HSDF THE PODCAST

Part 2 of 2: Incorporating AI into Acquisitions: Innovative Strategies & Challenges

• Homeland Security & Defense Forum

Share episode

Welcome to “HSDF THE PODCAST,” a collection of policy discussions on government technology and homeland security brought to you by the Homeland Security and Defense Forum. 

In this second of a 2-part series, the discussion continues with the potential of AI in improving acquisition and procurement processes within the government. It highlights the use of AI in planning acquisition programs, conducting oversight, and enhancing procurement processes.

Featuring:

  • Polly Hall, Senior Advisor, Office of the Chief Procurement Officer, Department of Homeland Security
  • Dina Thompson, Deputy Head of the Contracting Activity, Deputy Assistant Administrator, Contracting & Procurement, TSA
  • Kimberly (KC) Milligan, Component Acquisition Executive and Executive Director, Acquisition Governance and Oversight, Customs and Border Protection
  • Elizabeth Moon, Managing Director, Custom Engineering, Google Public Sector
  • Soraya Correa, Former Chief Procurement Officer, Department of Homeland Security 

This discussion took place at the HSDF’s 4th Annual Women in Homeland Security Celebration on March 22nd, 2024. 

Follow HSDF THE PODCAST and never miss latest insider talk on government technology, innovation, and security. Visit the HSDF YouTube channel to view hours of insightful policy discussion. For more information about the Homeland Security & Defense Forum (HSDF), visit hsdf.org.

Speaker 1:

Welcome to HSDF the Podcast, a collection of policy discussions on government technology and homeland security brought to you by the Homeland Security and Defense Forum. In this second of a two-part series, the discussion continues with the potential of AI in improving acquisition and procurement processes within the government. It highlights the use of AI in planning acquisition programs, conducting oversight and enhancing procurement processes. Featuring Polly Hall, senior Advisor. Office of the Chief Procurement Officer, department of Homeland Security. Dena Thompson, deputy Head of the Contracting Activity. Deputy Assistant Administrator, contracting and Procurement TSA. Kimberly Casey Milligan, component Acquisition Executive and Executive Director Acquisition Governance and Oversight. Customs and Border Protection. Elizabeth Moon, managing Director, custom Engineering Google Public Sector. Soraya Correa, former Chief Procurement Officer, department of Homeland Security. This discussion took place at the HSDF's fourth annual Women in Homeland Security celebration on March 22nd 2024.

Speaker 2:

Okay. So, casey, I'm going to ask you a question. As the Component Acquisition Executive at CBP, you have the responsibility for setting policy and conducting oversight of major acquisition programs. Can you share with us some of your thoughts on how AI can help you improve acquisition and related processes, in terms of, like the acquisition programs, establishing them, but also conducting the proper level of oversight?

Speaker 3:

Sure. So I talked a little bit about how long it can take to plan an acquisition program. So the best case scenario is you start your AI journey at the very beginning of the planning. You identify your capability gap for what you're trying to buy, you work with industry, you identify different AI alternatives that you can use and then you pilot and you do technology demonstrators and in doing that you'll know if it can help you meet your capability gap. But it'll also give you the opportunity to mitigate any risks associated with, you know, data security or privacy security we keep hearing that over and over again and you know we do that in CBP through our acquisition program managers.

Speaker 3:

They work with our chief technology officer and our chief information officer, going all the way back to IT always, and they set up several years ago AI Center of Innovation and you know they test everything from automated commercial environments to truffle machines and they have several use cases, including that they're working on right now, including identifying and targeting, you know, criminal organizations, identifying goods made of forced labor, maintaining the security of our supply chain. So a lot going on in that in that arena and we use that heavily in in the very beginning of our supply chain. So a lot going on in that arena and we use that heavily in the very beginning of our planning for acquisition programs. You asked me about some tools that we use for our processes. Going back to the procurements, the contracting of it, we have a few robotic process automation tools, which I'm told is AI on training wheels kind of.

Speaker 4:

Okay, good.

