HSDF THE PODCAST

Part 2 of 2 - Procurement Planning for IT Modernization and Emerging Technology

• 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 two part series, Patrick Walsh, Chief of Contracting Office at Immigration and Customs Enforcement, Nick Martinelli, Director of the Information Technology Contracting Division at Customs and Border Protection, and Bill Weinberg, Assistant Administrator for Contracting and Procurement at Transportation Security Administration, discuss the speed of acquisition around emerging technologies and efforts to support Artificial Intelligence initiatives.

Featuring:

  • Patrick Walsh, Chief of Contracting Office, Information Technology Division, Immigration and Customs Enforcement
  • Nick Martinelli, Director of the Information Technology Contracting Division, Office of Acquisition, Customs and Border Protection
  • Bill Weinberg, Assistant Administrator for Contracting and Procurement & Head of Contracting Activity, Transportation Security Administration 
  • Jonathan Alboum, Federal Chief Technology Officer, ServiceNow (moderator)

This discussion took place at the HSDF Technology Innovation in Government Symposium on February 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.

Jonathan Alboum:

Patrick, I don't know if there's anything you want to add there, but I do want to get your take on things you're doing to make procurement more effective and efficient. Go faster. I think everybody here would say these things take too long. Some of it's the back and forth, perhaps, but the government wants to get it right. How do we speed it up?

Patrick Walsh:

Yeah, I do want to put a disclaimer out there, because I know that is the general consensus that government takes too long to buy stuff. When we're doing an O&M services contract which I'm sure a lot of contractors in this room perform so long as we give ourselves enough runway, there's no sense in getting it done way in advance. We need adequate transition time. There are holdups with things like security processing, so we need to make sure we get our people on board with adequate transition time. However, I don't need to go out and procure a services O&M contract for two years from now. There's no sense in really doing that two years ahead of time.

Patrick Walsh:

Also, let's be honest, this system wasn't exactly created 100% for speed. There's a lot of legislation behind the requirements that me and my contracting officers have to checkmark before we are able to award something Whether or not speed is more important than those other interests. The legislation has decided that those other interests trump speed a lot of the time. Things like protestability vendor to vendor contracting a protest is not something that's possible. However, when I'm awarding a contract, I have to ensure that this isn't something that, as much as I can, isn't going to be protested. So there's a lot of other considerations.

Jonathan Alboum:

Operationally in your office, though you shared with me some of the things that you're doing just to run a better office, absolutely, and make those interactions better. Why don't you share that with everyone?

Patrick Walsh:

Yeah. So within the parameters of that space, we do try and go as fast as possible. So I mentioned Swift upfront. When we created Swift as an IDIQ, we saw it as an opportunity to really speed up that timeline to a word. We've gone from six months when we had to post out on a strategic sourcing vehicle to one to two months. I mentioned the seven to eight task orders upfront that we got done in one month.

Patrick Walsh:

This is complicated stuff Hyper automation platform as a service, various different things that we were able to evaluate and get on contract very quickly because we did the framework upfront, we did the heavy lifting, we did the FAR 15, and exposed ourselves to potential protests. But because we were able to get through it now we're able to do things a lot quicker. We can waive acquisition plans, for example, that otherwise would have gone up to, potentially to DHS. I would have to go for approval for some of these task orders, but we don't need to create that extra paperwork. Some of the things are what are called pill techniques that we've been doing for a long time. We do oral presentations, we do two-phase down selects, so during our second phase or second look, we're typically looking at three or four offerors rather than everybody who was interested. They're advisory, so people can go through the extra legwork of doing a full proposal if they want to, but that does narrow the playing field.

Patrick Walsh:

And then building in flexibility post-award. So one thing I mentioned up front was one of our contracts that had to pivot, the ERO-ONM contract that we had to do in NTA micro-rep. That hadn't been envisioned at the time of award. So building in that flexibility when you know you need to. Not everything can be a performance-based contract. You don't know up front necessarily. I want to build a rocket, for example, which is a very tangible thing that you could put as a performance-based contract. A lot of times we don't know exactly where this thing's going to land three, four, eight months down the road. So building in flexibility by having clans that are created to have either a personnel focus or a focus on surge. So we use surge clans for TNM and we've also done not only options at the end of the year but options within the year. So say, hsi comes upon some money and they had this idea up front, then they're able to exercise that plan like that and get surge work done a lot more efficiently.

Jonathan Alboum:

Flexibilities. We want to all embrace that. So, bill, I want to jump back to you for a second Part of this panel is about emerging technology. You have the most experience of anybody up here buying technologies. You've seen a lot of emerging technologies over your time. Today maybe it's artificial intelligence, but before it was cloud. Before that it might have been something else. How have you gone about advising your CIOs and IT leaders, program the folks with the requirements on how to embrace and use these things that we're hearing about, but maybe we don't know how to purchase yet?

