Artificial Intelligence

Lots of talk about AI, but are agencies spending money on it?

Federal agencies are looking at different ways to utilize artificial intelligence. But just how much of their budgets are agency leaders willing to put AI?

Eric White@FEDERALNEWSCAST

April 15, 2024 1:51 pm

8 min read

You’ve been hearing a lot about it lately, present radio program included. Federal agencies are looking at different ways to utilize artificial intelligence. But just how much of their budgets are agency leaders willing to put AI? Federal News Network’s Eric White talked about that with Professor Greg Dawson of Arizona State University on The Federal Drive with Tom Temin. Dawson recently conducted a study with the Brookings Institution to look at how the federal government has been purchasing AI.

Interview Transcript:

Eric White Absolutely. So why don’t we just start from the beginning and tell me what exactly you all were trying to find with this, was it a review or was it actually research? What did you all undertake here?

Greg Dawson Well, so this is our second paper is series. We’ve been working with folks from Leadership Connect, and a couple of years ago they provided us all of the information on AI spending within the federal government. And it had been, 18 months, two years since that last study. And so we wanted to see what had changed.

        Join us May 6 at 1 p.m. EST for Federal News Network's Industry Exchange Data where we'll explore how you can employ data to modernize your agency. | Register today!

Eric White Well, a lot has changed, well in the technology world at least.

Greg Dawson Yeah. Well we were surprised how much the spending changed as well. In the report we did 18 months, two years ago, Department of Defense was about, I think 76% federal AI spending. And they had about 250 contracts to this past that of data that we looked at. They are 95% of federal AI spending in the Department of Defense and now have over 650 contracts. Sort of our interpretation of that is Department of Defense has found some use cases for AI, and they are moving from the thinking about it as concept, to this is how we’re going to practically implement it. And we think that’s exciting.

Eric White Yeah, that was definitely one of the big takeaways I took away from reviewing your research. Does the 95% number take into account that the Department of Defense they do a lot of contracts anyway? Are they actually just purchasing the most AI in general, even taking into account how many contracts they do on a daily basis?

Greg Dawson They are just purchasing the most AI general. My colleague Jim Dunford, who’s a coauthor on this paper, had the best line that, if you look at federal AI expanding all other agencies other than Department of Defense are just rounding errors. The DoD is so prevalent in terms of where they’re spending is.

Eric White Gotcha. All right. And were there any agencies that at least came a little bit close to at least not being just a decimal point?

Greg Dawson No, not really. The next couple of agencies were NASA and HHS, but they are just such in distant second and third place that they’re almost irrelevant.

Eric White Gotcha. All right. And so how deep did you all dive into the kinds of AI technology? What are the numbers look like breaking down just the Department of Defense purchases?

Greg Dawson Yeah. So this is where it really got interesting, because up till now most of the spending has been on study related stuff. So the NAIC code I think was our code number 51, which is all about studies and things like that. The shift now has been dramatically towards code 54, which is all about implementation work. So we shifted from the study to the implementation. And and as I said, clearly the Department of Defense’s found some use cases for AI and they are moving forward aggressively and doing it. And we can also see that with the different vendors, the vendors that we looked at a couple of years ago did a lot of study work. And now, as we predicted in our last study, so kudos to us for that. The the integrators have moved in. So people like Accenture Federal Services who are all about implementation, they are now a big player.

        Read more: Artificial Intelligence

Eric White You use the prediction word. I’m going to hold you to that a little bit later. But first let’s reset here. Gregory Dawson is a professor at Arizona State University. We’re discussing AI spending in the federal government. Looking through those vendors, is it a pretty good spread, or was it like a lot of the defense spending where there was the big five or anything like that? It seems as if there’s a diverse group that the Defense Department is actually purchasing a sizable percentage from.

Greg Dawson Yeah. And I’ll go back to the P word, the prediction word. One of the things we noticed, 18 months ago was, again, there was a whole bunch of small guys who curiously were often co-located with a military base, and we thought the bigger guys would move in. And indeed, the bigger guys have moved in. And there’s still a lot of spending on smaller companies as well. But the big guys have definitely moved in. And again, our speculation is that this has moved from, what a quaint idea to we’ve actually got some ways to be able to approach this, at least within the Department of Defense. We talk about other agencies. I’m not sure the same statement could be made.

Eric White Either that or the big guys bought the little guys to move in.

Greg Dawson Entirely possible.

Eric White I know that you said it’s not even really sizable, but the agencies that are making those tiny purchases of AI, what are some of their use cases for it?

Greg Dawson So it’s not clear to us what their use cases are. And that’s actually something that concerns us. What are the things that we’ve talked about in our numerous Brookings reports, is the fact that the way the US as a country is approaching AI is sort of, to quote Chairman Mao for a moment to let a thousand flowers bloom. So the US as a country is looking very broadly about AI and where AI spending could be contrasted to China, for example. And China is all about they’ve got 2 or 3 things that they’re pushing forward. Our speculation is that, eventually, the US approach of very diffuse spending is that it’s impossible or it becomes much more difficult to learn. There’s a concept in education where the more you learn, the more you’re capable of learning. The problem is, if you’re doing a whole lot of spending across a whole lot of different agencies, you don’t have the opportunity to be able to learn those things. Department of Defense is unquestionably doing it right. These other agencies, I’m not sure that the learning is going to take place across like we want to. And I know there’s some federal regulations and boards set up to be able to address it. At this point, it seems like it’s a lot more of a concept than actually being seen in practicality.

Eric White Yeah, and that provides me a perfect segue into my next question, which is not every agency has one, but a lot of them have put out artificial intelligence strategy. The White House has said some things about how they want agencies to use artificial intelligence. If you put those strategy side by side with your research and the strategies on another screen, it doesn’t seem like they’re matching up in agencies are doing what they set out to do regarding AI.

        Sign up for our daily newsletter so you never miss a beat on all things federal

Greg Dawson Yes. But so are they approaching it that way? Yes. But I gotta tell you, a lot of those plans are very aspirational in nature. We want to be all things to all people and do good with it. That’s good, but in terms of the practical implementation, that feels a little bit weak.

Eric White Gotcha. All right. Well, so now I’m going to hold your feet to the fire. What are some of the predictions you think you all see or are you planning on continuing this research? I imagine you are, it’s fascinating and I can tell that you’re you’re very interested in it. What are you all expecting to see the next go around?

Greg Dawson So, yes, my colleagues and I will certainly continue our relationship with Leadership Connect, who’s provided this. Provided all of the data to us. So what we expect to see in the future is, I think the Department of Defense models perfectly. What other agencies are going to do. So the Department of Defense sort of nibbled around the edges for a couple of years looking into AI and has now found some use cases that it clearly used is worth exploring. We think that’s what the other agencies are doing now. They’re just kind of dipping their big toe in the water to test their temperature. We have little doubt that they’re going to be able to find some appropriate use cases and that they’ll probably be a couple of years behind where the Department of Defense is in terms of actually being able to implement these use cases. But again, I’m going to circle back to, man, this has got to be done in a coordinated fashion. If we’re doing one offs across all the different agencies, we’re not taking advantage of the wisdom that can be gained from a more coordinated effort. Again, what the White House would put out, in terms of their strategy is quite good in terms of coordination. We would really like to see that stronger and enacted more fully. But mark my words and let’s have this conversation in a year or 18 months, I think we’re going to see this shift to be fairly dramatic, in terms of what other agencies have moved past the dipping their toe in the water aspirational to something where they’re really no doubt, implementing AI in the true sense of the word.