Speaker 3:

And three of them that come to mind are that's a brand new one for congressional notifications. If contract action requires congressional notification, our contracting officers know about it because they use this RPA. We have the obligation RPA that helps significantly at the end of the fiscal year, and then another one that we just started that helps us with our PR packages. That might have stole that from you I heard you talking about it but again, it's really about continuing to work with industry as we're moving forward and holding those industry days where we can talk about CBP's views on AI and then what our acquisition programs are that are coming out in the next few years, so that we can identify AI potentials.

Speaker 2:

So you raised an interesting point in your comment to Dina, which is one of the things that I think is happening more and more across the government at least I know in our profession and the acquisition profession is happening more and more is the sharing of the technology and the data and what we're learning.

Speaker 5:

In other words, we're not. In the past, sometimes we worked in stovepipes. I don't know, polly, what's your experience in this area, please? I think we're doing a much better job right, even over the last, you know, five years of working across not just the inter department right. The efforts that CBP has undertaken around RPA and some machine learning capability for these process improvements I mean, stemmed actually from a crowdsourcing concept right that we launched at the department level and it was using a process to bring user-centered ideas to bear on improving our acquisition process. Cbp saw what we were doing at the department level and said, hey, can we use your system to launch this idea at the component level? And so even just sharing the frameworks where we solicit ideas from our workforce is a piece of that.

Speaker 5:

A few years ago and I know some of you guys in the audience were part of the journey with us but we undertook an effort to work across nine federal agencies to experiment with our artificial intelligence around accessing contractor performance reports from our CPARS contractor performance assessment reporting system and had some really fantastic proof of concepts and initial rollouts.

Speaker 5:

It kind of evolved because of some policy to market research capability, but today we fielded that and you know, we led an effort, but it was co-funded by nine agencies, and now they are commercial products that were built by the procurement community with industry, federally available for any agency to buy. And so, you know, dina talked about the fact that we are under-resourced right in the mission support departments of procurement and acquisition, and so we have to lean on each other and share the demos, share the costs, share that engagement and then benefit for the greater good. And so I think there's a lot more we can do in this space, but we've had some really good early successes. We've learned a lot along the way, and I think that is our only path forward, especially, I mean, as we sit here wondering if we're funded right as of as of midnight, like we don't have a choice. We have to be smart about how we build this capability and then, and then procure it.

Speaker 2:

Yeah, no, I agree, I'll hardly agree. So at this point I'm going to turn to the audience to see if you all have any questions for our panelists. Okay, we have some. Do you have a microphone? I'm going to get your microphone and I think this is probably to Polly, but probably to all of you.

Speaker 2:

As Sarai mentioned, we have a lot of data, but we haven't turned it into information. So we have these two sources of vehicles, a lot of pricing information out there, and right now, old people like me use websites, right. So is there some way that you guys could look at some effort or maybe it's going on now, I don't know To put that information somewhere, look at it from a AI standpoint and then allow us to be able to give a tool to someone like Kayla and Sarah as a legacy, because this is going to be the environment that they're going to be working in so that they can be better informed buyers. I mean, we have the data. We just don't need them to be searching 10,000 websites in order to do market research. So there's a way we can do it better and there's a way that we can create better and informed buyers so that we can make better buying decisions quickly, because that's what we have to do, robin, I mean yes right to do, robin.

Speaker 5:

I mean yes, right, we all vote yes, and it's funny because you mentioned prices paid. The reality is, right now, data is everything right, but our data is disparate. It's in different systems that are not interconnected, and our fields right. Any database administrator would be horrified about the inconsistent approach that we have across the federal sector, and any vendor right who works in multiple agencies or even the same agency with two different contracting officers.

Speaker 5:

Right, knows when structures differ, right, how we segregate costs under contractors differs, and so getting our data to a consistent standard so that it can be leveraged across different systems and aggregated and utilized by the AI is our first step right, and I bet you could talk for hours, elizabeth, about this as a challenge, but I mean, that is kind of our first step right, and luckily we are paying attention to it and you know, at least for the next year, right, the better contracting initiative that's been stood up by the Office of Federal Procurement Policy.

