GSA pilots AI tools

We are going to be building the playbook. We're already happening. We're sharing this across the government. We just changed our logging of data, which is going to save us another $4 million. So, this is low-hanging fruit. We're trying to shine the spotlight around and just trying to find these quick wins. Innovation last six months has been incredible. Zach, I'm not going to steal your thunder, but the GSA AI bot is really cool. It looks like ChatGPT, Gemini, Claude. The innovation isn't this UI. The innovation was the thoughtfulness of saying GSA should not be building the model itself. We should be Switzerland here. OpenAI, Google, Claude, we'll take the best-in-class models. We will do what the GSA does best, build a layer on top of this around security, compliance, auditability, and by doing that it is now compliant to work at the federal level. And now we have agencies. It's been overwhelming. It's been a week now and people have heard about this and are asking how do we get access to this? So well done. We'll get more of a demo to come at our Friday engineering demos. There's a lot of builders in this audience online. So thank you. I've had an opportunity to meet about 2,000 of you or engage with you last Friday. As we think about the slimming down phase in this efficiency play, there's a slimming down of spend but there's also a look around the corner. There's a build back phase. Build back is around modernization efforts. This is how we bring new tools in to allow us to do more with less. So these demos are a great way, super informal, super optional, but if you have ideas, please demo it because it's a great thing for other people in this room to think about what's possible on the business side. You're like, "Hey, I may have that idea. I can apply it over here." Cody bot is a good example of that. And then RPA, this is a great example of the power of the GSA to build a playbook and export it throughout the government. We stood up an RPA practice maybe a couple years ago and we were the first ones. Fast forward, we built a community of 1,600 members building RPA bots throughout the government. So that's incredible. So, now taking all this information, all the learnings, we're thinking about the path forward and I think you get a sense where we're going. But let me just enumerate. We'll double click on each one of these. Welcome Josh, Mike, David and Thomas talk about the IT and software side, but just the quick bullet points here. Number one, right-sizing the portfolio. We have to get rid of those buildings where we'll never get enough money to fix them up. Those maintenance liabilities are increasing every single year. It doesn't benefit us. It doesn't benefit the community. We can do better there. Occupancy return to office is a thing. It's happening, team. Our goal is to get north of 80% occupancy throughout our portfolio. My commitment is I know this is disruptive for people who've kind of moved out of the cities. They have to commute in. I get it. It is a mandate. We'll do the best we can to make this an enjoyable experience to work in across the country and as convenient as possible to get into the office. And then I also just want to highlight here as part of these reductions, I'll acknowledge that the PBS team, you've been hit and I know I've been getting emails from your teams. My commitment is I want to hear from you. So if there are any issues, me and Mike Peters, email us. You've already been doing that which is great. I can't fix things if I don't know about them. Please email us. We will get on it immediately. Also one of the goals is to get in the field. We've already started doing this with some regional offices. We want to go out and meet you in the respective regions. So that's a commitment we'll be doing. We'll come back with a timeline to go to that. But I'm excited to meet in person. Centralized procurement again, one buyer on behalf of the government. You see the benefit of that. That is happening. I'm hoping to have some really good news in the next day or so that we should be coming out. But there's a lot of eyes from the top down on this. So again, GSA is in the spotlight. We're being elevated and being asked to contribute towards this mission. Compliance burdens procurement. I wanted to show this. This is a huge effort. And just give you a sense when you think of things like FAR, I'm not sure I didn't appreciate it. This is FAR. This is 2,000 pages of regulations. This is what every CEO has got to be trained on. It's complex. It's conflicting at times. And so we have a huge opportunity with the alignment of OFP, DoD, NASA to make a meaningful change to this. And why do we want to do this? We want to make it easier to bring best-in-class products at the best prices, the best value into government. We have an opportunity to do this today. On the same side, you think about software. The software market is changing so rapidly. We want to make sure the government has access to the best tools, things like FedRAMP is a huge opportunity to make it easier for younger businesses that don't have a huge track record in government to be able to sell into government. So, I know Pete and the team is actively working on this. We had a demo last Friday. The team talked about this huge effort. And then on the IT software services side, I think it's just rationalizing it. So consolidating one system for each job, centralizing our data. This is like EDS if you guys know that, electronic data system close. Yes. As you think about the future AI, in order to actually think about using any of these tools at scale, we got to get our data in one place and make it accessible. We have to have the protocols and standards so these different systems can talk to one another. And the benefit of doing this in the future is you think about tools like OpenAI operator. Has anybody used operator by the way? Raise your hand. You guys know about it? Oh my god. Okay, check this out. It's like RPA++. Really cool. That's going to be possible in the future as we open up these systems and have common language like little JSON. There's a lot of things we can be doing here but we got to get the data in one place. The cybersecurity IT team here is like the envy of the government. Everyone's looking towards you as a model of how to do shared services and cybersecurity. But I can tell you, look around the corner, just like we're consolidating and centralizing procurement, there's an opportunity within IT side as well. We already have agencies reaching out and saying, "Hey, can we offload to the GSA?" Yes, we should be saying yes to that. So that is the road map ahead. Again, these are just words on a page. There's so much context and substance behind these. I will probably be encouraging lunch and learns with my respective leaders to dive into the details. But it's really exciting and I can't stress enough like the moment is now. We literally were built to handle this mandate from the government around slimming down, becoming more efficient. I get asked this a lot. How can people help? And I want to just enumerate a couple points. I'll just unpack these. We saw this on Friday. There's a lot of people have context and skills and they care. That's a really powerful combination of traits. Be a problem solver. It doesn't matter if you code, design, do spreadsheets, or policy. Everyone can solve problems. And think about championing innovation in your own right. That's the only way we bend the curve on spend here. Collaboration obviously. There's a return to office. GSA is multiple businesses in one, but I also say since we touch so many agencies, let's overcommunicate, let's share best practices. We started doing these memos. They're words on a page that have a huge impact. We're literally telling people how to identify software rationalization, procurement centralization, policy normalization. So there's a lot to do on collaboration. Customer excellence, that's something I think is a hallmark of the GSA. We think about customer outcomes whether it's other agencies, our vendors, all the stakeholders and we work backwards. So let's just continue thinking about the outcome we want to drive for our customers and we'll find the best solution for that. The last one is an ask of me to you. We have a lot of authority right now to think about EOs, policy changes, potentially statutes. If there are things that are holding you back from doing your mission, please let me know. We've written ones on FAR already, on procurement, on real estate occupancy. There's a bunch being written right now. But yeah, there are people who want to listen and I need you, the experts, to give me feedback and we can write these together. But yeah, that's my ask of you. And ultimately, just again I want to say, oh last thing, Steve, you said a lot of things here. How do I know you're actually accomplishing them? Well, I have this nice little scorecard now. And this is pretty incredible. This is thank you to the finance team, comms team. This is a whole interagency effort. This takes companies, public companies years to get a scoreboard like this. So, this is what's going to be available to all of you to keep me accountable from cost reductions to automations that we're doing to deregulation efforts to disposition of real estate. There's a lot. And this, by the way, draft top right. I know this is going to be sent out to the world, draft. These are not actual numbers to date. We have to finalize this. I'm just trying to give you a draft of the format that we're going to be sending out to you. But I think it's really good. It's good to be accountable and have specific metrics to drive towards. And with that, I want to end with where I started by just saying thank you. This is the first hopefully of many town halls like this. I want to make this much more interactive in the future. I want you to ask questions in a couple minutes, but I want to hear from you. My door is always open. Do not hesitate to reach out, grab me, and I look forward to meeting you all in person. So, thank you very much.

