multiple siloed vehicles and customers. It was, he frankly said, a poor business model. He couldn’t plan for the support the department clearly needed, and the department couldn’t leverage enterprise scale and predictability to drive down the cost. It was bad for the taxpayer and bad for the company to be on this unpredictable seesaw of demand. In parallel, in August 2023, the department highlighted a critical shortage of translators impacting some key missions. So, in November, my office onboarded this company across our platforms on an enterprise vehicle with enough ceiling to support demand across the department. It is critical that this type of speed of execution matching AI and digital solutions with real mission impact moves from episodic to routine.
AI and digital solutions are increasing the pace of change and advancement across key sectors of American society, including the Department of Defense. Advanced data applications that leverage AI are a critical part of defense modernization efforts, whether we are looking at tools that support warfighting, data management, or anything in between. In physical platforms like fighter jets, ground vehicles, or autonomous vehicles, and in functions from logistics to health care to financial management, these digital solutions are not merely layered on to military hardware — they are part and parcel of a weapon system or enterprise tool. Foundational to justified confidence in AI-enabled systems is ensuring they are equitable, traceable, secure, reliable, and governable, and can be responsibly used by both operators and leaders. The United States has world-class talent, innovating in an open society and with access to the free market of cutting-edge solutions. As part of U.S. defense modernization, the Department of Defense should continue to leverage these national assets and ensure a lead not only measured in speed and scale, but in responsibility.
As the head of the Chief Digital and Artificial Intelligence Office, I have been focused on driving advances not only in these critical areas but in changing how our office procures software to better comport with the principles of interoperability and replaceability. This will allow best-in-breed technologies to enable our warfighters and business systems. As AI rapidly matures, this office should be structured with the right investments to adapt to shifts in technology, and the underlying infrastructure and offices to enable, speed, and scale growth and adoption. While I will be departing government in January, capitalizing on the advances in AI and technology on behalf of the warfighter and the taxpayer is a clear imperative. I hope our Open DAGIR has set the foundation for successful procurement of advanced technologies for years to come.
Hardware Solutions in a Software-Driven Environment
The Department of Defense has not yet fully realized the promise of AI capabilities. Technical solutions have not been fielded at scale due to three primary challenges: the hardware-centric procurement processes favored within the Department of Defense, proprietary network-centric approaches to procurement that stifle future interoperability, and a bias towards risk aversion that leads to inaction and, conversely, may be creating bigger risks. The department should down-risk and say “yes” when it counts, starting with these three challenges.
Software Procurement in a Hardware-Centric Department
First, the Department of Defense continues to orient around hardware-centric procurement processes that support its major platforms and weapon systems. These processes are critical tools for fielding platforms and systems that may be used for decades to come and stand the test of time. However, these same processes are often not well-suited to the rapid fielding of more software-centric capabilities. The department rightfully procures hardware based on extensive (and thus long-lead-time) planning, and those acquisitions and subsequent sustainment programs need to essentially build towards a specific technical capability that is relatively static. In contrast, software-defined functionality is often rapidly evolving and developed iteratively. This means major program acquisitions, that focus on a linear, longer-run capability-development approach, create impediments to using the agile frameworks more common in commercial emerging technology settings. Moreover, because the software in major program acquisition is embedded in proprietary hardware solutions, changes and upgrades critical to enhanced functionality and performance can be difficult to incorporate without major changes to the underlying hardware.
The United States needs a different way to buy these much more rapidly evolving, but less enduring capabilities. In line with that need, in 2020, the Department of Defense aligned the range of processes and approaches into a more tractable acquisition framework, the Adaptive Acquisition Framework, with six acquisition pathways, including one for software. The software pathway emphasized modern software development practices, but applying these at scale, especially with legacy programs, remains a challenge.
Moving From Network-Centric to Data-Centric
As part of the traditional design process, each platform, sensor, or weapon system had its own proprietary network which leveraged specific data that adhered to custom-defined sets of standards. For example, the Aegis Weapons System has Cooperative Engagement Capability, the F-35 has Multifunction Advanced Data Link, and the F-22 has Intra-Flight Data Link. Even systems that communicate over Link 16, which adheres to “Military Standard” 6016, have more than 50 message formats for different data types used in surveillance, targeting, weapons management, electronic warfare, and more. This approach to data sharing would be like AT&T, Verizon, and T-Mobile deciding to each provide a specialized device and cell network with unique data formats depending on whether you want to make a phone call, send a text message, or browse the web. You can’t have it all, can’t easily switch, and would have to choose based on which kind of medium you (and your friends!) may use most.
