ROBOTICS AND AUTONOMOUS SYSTEMS
VIEWPOINT: The Monumental Task of Tackling AI at the Pentagon
The Pentagon is creating a new position — the chief digital and artificial intelligence officer.
Intended to be the successor to the Defense Department’s Joint Artificial Intelligence Center, the new office will better align data, analytics, digital solutions and AI efforts across the department, reflecting a “shift in organizational concept,” according to Deputy Secretary of Defense Kathleen Hicks.
The department’s Chief Information Officer John Sherman has been named as the acting chief digital and artificial intelligence officer, while the search for permanent leader continues — targeted for June 1 hire or earlier.
Whomever takes on the permanent role will face an enormous task: establishing foundational analytics goals for the Defense Department, helping it better define AI for its enterprise operations, and providing best practices for building out tools, processes and reporting to drive toward those goals.
Some distinct skills and strategic actions will be needed not only for the new director but for the estimated 200 to 300 individuals who will work under this office. Those are aligned to four domains: the director themself; the team they will build; the policies that will need to be implemented; and the tools and technology to get there.
Whomever takes on this role needs to build a strong foundation for the work. Perhaps the single most important step will be building relationships with the massive and complex Pentagon procurement structure from Day 1.
After going through the heavy lift of determining what will be needed to achieve objectives, it will be essential to ensure that carefully made plans are not unexpectedly altered because procurement takes a different acquisition direction. Pre-establishing dialogue and building trust will go a long way in the skillful negotiations that may often be required.
So will having prior understanding of all contract vehicles that are available to the office for consideration. Someone new to a role such as this usually does not have those relationships, so the effort should be undertaken as early as possible.
The director will potentially be responsible for disseminating information to four-star generals and other very senior stakeholders. The individual must be someone who can look at a large volume of data, interpret the information according to Pentagon objectives, and provide data visualization that is meaningful to those who need it.
Given the breadth of potential applications, it is necessary to get very focused on the high-level strategic goals of using that information.
Assess the main objective from the perspective of AI and machine learning analytics capabilities.
Is the goal to prevent threats? Is it to warn people about natural or other disasters? Who are the end users that may hold needed information?
How will the director best be able to visually represent that information to get concisely to the point?
While the director does not have to be a deep technical guru, they must effectively build and manage a team of people who are. It will be necessary to guide people possessing the right skill sets to take on the volumes of data the department collects, and create interfaces and reporting that allow for accurate, rapid access to exactly what is needed. Life and death may depend on it.
The personnel who will be the lifeblood of this office will take on something unprecedented. They must be of a temperament to appreciate and welcome the responsibility and the opportunity. Building and sustaining a cohesive, specialized team with such rare capabilities will involve some thoughtful leadership practices.
There is no need to start from scratch. It’s better to optimize the existing resource pool. Identify all incumbent resources and assets to determine both existing strengths and the gaps that will need to be filled to create and maintain a robust data organization.
Beyond data governance skills, the director must consider developers who are able to use the tools most effectively to produce the needed information. Look to other organizations that have implemented similar efforts — perhaps within the Pentagon or elsewhere in government — to see what they have done, what has gone well and where challenges lay.
What new talent is required must be thoughtfully targeted. The technical talent shortage will likely pose a challenge to the new director as it does for countless others in both government and industry. While there is considerable competition for AI and digitally savvy experts, roles such as this within the Pentagon have some unique recruiting advantages that must be touted.
The access to the breadth of data the Pentagon can offer is almost unmatched. The work that members of this team will do will impact national security, so the weight of the analysis and subsequent decisions based on it will be life-changing for others.
These roles carry much responsibility, but also the honor and privilege to work with information that could protect untold number of lives in the event of threats or hostilities.
It will be important to establish very clear definitions of what each role must do and create training to make sure everyone understands what is expected of them and how to do it. Beyond identifying and recruiting talent, build a redundant talent infrastructure so when people inevitably do depart, there is a “next man up” readily available to prevent disruptions in operations.
Part of the overall team must also include a data governance organization. The director will need data stewards who can develop a thorough understanding of a specific data domain and their specific department’s interests — what data the group produces, what the information contains, how it is used, and who has access to it.
The team needs data administrators who can support and manage all the architectural infrastructure required to enable appropriate access. It will also be necessary to have a data governance council that convenes frequently to discuss the types of decisions they want to make and the data they will need to inform those decisions. The council should know which data steward should be tapped to pull particular information, and what access should be given to what Pentagon departments based on their final intention for using particular data.
However, data management is not just about people, it's also about policies. In parallel to building the organization, the new director will need to assess existing data governance policies to determine if any changes may be needed.
Particularly in this domain, experience matters. The Pentagon will be well served to select someone who has already done this work at another government organization. The new director will need to articulate very clear policies regarding how data should be utilized and who should be accessing what.
It may be helpful for the director to seek counterparts and liaisons in other agencies for consultation and collaboration, and even to emulate some of their best practices.
A unified approach will be critical, but adoption may be challenging. It is well known that each military branch has its own way of doing things. The director will be tasked with setting key policies that will be disseminated across the department with the goal of all branches following suit. For example, he or she can set higher-level policies for how to clean data, how to catalog information, how to disseminate that information to other federal agencies, and even suggest compliance with a catalog of specific data governance tools.
Getting buy-in from across the branches is an immense and frankly formidable task. The director will need trusted, skillful liaisons and data stewards that can build cooperation and collaboration with the groups they will represent.
As mentioned earlier, any new director does not have to be deeply technical. But the individual must have an adequate technical understanding of the specific analytics tools that will be employed, the value they provide and trends that will be needed to get the job done.
That doesn’t mean starting from ground zero. There are technologies already in place, decisions that have already been made, and legacy contracts with a breadth of providers. The director will need to take an organizational deep dive to see what is already in use and what may need to be acquired.
The same applies to data. The director must determine what information exists where and its level of sensitivity. A tools rationalization exercise can assess what tools different departments within the Pentagon are using, if different tools are being used for the same purpose, and what those specific purposes are: for example, data visualization, data management, AI/ML, and data warehousing. Which tools and datasets are cloud-based? Which are on-premises? What kind of hardware and software is used? Which tools are working the best?
After an exhaustive assessment, formal recommendations and actionable guidance on the best set of tools for everyone to adopt can be issued, substantiated by data and facts.
Adoption of AI is already happening within the Pentagon, but it is about to significantly increase. Success will depend on trust – in the data and the recommendations that come from it.
Currently there tends to be a high level of mistrust in AI and ML because people see them as a black box they don’t understand. The director will need to demystify these critical new technologies for those who need their output. They must understand the Pentagon’s key objectives and be able to explain how they are applying leading-edge technology — using what specific information and building what particular model to predict what specific outcomes.
Consistently and patiently providing that level of visibility will build trust over time.
There is a tall order waiting for the individual who takes on this role. Whomever that is should come with a vision that empowers them to set achievable targets for their first 100 days, the next 100 days and what comes next. Think two years out to what a successful Pentagon AI and digital program looks like, then build from there. The right team, tools, policies and confident persistence will get us there.
Aparna Kumar is the analytics practice lead on the public sector team at World Wide Technology.
Topics: Robotics and Autonomous Systems, Robotics, Defense Department