Artificial Intelligence in Government: Where to Start

Just don’t expect a genie to grant all your AI wishes. 

If the federal government invested in artificial intelligence, it could potentially save 1.2 billion in labor hours, $41.1 billion annually and increase mission-delivery speed by automating processes, according to a recent Deloitte study. It’s difficult to argue the potential of advanced technology considering its growth in commercial industry, but is government prepared for the adoption of AI?

To answer this question, we asked an expert in both industry and government to comment on the value and current standing of AI in the federal government. Justin Herman is the lead for the Emerging Citizen Technology program at the General Services Administration, which focuses on AI for citizen services, blockchain, virtual and augmented reality, and social tech. The program intends to help agencies adopt new technologies by building the road maps, shared services and policies to do get there.

Greg Castellucci is a systems engineering manager for Brocade’s federal division. He’s been with Brocade since 2010, supporting federal customers both on the system integrator community side, as well as directly with federal customers in the defense and intelligence community.

GovernmentCIO Magazine: In your experience, how are you seeing the expanded use or interest of AI in government? Where do you think the federal government can find value in AI technology?

Herman: I can barely keep up with the interest. Every single day, agencies are reaching out with new ideas on how these technologies can improve their mission. There are two large reasons for that. One, it’s part of the success of data science initiatives, because now agencies have more and better access to their program data than ever before. Two, technology is evolving at a pace that is better and faster than ever before, thanks to the advancements of U.S. businesses. So now we can take something like AI, which the U.S. government has been working with for decades, and bring it out of research and development, out of the laboratories and put it in these solutions and into the hands of any program manager.

Castellucci: The end goal of artificial intelligence is basically where everyone is trying to get right now. The idea where you can have a machine make the decisions that a human would normally make to speed up resolution and actions is really the goal of any agency, to get the human interaction out of the way. '

The problems that we see today tend to be human related. The hands-on work from the command line that a lot of network engineers still do are very error prone. The industry is looking at artificial intelligence for this, and there are a lot of things happening today that are moving the government agencies toward that idea.

GCIO Mag: Where should the government start with AI?

Herman: We start with focusing on the problems in government. If an agency or company acts like their problem is that they’re not using AI yet, that’s not sound. We want to be able to articulate the challenges of government that potentially emerging technologies like AI could provide a solution for. It’s focusing on the mission need, becoming best friends with your data scientists, program managers and privacy and security experts.

In order for us to get smart and get ahead of emerging technologies like AI, it requires identifying skill sets and missionaries within our agencies that may not have been at the table before, but have to be to get better outcomes.

I hear this every single day from our federal agency colleagues … they are being asked to do more with less. AI, machine learning, smart automation, AI augmentation of programs―these are all things that will empower public services to do more with less and do better with whatever we have.

Castellucci: In order to get there, the first thing that has to happen is adoption of the hardware infrastructure that can actually talk the right machine languages. This is done with application program interfaces. What we have today is a mishmash of hardware across the federal space that is very old infrastructure. The hardware layer has got to become smarter to support things like intelligent augmentation, which is actually one of the first steps to get to AI.  

GCIO Mag: What challenges is government facing to get there?

Herman: In our interagency AI community, we discuss things like ethics, privacy and security―the same things that federal managers focusing on cloud, cybersecurity or website development discuss. These are the same questions that we have for AI, right down to even accessibility for persons of disabilities. These are things that we not only have to address, study and come to a consensus on within the federal government, but also we believe that AI and machine learning can play an integral role in supporting solutions for all programs.

In turn, AI can help improve effectiveness, accessibility, efficiency, security and privacy. So AI, in a way, is a catalyst and a not-so-secret ingredient to the success of so many of our programs moving forward.

Castellucci: I think the biggest challenge is the disjointed communities that are all within the same agency. In order to be able to implement AI or take intelligent action, you need to be able to have visibility, to not only what you have, but also what your data is doing. If you’ve got dozens of enclaves of communities of interest within an agency and disjointed networks, how are you going to correlate what your data is doing to make decisions?

