Foundations of AI

Published: August 21, 2019


A group of government experts gathers to provide insight on how the federal space is preparing for and implementing AI.

It is no secret that the era of AI has arrived, where the private, and more so the public, sectors are just scratching the surface with this advanced technology. AI has the potential to revolutionize any environment. According to Brett McMillen of Amazon Web Services, AI will work in a variety of settings within government - science, back office systems, customer experience – ranging from providing increased knowledge to understand the sun better, to Congress members more effectively responding to constituent calls. First, however, several factors must set the stage for how AI will expand in the federal space. A webinar conducted by FedInsider last month interviewed a team of nine experts in AI to expand upon the technology’s role in the government space today. Titled, “Building Blocks of Artificial Intelligence in Government,” the panelists explored the foundations needed to establish AI throughout federal agencies.


Data management, data councils, data inventory and freeing data from silos are all key factors in building a foundation for AI. Without good, clean data, AI technologies such as machine learning are tainted and unreliable. Pulling the data together is a top priority for AI government experts to pay attention to, according to McMillen.

Ted Kaouk, Chief Data Officer at USDA, stated that his agency is doing just that. Since 2018, the USDA had been undergoing a dashboard effort, unifying data from across the agency to ultimately understand data priorities and conduct maturity data assessments. Kaouk shared that USDA worked on unifying admin areas of data in 2018 and has been working on mission area data in 2019. When completed, USDA is hoping the dashboard will lead towards trend analysis and predictive modeling with AI in the future.

Strong data will also help with another aspect of AI – machine reasoning, explains Todd Dabolt, Chief Data Officer at Interior. Machine reasoning is structuring information in a way that is semantically understandable to the machine, stated Dabolt, so that the machine can reason when asked a question. Interior wants to apply machine reasoning towards improving customer experience. The agency found that customers were navigating through burdensome websites and sifting through piles of data to find answers to their questions. With machine reasoning, the data is made understandable by both machine and human so that the machine can quickly answer questions directed by customers.


When asked how GSA is evaluating AI with an eye to acquisition, Keith Nakasone, Deputy Assistant Commissioner of IT Acquisition stated that the agency plans to utilize public-private group settings to develop AI acquisition standards. Nakasone explained that in the past, learning from private sector’s business and use cases as well as challenges with different technologies has benefited GSA and feels the same will be true with AI. Additionally, GSA is planning a workforce buildout for AI so that contracting folks are prepared to understand the technology to quickly get AI to customers by emphasizing Statement of Objectives vs. trying to provide the requirements in acquisitions.


In a panel with DOD experts, when asked about the potential of AI, panelists harped on the democratization of AI. DOD’s Joint Artificial Intelligence Center, described Capt. Michael Kanaan, Co-Chair for AI at the U.S. Air Force, is the perfect model for AI because it will revolutionize technology sharing among the services, then other government agencies, and eventually international partners. Moreover, AI should not be limited to just data scientists, explained David Levy of Amazon Web Services, but knowledge and use of AI should be for the everyday , common developer as well.

When it comes to AI and the workforce, technician practitioners should have a two skill sets: application and research, explained Capt. Kanaan. Within the Air Force, the agency has started valuing computer language capabilities just the same way they do foreign language. For example, knowledge in a number of digital computer languages is treated equally valuable and functionally as those with extensive knowledge of various foreign languages. Jesse Rowlands, Data Scientist at the Defense Logistics Agency, also stressed the importance of spreading AI skill sets across an organization. His enterprise does not need many data scientists; rather it needs to upskill AI to the resources that will surround data scientists such as those in cyber, IT, cloud, and data stewards.


Guests throughout the webinar explained various AI expectations and initiatives taking place throughout the government. Some of those cases include several initiatives at USDA, to assist in automation, predictive modeling and improving customer experience. According to Ted Kaouk, USDA is standing up the Office of Robotic Process Automation to implement pilots related to financial management and to develop a framework for future AI endeavors. Additionally, USDA hopes to utilize predictive modeling in areas such as farm loan programs to better anticipate farmers’ needs ahead of disasters. Finally, as part of the Centers of Excellence activities at the agency, USDA is preparing to implement AI technology to improve customer experience by making it easier to find information across its 19 subagencies. On the Defense side of things, Capt. Patrick Schreiber, Commanding Officer at the U.S. Coast Guard’s Intelligence Coordination Center, stated that he foresees AI’s heavy use in readiness. For instance, the U.S. Coast Guard is looking to leverage AI with its International Ice Patrol, which monitors iceberg activity to provide warning to the maritime and military communities. If USCG can use AI in commercial satellites, explained Capt. Schreiber, the technology will help locate icebergs faster or even predict the weather better, and in turn help to protect aircraft and seamen and enhance readiness.