Bill Weinberg:

What's interesting about that is that emerging technology, embracing emerging technology, really involves taking on risk, and some CIOs organizations are risk averse, very risk averse, and some recognize the value of taking on risk to meet the mission. When you work in an organization like DHS, which is very threat oriented, mission centric, you have to be able to take on risk because the adversaries that you're dealing with have no risk parameters at all. You know they have no. You know I've often said that if we could get the enemy you know, the threat actors to go through the same type of budget and approval process that we go through, we might be more on a level playing field. But by the time we get out of our own way in the government, they're on to the third generation of whatever it is they're developing. So we really need to be more agile and so being aware of what's available and being able to, through the acquisition and I won't even just say procurement process, but through the acquisition process to be flexible enough to either pilot and or bring on new technologies quick enough to make them matter.

Bill Weinberg:

People often criticize the acquisition, the length of the acquisition process, and procurement's really a fairly small part of that, because by the time. You know people visit me all the time and say you know, we've got this great product. We think you could use it, this particular division or whatever To turn that into a requirement, get it funded. That takes a long time. We can go buy it in a couple months probably, but getting it into the acquisition framework and you know, and DHS some years ago evolved their own DOD 5000 structure to. I wouldn't say they did it to enhance procurement or acquisition. It's more like the DOD structure and we're fighting threats every day.

Bill Weinberg:

I mean we need to have a construct that allows us to say I saw, I was at a vendor and our administrator will do that. I was up at someplace in Boston or Silicon Valley and I saw something we need. We need to buy that right away. How do I? Can't wait three years to get that. I can't wait six months to get that. I want to buy that right now and start fielding it. And we have the tools to be able to do that if they can get the money. Money sometimes is one of our biggest obstacles, but understanding what's out there, understanding what you're trying to do and working with the procurement people, you know, as has been said on the panel also we can. We're sort of I always tell people, we're kind of like the Make-A-Wish Foundation You've got something you need, bring a definition of it, bring the money and we'll make it happen for you.

Jonathan Alboum:

So let's apply that to artificial intelligence. I'm sure everybody here is wondering well, how are these tools going to— especially generative AI tools make it into agency missions, into the way agencies operate? The executive order on artificial intelligence is, I think, a great start, but there's a lot of unanswered questions in there and, as providers of technology, we're working to answer those questions with government. So what's that conversation like in your offices today? Are you getting requests? I want to use ChatGPT. I want to use this GenAI tool.

Bill Weinberg:

I'd say we're probably at the beginning of what you're hearing. We're using some robotics RPA to automate some very manual, less thought-involved processes today very manual, less thought-involved processes. Today, ai may be being used out there in the Intel community to try to analyze data better. I think there's more risk attached to it today than there is. I think we're more looking at how to counter it as it tries to subvert our effectiveness as threat responders. So we're at the beginning stage.

Bill Weinberg:

I think later you're going to hear from Eric Heisen, who's our new AI official, and I'd be interested to hear that also, but I think we're starting slow. We have to determine what AI can do for us and we have to make sure it's secure. I mean, I think we're in the middle of embracing and implementing cybersecurity everywhere to protect the data. You know a lot of the DHS components create and store a lot of data. We have to protect that number one and we don't want to introduce things that we don't fully know what they're going to do. So I think that's a little bit of a risk base.

Jonathan Alboum:

So how about a CBP Nick? Any clear use cases? Or you know, what are you hearing from OIT?

Nick Martinelli:

Well, I think that the you know Bill's tease for Mr Heisen was very good In the procurement shop. We have had conversations about AI, but it's very much in its national stage in terms of its use. We use RPA and I think that with a know, we're with a lot that we do in the mission that we have. We have our you know, our heads down meeting the need of the day and we're relying on you know, our partners in OIT to assist us with the AI piece, but it's so soon for us in terms of that type of using it within our office. This is nothing against. This is part of how the government works. But we just moved to electronic contract files a couple of years ago, so that leap, will it come? Yes, but we're not there right now.

Jonathan Alboum:

Can you envision and, patrick, I'd love to get your perspective on this too you envision an opportunity to use generative AI capabilities to make contracting more efficient? Could you summarize proposals? Could you generate RFPs? Are there ways to? It's a laborious process at times, right? Could you summarize a proposal? Rank or something? Are those things that are conceivable in your office? What do you think, patrick?

Patrick Walsh:

I'm going to take that, Nick. I wasn't sure who that was directed towards.

Jonathan Alboum:

I'd love to get Bill was directed towards you start. I'd love to get Bill's perspective on it too.

Patrick Walsh:

I think there's huge potential. There's the short term and there's the long term. Short term thankfully, dhs gave us permission recently on government computers to use generative AI, a few different large language models, I think. The short term we're using that for market research. We're using that for all. Tell me how you use it for market research. Okay, yeah, and I do want to say we can only use that for non-public data. So all the data that Bill's referring to off limits.