Speaker 5:

The high def data environment is a big one, and Dina can speak to how TSA actually shared clinical level pricing data as part of a pilot so that we can understand the manual nature of how that has to be pulled now so that we can learn about how we can develop better, consistent methods to collect data so that it can be leveraged by these tools. And I think we are very nascent from a procurement community in that kind of growth process, but getting a hold of the data structure is our first step, and I think in that kind of growth process, but getting a hold of the data structure is our first step, and I think that's kind of where we're spending some time these days. I wish it could be faster, but that's the reality of where we are.

Speaker 7:

Yeah, I'd love to weigh in. This is a great challenge for us to potentially partner on, because the technology is really improving leaps and bounds this field of generative AI. It's so exciting because of how much we're able to do and how fast things are going and new inventions keep coming out. So actually now we do have tools now that can go in and read very different types of data and understand it immediately without, and you can also train the AI on your specific you know forms or structures. But we really should consider working on that. We could, you know, you can actually, you know, for example, with our tools you can actually, you know, comprehend a form you know. Just point the AI at the form and database and and really now there are tools that will allow you to synthesize that information and sort of bring it into a place. That, even without as this is a big difference between the old AI and, you know, generative AI is that now you don't have to do quite as much of the data prep we still want to make sure that we're thinking about what you know, security of it, and we have tools to do that as well. You know, this is for an internal audience. This is for a different audience, but the tools have come, really come a long way recently and you can do a lot more.

Speaker 7:

We also introduced recently our you know, and you'll hear a lot of the announcements it's like what does this mean? We introduced our Gemini 1.5 model recently, which includes a very long context window of up to a million characters. What does that mean? Well, the context window is the information you're giving the AI that you can then analyze in real time. A million characters is enough for a couple books or the entire US tax code, for example. I'm sure there's a lot of DHS policy documents that you can enter right into the prompt and then you can actually have a conversation from there and ask questions. And our tool does provide citations so it can link right back to where the information was found so that you can do that due diligence and check to make sure that the information is what you were looking for. So there's a lot of opportunity here. We'd love to partner on this particular challenge. It's a great one for improved productivity.

Speaker 2:

Yeah, the only thing I would add is because we've been talking about pricing for years and one of the things, one of the challenges with pricing is that no two prices are actually the same, and so when you're trying to do that comparison, you know, when you're trying to use that pricing and making a comparison, you have to understand what's behind those prices. What was the basis for the price? And that's where the AI can come in. Very handy that it can take this AI and, for lack of a better term, tell you the flavor of truffle that you're looking at. Right, what's the flavor of that price that you're looking at? I got my truffle in. I got my truffle in. Okay, next question we have a question over there.

Speaker 4:

Yes, I'm Gigi Bakhtar from Hive Group. I'm going to cheat and squeeze in two questions. So the first is that what is your methodology for overcoming across the workforce? Overcoming fear and resistance at all levels of the workforce is when you speak of teams. Also, how do you find sometimes, or how do you attain a common language, that different members of the team can speak and understand the same language? Because I'm sure that you've faced this before.

Speaker 8:

So thank you, I can try. So as far as the fear, I think that as our workforce we get new people like Kayla and Sarah. They wish we would hurry up. They think we're too slow. They can pick up the processing and the way that you use these things quickly my kids do all the time. They're just like mom you're taking, just give me your phone, mom, we'll do it.

Speaker 8:

So I don't think that's going to be. It's going to be the old fogies like me that are struggling. Rick, tell me again how to do that. Can you give me some steps that I have to take? So I just think it's probably going to be something that is going to be introduced and we're just going to have to learn how to use it. And then, what was the part two? Common language in IT, at least in TSA in our IT area that are collaborating with folks in my office that are developing draft sample clause language that will kind of tell us and our contractors how we intend to apply AI. And then, if we get it right or halfway right, we'll elevate it up and it will go, hopefully, to the FAR councils so that they can incorporate it into our federal acquisition regulations and, I'm assuming, in IT regulations as well.