This kind of approach may be necessary for the last tactical mile, where fire-control quality and safety are paramount. But for every other function, especially command and control, it creates a web of options that are not optimized for joint (involving multiple U.S. armed services) or combined (involving allies and partners) warfare. It makes the integration of new data sources and interoperability with other platforms — especially across military services — a complicated and expensive effort. The solution to this is for the government to establish an architecture that is open, modular, and data-centric. This would require systems to be able to make use of data from any source over any network. Instead of relying on proprietary “ontologies,” which require the data be in specific standards or formats, systems should be able to send and receive the right data fields for the mission. If built into requirements and procurements on new platforms and incorporated into legacy systems, this would shift the burden from the government having to define the integrations to the vendor and their technology within any network (proprietary or not) having to be agnostic to the data regardless of origin or type.
A Unidirectional View of Risk
The adoption and scaling of AI-enabled solutions are hobbled by a pervasive culture of risk aversion in the defense enterprise. Risk-minimization culture is in part necessary because the government cannot sustain a level of risk akin to the private sector, given the number of no-fail missions and responsibility for stewardship of taxpayer dollars. Additionally, the Department of Defense faces a wide range of risks that need to be considered and mitigated to ensure critical capabilities are securely and reliably available during all contingencies. At the same time, this risk aversion is premised on the notion that integration and deployment of new technology increases risk. That unidirectional view of risk, perversely, may increase overall risk through tech debt and lack of key new capabilities. To be explicit, there are risks from not adopting technology that are often ignored. There are a number of efforts to address this generally, and specifically as it relates to “authority to operate,” including reciprocity and continuous authority to operate processes. Authority to operate indicates a technology meets certain criteria to be allowed on and/or exchange data with a Department of Defense system and can take months or even years to achieve. Efforts surrounding reciprocity and continuity aim to ensure that technologies that have been determined to meet these criteria by one departmental organization do not have to be re-certified to be used in another area of the department. While these have helped create policy support for a more balanced risk management approach, the streamlined approaches to authorization ought to be “productized” into repeatable processes to enable scale.
Given the complexity of the problem and the scale required for true digital transformation, there is no silver bullet or single solution. Addressing and overcoming these three primary challenges and getting to a legal, ethical, and meaningful “yes” on all three will accelerate this transformation and deliver true next-generation AI capabilities to both the boardroom and the battlefield. Much has been written about activities that will need to occur to address these issues and, to that end, the Chief Digital and Artificial Intelligence Office has introduced our framework for Open DAGIR to drive progress.
Open DAGIR
In May 2024, the Chief Digital and Artificial Intelligence Office launched an initiative focused on improving how the Department of Defense ensures reliable access to new technologies that form the department’s digital foundation. Open DAGIR is not a single procurement or program — rather it is a combination of principles and processes necessary to acquire, build, and manage modern data and AI technology at scale. The foundation has 5 principles: clear and accessible acquisitions pathways; government-owned technology infrastructure and services; clear requirements, policies, and processes to balance the department’s cyber security needs with the imperative to develop and deploy prototypes; rules and tools pertaining to standardized data, analytics, and an AI marketplace; and a transparent experiment-based operating model.
The Open DAGIR framework is grounded in the principles of interoperability and replaceability — the Department of Defense needs different technologies to seamlessly interact with one another and with its data, agnostic to the underlying infrastructure. This allows the Department of Defense to take advantage of its federated data ecosystem and diverse set of use cases by utilizing the right data architecture (e.g., relational or object-oriented) while ensuring that use cases can draw from multiple stacks (e.g., logistics use cases can leverage both relational and object-oriented data). Second, it separates procurement of the data from the development of applications, where appropriate, to ensure that applications can seamlessly leverage a variety of infrastructure solutions and data infrastructure can be leveraged by a more diverse set of applications. The first two aspects are critical to resist a natural tendency for technology to fracture into silos by allowing applications and data infrastructure to become more interchangeable.
Marketplace Environment
Open DAGIR breaks apart the vertically integrated tech stack that the Department of Defense has an unfortunate habit of buying in its entirety, at great cost and lack of agility. Imagine you bought a phone with a set of applications, and then every time you wanted a new application, you bought an entirely new phone. It would be exorbitantly expensive, inefficient, and frustrating. Instead of buying the entirely new phone (or tech stack) you should be able to buy individual applications, retain individual applications, and remove individual applications, all with the same hardware and underlying infrastructure. When you want a new phone, you should be able to port these applications over — this is “interoperability.” To enable this, we have developed rules and tools to promote standardized digital tools and products in an AI marketplace environment. Think of it as being able to have the same app on multiple types of phones and being able to move all your apps over to a new phone and seamlessly access your distributed data (e.g., contacts and emails). This is the type of flexibility required to move at speed and scale for digital solutions to warfighter problems.