Patrick Walsh:

But as far as like using it for market research, I had a zero trust requirement that I was in. This was actually pre authorization, so I was doing it on my personal computer. But so how do we assess a sole source justification if we haven't adequately done our market research? Chatgpt, who are the leaders in this zero trust category? Because I had been told that there was really only one player in the game. I consulted the Gartner Magic Quadrant and that gave some information, but ChatGPT kind of got me started down the path of hey, here are the other players in the game for this.

Patrick Walsh:

So long term, yeah, absolutely, there's going to be. I think the sky's the limit if we have, if we kind of have our handcuffs off. So right now we're handcuffed to using only certain kinds of data. But if law, I mean we're law enforcement. So if law enforcement officers could go in there into a Chad GPT and get CBP, get TSA data and ICE data just with a simple prompt, I think it's doable. It's a question of whether or not we're comfortable enough with these language models, sort of what you're describing are domain-specific large language models right.

Jonathan Alboum:

They're smaller, they're focused on a particular kind of data, they're more curated, they have a you know an intention. It's not all the data on the internet. So I mean those smaller large language models are. I think we're going to see a lot of those come from industry into. That's something I see coming.

Bill Weinberg:

Well, and the reality is, when we start seeing proposals generated using AI, why not evaluate them using AI?

Patrick Walsh:

That doesn't scare me. I mean, there's inherently governmental functions, so inherently, the contracting officer is going to be the one who's obligating money, and that's why I'm so happy that we're up here. We're the gatekeepers. So whether or not Chad GPT churns out a brilliant interpretation of these proposals and churns it into something that's passably good English, it's gonna be the contracting officer that puts pen to ink, so keeping that human in the loop. Also, I think a lot of the fear is that in a lot of really smart people are afraid of bias in large language models. They know a lot more about it than me and I respect their authorities on this. There is the human element of bias as well in reviewing a proposal. So you can't eliminate bias Just. Everybody has preconceived notions that go into your interpretation of a proposal, and the fact that that's part of the calculus in bringing on board these large language models, I think is only going to make government.

Jonathan Alboum:

I appreciate that we're talking about it right. You know that there's guidance from DHS and an approach to begin a conversation because it's coming right. You know it's going to be impossible to avoid, it'll just take longer and it won't turn out as well if we're not engaging in this kind of dialogue today. So that's a very positive sign. Nick, I don't know if there's anything you want to add. We want audience questions too, nick, why don't you share?

Nick Martinelli:

No, I think that I like having the opportunity to work with my team and hear from them, and they all bring unique perspectives to the jobs that they have. I rely on them. They do a great job and I want them to continue doing a great job. So I might be the opposite, but I don't want them being removed from the process. I don't want them being removed from the process.

Nick Martinelli:

I think that they're bringing business acumen to the process and I'm not aware of a chat, gpt or an AI that's going to bring all that different perspective and experience to the process, where that is going to be able to make a fair and reasonable determination, which is the soul of our job and be able to do the things that they're able to you know that sort of unique experience and to the you know to the process. So you know, I like the job that they're each doing. I want them to continue doing that. You know I like the job that they're each doing. I want them to continue doing that. You know what they're doing and I think they hold a unique place in the process and you know I kind of shudder that we would use something other than you know, a government official to rank a proposal or understand what each you know contractor's unique technical approach is, and that goes for our evaluators as well.

Jonathan Alboum:

It'll be interesting to see how this falls. We have about five minutes left. Do we have any questions from the audience?

Bill Weinberg:

Good, no questions, so good.

Jonathan Alboum:

All right, we have minutes left. That's it. Just share your name and where you're from.

Audience:

Yeah, good afternoon. My name is Joanne Newkirchen, I'm a federal employee with FEMA and I'm actually going to raise a question that's on the topic of IT modernization but you have not yet discussed today and that is related to the DHS CIO's strategic objective to support modernization in place, and I'm wondering whether that's a topic that you've engaged with on the contracting side and how you see that potentially changing operations and maintenance contracting going forward.

Bill Weinberg:

Well, I'm sure if Bobby McCain were here he would say no I bet you Bobby McCain was supposed to join us on the panel.

Bill Weinberg:

Modernization some of it's being forced, you know, by the data center closing down and everybody moving to the cloud, implementing cybersecurity, modernizing to use workflow tools that are emerging, you know. So we're certainly having those conversations. We have a relatively new CIO at TSA and he and his new team, as they're coming on board, are taking some new viewpoints to how we ought to be doing IT and governance and modernization at TSA. So over the next couple of years, I think you're going to see a transformation to different ways of doing things, a little bit more enterprise approach to IT services, not having siloed platforms.