Speaker 2:

So just a quick comment on the whole fear thing. It's change management. It's change management like any other change management process, and it's about talking to our workforce, educating them and showing them what this technology is, giving them the opportunity to experience it. All too often we talk about processes and data and technology and we talk about it like, at this level, we need to put it in their hands and let them experience it and see it. That's what removes fear. That's the best way to remove fear. With respect to common language, when we talk about common language across AI, that certainly falls to the IT community, but one of the things that the Defense Acquisition University and the Federal Acquisition Institute have done is now joined forces to come up with a single certification for the contracting profession, and that ties right back to some work that was done by the National Contract Management Association to standardize the contract management language so that we speak in the same language across the government, both buyer and seller community, and I'm pleased to say that a lot of folks are adopting that contract management standard, which means that we're all going to learn the vernacular in the same way, so that when you're looking at our contracts, there is consistency. There's consistency in the way we present concepts, the way we train people and, more importantly, how we interact with industry. So I think those are some areas where certainly we can improve.

Speaker 2:

That's a great question and I agree on the fear and resistance. I agree with Dina, the fear and resistance. Us old folks and, by the way, I'm all in but it is the newer generation of worker that's coming in, that's pushing us in this direction, because they're going. You're using a PC, they use a phone, they text, right. They'll write a whole contract on a phone right now, you know, because they also want a contract that's like one page, right. So I don't think there's fear and resistance there. So I don't think there's fear and resistance there. Looking at my watch, we're getting close to the end of our time and I want to be conscious of that. Is there any other question, any other pressing question that we can answer? I was looking over here to see if there was anybody there. Okay, oh, there's one. I'm sorry.

Speaker 6:

Thank you, ken Phelan, with Storm Center Communications. As you know, as the federal government dives deeper into technology modernization to respond to its AI big data language modeling needs, just find that it needs to engage with the private sector and industry, and as a part of that I understand there is the Congressional Appropriated SBIR Small Business Innovation Research where the federal government is engaging with technology companies that are emerging that can help solve their technology needs. So, with that perspective, are you or what's your take on engaging with companies that are going through the small business innovation research process and how to leverage those innovative and emerging technology companies to respond to AI and other technology platforms?

Speaker 5:

I can talk.

Speaker 6:

Also, are you beginning to, I guess, get your contracting and procurement specialists up to speed on how to leverage that powerful technology or that powerful status?

Speaker 5:

Well, I think that's probably what I can talk to most, and then I think some of the folks from the components might be able to give some specifics. But we actually recognized the fact that we have great authorities at the department. Right, we potentially have a bill in front of the Hill that looks like it's on a path to passing that would extend our OT authority for seven years instead of annually. Right, like whew, that would be helpful. We have our commercial solutions opening pilot program that is in play that has a multiple year authority. Now we have, as every other federal agency right, access to support companies and bring capability on board through the CBER program and the STTR, sbir and STTR programs.

Speaker 5:

But we have done maybe not as good of a job as we needed to do in informing not just our contracting professionals but our acquisition professionals, right, our entire acquisition community, about what those authorities are, good use cases for those authorities, how and when to use them, how to make decisions about.

Speaker 5:

You know, as you write and frame mission problems into requirements documents and shape your acquisition strategy, how to consider that whole portfolio and toolbox. And so very recently our chief procurement officer challenged a team of us at headquarters to revision how we're training and upskilling our acquisition community at the department on that, and I think you're going to be seeing and this is going to be part of our discussion at Thursday's strategic industry conversation as well there's a panel with Sarah Todd, our policy director, and Katie Crompton, our new procurement innovation lab director, talking about some of this work. So I think you're going to see more in the coming year to two years on knowledge and capability of our acquisition teams to bring these authorities and programs more expanded across the department so that we can access the capabilities and the companies that would bring solutions appropriately through those authorities as well as our traditional acquisition authority.