Acquisition Pathways
However, this flexibility only becomes a strategic advantage when paired with the acquisition pathways to deliver at enterprise scale. The next principle of Open DAGIR is clear and accessible acquisition pathways to allow both existing and new vendors to build prototypes and deliver services with pathways to production and scale. Open DAGIR offers clear pathways to pitch, test, and when applicable deploy at scale applications from third party developers who previously may have been “boxed out” of locked, proprietary ecosystems. By offering a wider ecosystem of digital solutions the opportunity to deploy their technologies into an ecosystem that protects their intellectual property, while offering access to Department of Defense data, there is greater commercial incentive for developers to create technology for the department.
Experiment-Based Approaches to Scale
Further, the Open DAGIR model allows for an app to be scaled if there is sufficient demand. This is based on the principle that the best way to arm the warfighter and senior leaders with the technology they need is through a transparent experiment-based operating model and a clear process for vendors and government users to navigate application options. For example, an app may be initially fielded as a prototype through small business innovation research funding or other transaction authorities. Through experimentation with real users on live, operational data, the department can then assess performance and the size and scope of demand. Where applicable, this may scale to enterprise license contract solutions. These types of enterprise licenses allow the government to leverage its unique scale to drive down cost and ensure good stewardship of taxpayer dollars, while also providing industry with a more consistent and scalable solution to a software as a service offering that previously may have been purchased in a scattershot manner or in independent siloes.
Government-Owned Infrastructure and Services
Underlying Open DAGIR is our managed technology infrastructure and services that are interoperable across platforms and can integrate applications at scale. Government-owned technology infrastructure and services should be interoperable across platforms with clear mechanisms to define, protect, and compensate use of commercially available vendor intellectual property to align with Open DAGIR. This includes ensuring clear lines on what intellectual property belongs to which company and what are government-owned elements and creating clear business processes for customers across the Department of Defense to find, acquire, and integrate new software into their programs. The principles are translated into procurements to ensure that moving forward, key technology providers are held to a standard of modularity that will allow third party vendors to compete, and the government to choose between vendors to ensure the best fit, rather than being locked into a particular technology stack.
Balancing Security with the Imperative to Develop and Deploy
A critical objective of Open DAGIR is enabling the shift to data-centric enterprise architecture. Interoperability is defined by ensuring that defense networks can store, transform, and access data regardless of standards or formats and by using zero-trust architectures. This shifts the focus from data compliance with fixed standards to a meaningful assessment of whether applications and digital technology within the network can process and use of all the data regardless of origin or type. In that context the work is to establish a government-owned environment in which the development, security, and operations activities can occur. This enables the government to conduct a robust risk management process with integrated cyber security work in a way that is agile enough to allow capability experimentation, adoption, and operations. Therefore, rather than a process based on fixed standards, we can do a more iterative identification of, mitigation of, and recovery from threats or vulnerabilities. Part and parcel of this is the streamlined authority to operate process which can accelerate deployment onto department-owned networks. While this does not fully address the issues that arise from often slow processes, the integration of the risk management framework into the DevSecOps pipeline creates a streamlined and ultimately more predictable pathway.
Continuing Digital Transformation Progress
Ultimately, the objective is to create a federated ecosystem that has the right technical foundation, is accessible to the best developers, and delivers timely solutions to the warfighters. To that end, within our office we have leveraged the acquisition authorities granted by Congress to develop a streamlined competitive selection process through the Tradewinds Solutions Marketplace. We created a single front door to the commercial sector through a formal partnership with the Defense Innovation Unit. And we have shared this information routinely with industry through quarterly industry days.
While Open DAGIR is an important step in linking the technical requirements, capability delivery, and acquisition processes, it is dependent on evolving the Defense Department’s underlying approach to technology. In addition to the hardware-centric rigor, the department should apply a software-centric mindset to major acquisition programs such that the principles of interoperability and replaceability are built in from the beginning. Applying the Open DAGIR tenets across a spectrum of future procurements will enable the department to have access to best-in-breed technology solutions critical to the warfighter while also driving responsible stewardship of taxpayer dollars through competition within industry and leveraging the massive scale of the defense enterprise.
Radha Iyengar Plumb, PhD, is the chief digital and artificial intelligence officer at the Department of Defense. Prior to serving in this role, she served as the deputy under secretary of defense for acquisition and sustainment, and as the chief of staff to the deputy secretary of defense. She also led technical organizations within Google and Meta prior to her most recent government service.
Image: Chief Digital and Artificial Intelligence Office
Jesus Christ is King