Put your hands down.

We owe it to you guys to make sure that you have the best tools. But we also owe it to make sure that you feel like you're empowered to try things out, to experiment. And so, that's where this is coming from. Now, just to look at it, we tried to make it look identical to most chatbots. We're trying not to reinvent the wheel. We want this to look and feel exactly like you've been trying it out with ChatGPT or Gemini or any other tools that we've allowed for in the familiarization use cases. There are some subtle differences mostly in that we've pulled some stuff back. So if you look on the left hand side you have your basic chat history. Towards the top here you can select different models. We'll get into that a little bit later. And then the basic interface. So you can ask it general questions. I know I'm preaching to the choir for many, but just to make sure that we're all level set, we wanted to explain just what this thing is. And the amount of control we have with this is truly remarkable. We are able to tailor it to control for how it's going to speak to you, how authoritative it feels, what it knows, and importantly, what it doesn't know. So as we can see here, it knows what it is generally and it knows that it can answer some general questions but you can ask it like, "Are you an expert in GSA domain?" and you can... Hopefully I have it. I have a fallback in case the internet dies. Oh no, we're good.

You like that?

Who likes that? Some people like that. Some people wish that. I don't care. I'm a Bills fan. So my place in this world is to be disappointed.

You want to do a demo with the audience?

Does somebody have a meeting? I'm showing a good use case like you're preparing a big meeting. You need to outline it. Big presentation. This won't do the meeting, but it will create a good outline for you to start filling it in. Is there a big meeting people are preparing for that you want to volunteer for this or an outline or something you've had to do yourself? You want to call in Saul at some point? Is anyone else? I know Saul has one. Saul, what was yours? What were you thinking? So, let's take it. Get ready. Write me the strategy and a tactical plan.

To rewrite FAR. Again, it's not going to do it, but this is like a first meeting and just get an outline. See how this is. We're doing it live. Simplification techniques, modernization. So, it's pretty good, right? Like this is go. Yeah, I need to engage OFP, DoD, NASA. Here's an outline of it and we can start filling it in. But this is like the smartest tutor you ever have that's now available 24 hours a day, 7 days a week. You may have started using this at home personally. You're already using it in your inbox and email. Please just try using these tools. Just like the movement from the typewriter to the personal computer, the velocity this will be changing and improving is just so rapid. So we'll be showcasing examples, use cases. We'll maybe talk about the road map as well going forward. But this is a very powerful tool if you want to use it to gain a lot of leverage. Okay, great. It's bigger. Eventually you can export to a Word doc. It'll do videos and pictures and audio. But maybe talk about the road map now, like look around the corner. RAG, the FAR bot. I know there's tons of use cases around that.

For example, OGP is literally all these policies we're sending and reviewing. That's all being done by hand today. Thousands of policies are being read through manually. We can upload those documents, all of our policies into this model and just search across. So just like we're doing, we're searching across a corpus of data that we can upload. So that can be done across FAR policies. Department of Education had a great use case. They have support people in a contact center that have access to their own little bot that are looking up against an FAQ. And so their agents have access to this, but their customers do not have access to it. So they're saying, "Hey, Zach, can you actually take this at our FAQs and put it on the website so you can avoid the call completely?" So these are things that will be available and eventually voice mode, multimodality. Anyway, there's a lot to be done. We might be at time but do we do a black belt like do you have a code example?

It's all done. So this is incredible. Just see where it's going. A prompt generates code, generates an actual web page. This is not going to be eliminating jobs. This is literally going to say no developer wants to be doing this work. They want to do higher level work. So that's going to be these new tools and be available to developers and common citizens, citizen developers. And this is a very powerful set that you should be thinking about how to use it for your respective functions. A lot more trainings, a lot more lunch and learns. We're open to feedback. There are use cases. Please let us know. But there's a huge opportunity here.