Speaker 2:

So use cases these guys might have more practice on, but I do think that it's really One more question over here, so we'll try to do two in one do a use case as well as answer the question.

Speaker 3:

Hi all, china, sherry, ecs Federal. I think that implementation of AI is necessary to create efficiencies, but do you all see it increasing the risk of protest as these tools are increased and, if so, how are you mitigating against that?

Speaker 8:

So I can try. We're not there yet, but in my mind I don't think it would increase protests. I think it would decrease, because we can make sure that we're doing things consistently and more accurately and reduce the likelihood of us making mistakes that humans just make right. Humans get tired. Humans get distracted. They lose track of where they were. The use of the AI can help keep us consistent and on track, so I think we'll be better off for having incorporated that into our processes.

Speaker 2:

And I'll give you a use case. If I were queen for a day, we would have generative AI generate the debrief. Think about that.

Speaker 1:

AI generate the debrief.

Speaker 2:

Think about that. Generate the debrief.

Speaker 5:

Now you go challenge the AI and leave us alone.

Speaker 2:

But think about it. I mean I say that partially in jest, but think about it. You could feed it the information and it can generate that debrief document for you and it will probably comply with regulations and all that good stuff and make it a little more difficult for companies to just challenge for the sake of challenging. There are protests out there that are valid and I will always defend a company's right to challenge the government if the government has made a mistake, because we owe it to our citizenry to do the right job. But there are folks that file frivolous protests and I think a generative AI tool that generates that debriefing and, by the way, then Dina doesn't have to explain it here. Here's your tool Go talk to the PC or the phone or whatever.

Speaker 2:

So in these last minutes and I know that we're running out of time, but I can't resist I got four incredible professional women here and I just want to ask each one of them, you know, to please give us one good nugget of advice for the women in the audience who are either in any acquisition field or even in the professional, in any other professional field, a good nugget of advice about navigating their careers. So I'll start with you, dina, how's that Okay, so I could.

Speaker 8:

I've done whole conversations on this, but I'm going to just pick one. It's relationships. It all comes down to relationships. Build your relationships, expand your network to the extent that you can Come to forums like this so you can meet and greet people and you don't want to wait for an emergency before you ask for someone's business card. You need to know who the? People are. They need to know who you are when they hear the name Soraya. You want them to go.

Speaker 3:

Oh yeah, soraya Not.

Speaker 8:

Oh, soraya. So you want to get out there and you can't just sit behind your desk and work hard. You got to know folks too. Exactly Right. Thank you, I love that.

Speaker 7:

I would just say jump in, just jump in, don't don't think um, just just go because, um, there's a lot of opportunities to embrace and we're sitting on one right now. Generative AI is going to really change technology and the landscape of the way that we work. Really, the way that we work will change. How many of us even heard the word large language model, like a year and a half ago? Probably nobody in the room or very few of us, and I was at a big tech company, so I think you know there's always a great time to just jump in and learn and you know, that's kind of one of the things I've used in my whole career.

Speaker 3:

Thank you, I always say the same thing, but it's take care of your people, I mean. It always comes down to that, thank you.

Speaker 5:

Maintain a learner's mindset right. Be willing to learn, hear new ideas, adapt, change your position and perspective, learn from your team and your network. That will help you grow and be able to adapt in your role and bring your best self to your role, but it also will set the tone for your organization and help your culture and your organization adapt and grow and continuously improve. So that's my two cents. Thank you, Molly.

Speaker 2:

The piece of advice that I always try to give folks is be authentic, be yourself. That's what your folks need, that's what your leadership needs, that's what your peers need. So just be authentic, be true to yourself, believe you, me, you will be successful. So with that, I want to thank these four wonderful ladies.

Speaker 1:

Thank you for tuning in to HSDF the Podcast. You can follow this podcast on any major podcast platform. Visit hsdforg to learn more about the Homeland Security and Defense Forum.