So the biggest challenge I think in an agency-by-agency basis is first consolidating and shrinking down all the numerous enclaves of communities of interest. Once you have visibility of what your data is doing, of everything in your infrastructure, then you can get to a place where AI can actually be implemented.

I think one of the biggest drivers for a lot of the federal agencies move to the cloud is exactly for this reason. They need to be able to get the visibility of their data to implement artificial intelligence ideas on those data sets.

GCIO Mag: How can agencies identify the internal processes that AI could make more efficient?

Herman: We’ve seen over the last two years a major emphasis on improving the customer service of federal programs. Smart automation through machine learning and AI is a major component of that; it is one of the ways in which programs are going to benefit from AI.

I think in some people’s minds when they hear the words “artificial intelligence,” they think they’re going to rub some magical lamp and a genie is going to come out and grant all their wishes. But truly, what we see happening in agencies are these thin layers of applications, and you don’t even know that you’re using AI, but you do know that the process just went a lot better.

Another is workflow process. The potential is everything from predictive analysis and research, to improved customer service and improving the processes of how we do business itself. It can help operationalize that data and make it actionable and digestible.

Castellucci: Take a look at a workflow that is very tedious, that you have multiple hands doing, and then see if it’s possible to automate that one little hanging fruit. One workflow that we are able to automate and agencies can look at is provisioning; provisioning of users, hardware, appliances or local devices on the network. There are tools out there right now that can automate those kinds of workflows very easily.

Once you get your groups up to speed on how to use these types of open source tools and create these workflows, understand programmatic languages, you can start to grow and eventually automate your entire environment. Find those really simple workflows, and then implement those open source tools or automation tools to simplify those workflows; those tend to be the easiest paths to get there.

GCIO Mag: Where is AI already being used?

Herman: Check our website.

Castellucci: The AI that we see today that actually is in use is not as much in the networking environments, enterprise or the data centers, as much as they are in focused programs like mission-system environments, cybersecurity tools for finding anomalies and specialized programs like guide systems and targeting systems in the scientific community.

In my industry, I think we already are seeing it in the cybersecurity space, more referred to as machine learning technologies being applied to hacking attempts. The AI systems in cybersecurity, or the machine learning capabilities, are spotting these anomalies right now.

GCIO Mag: Do you find that agencies are hesitant to adopt AI?

Herman: I’m well-equipped to deal with the hype cycles, and while there is a lot of hype around these solutions, it’s for good reason. That’s part of what we do, we distill those practical cases and those steps forward that truly will impact programs for the American people, without getting caught up in the hype.

I think a lot of what we do hear is cautious optimism. Optimistic because we know the use cases, we know the availability and impact they can have, but cautious because we’re still learning, companies are still learning, and we have a long way to go in developing the policies, road maps and resources that we need to truly mainstream. That’s what we’re doing, that’s what we’re dedicated to, and we will achieve it.

Castellucci: I do not; I think they are all looking at exploring it right now. I think what they’re learning is we got a ways to go to get there. There may be small pockets where you can implement AI and intelligent automation-type technologies really early and build that from the ground up to be AI-ready. That, I think, is where most agencies are kind of looking at being able to do AI, but they’re picking small pockets, and they are looking into that, absolutely.

GCIO Mag: What advice would you give to agencies wanting to get into AI?

Herman: There are two sides to this. We can’t just use the same ingredients; we need fresh ideas, and these solutions might come from small businesses and startups that have never worked with government before. And so on one side of it, if you have a need that hasn’t been met yet, there is a whole community of people just like you waiting for you. Whether you’re a data scientist, coder, program manager or HR specialist, whether you’ve been in government for years or you’ve just started with the new administration, there is a place for you at the table.

The other side of it is for the businesses that we need so badly to come to the table to help provide these solutions to improving public services. We can’t do this alone, and we can’t let the federal government fall behind.  

The answers have been edited for clarity.

Add new comment

This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.

artificial intelligence
justin herman