The AI Gold Rush: Models Are Shovels, Data Is the Gold | RAISE Summit 2026

0:04 Nice to see everybody. My as noted my name is Akquil Ahuja. I run the enterprise tech banking team at Bank of America and I am incredibly excited to moderate this panel today. I thought we would start by having our esteemed panelists here introduce themselves and maybe just a line or two about what your company does. Gareth. >> thank you. My name is Gareth Davis, chief product officer at Octa and North Zero. And we are an identity security company. We help the world's largest organizations and developers both secure their human workforce and increasingly their agent workforce and consumerf facing agents.

0:42 Alex Wizori, founder and CEO of DDN. we deliver business and financial outcomes through AI. DDN is to data what Nvidia is to compute. We are deployed across hundreds of the largest and most successful NCPs, AI natives, sovereign implementations and enterprises worldwide. And it's all about data as you will hear from the panel and from our friend here.

1:13 >> Mark Surman and I'm the president of Misilla. You may know us from Firefox, but what we're in the middle of trying to do now is do for AI, what we did for the web, which is democratize it, make it more flexible, and certainly build more competition into the marketplace. And I the biggest thing to know about here on this panel is we've spun out a company, we spun out a number of new companies. One called the Misilla Data Collective, which is a basically a fair data marketplace for people. Let's say you have a use case which is fine-tuning and maybe you have data that's very specific in a specific language. We match those people up and I would say you know what we would argue is it's all about data but it's also all about open source.

1:54 >> Yeah, my name is Ditri, founder and CEO at Emma. Emma is a cloud control plane that sits on top of a different cloud infrastructures and we help enterprises to streamline the way they interact with different providers, how they deploy the environments across the globe and how they move the data across. >> Hello, I'm Will, CEO of Exa. So we're a search engine for AI agents. So a lot of you have AI agents that could probably use this and yeah we serve thousands of companies with like super high quality information from the web and also now data beyond the web.

2:24 >> Perfect. Thank you everybody. And I think one of the unique aspects of this panel is we have so many different viewpoints from the data platform to search to security to cloud management. Excited to explore all aspects of this with our with our panelists here. I thought where we could start is I think the title of this panel is is is in particular interesting. AI gold rush models are shovels. Data is the gold. So maybe where we start and will I'll ask you to kick it off. Do we all agree that models become a base layer and the predominance of value is in the data itself?

3:00 >> Yes. I mean I definitely agree that. I'm a little biased but I do think this is true. I would say like the interesting thing is that as models get smarter they cross a level of intelligence where you just don't need that much smarter for most for most knowledge work right like open source models are good enough for most knowledge work right and there's not that much differentiation so then the differentiation between a lot of applications becomes the data they have access to and that's very exciting world I think because you know it means that we don't have a few companies dominate and really becomes more decentralized like because everyone has different data you know all your companies have access to data that you know bigger companies might not have access And so you could build differentiated products by taking the commoditized intelligence and combining it with your differentiated data. So that's very exciting.

3:40 Demetri. >> Yeah. I mean yes and no. And actually when I was thinking about this question, I I found an interesting fact that the guy who who actually was one of the the sellers of a shows in California and he became the the the richest man of California back in the days, he died broke. that actually illustrates Yeah, I mean that's that's true. Mr. Brandon he he he lost everything but he at a certain point of time he was the the the first millionaire in California. the thing is that all the money they actually moved from let's say the show wells to actually the transportation companies and to the banks. So yeah to a certain extent data is the new gold and on the other hand there are also the other layers where we we we actually need to build the right let's say regulations and we also need to build the right technologies to support the data flows from one point to another point from one human to another human from one one agent to another so yes and no >> I I love the I love that story and I mean I often wonder whether anthropic or open AI you know are they the next Google or are they the novel and you know at this moment they may be the people who you know feel like they control the the puck because it's it's the models but the models are also commoditizing and where where that plays out is going to be interesting to to see. The the thing I would just ask quickly is how many people in their including people on the panel in their companies or in companies you invest in like build on top of the Linux ecosystem. Just a little poll like it's clearly a group of liars. because you know that almost 100% of companies build on the Linux ecosystem. And at the heart of it the Linux kernel is really just a commodity. It's a commodity that everybody contributes to or many people contribute to. but it's a trusted infrastructure that we all rely on and then value is created on top of it and models you know especially as we have better and better openweight models start to become that kind of shared infrastructure and commodity that then the value you create on top of them is the opportunity and I would say data is one piece of that. So I can fine-tune something I can use the data in lots of ways to add value to that model. It's why we set up this data company. I would also say that lots of other pieces whether that's orchestration evaluation or just integration making it easier to use that commodifying stack is another key piece to be looking at.

6:20 >> So I think when you look at AI and the value creation and the opportunities that AI can deliver to enterprises across all sectors to nations. I mean we happen to be in France today. the geopolitical landscape is complex and sovereign AI is something that is extremely important. And so you look at in the context of models what does sovereign look like? can you have models that traverse countries and still provide the air gapping and the safety and the security and the integrity with respect to data that is required from nations to operate AI natives and the value that they're delivering so the way we look at it the way DDN looks at it is the adoption of AI and accelerating the adoption of AI requires value creation that value creation for enterprises is business outcomes and financial outcomes. In order to do so, you need to rely on the model part of it, whether it's closed or open. There's a pre-training phase, there's a post-training phase. The post-training phase is happening at the edge. We will increasingly see distributed disagregated models that live at the edge globally worldwide. I think we'll continue to see foundation models that are very powerful and require very significant capex investments. But I believe that all the value creation for whether it's enterprises or consumers or nations relies in transforming data which is anywhere into insight and that insight results in better products better services better security for nations for citizens of nations and the problem that DDN has set out to solve and I think we've done it successfully from Nvidia who is standardized on DDN to XAI which is running Grock on DDN is ensuring that you deliver the business and financial outcomes with the safety and security that is required by all types of organizations. So I don't think it's really a matter of shovels and gold.

8:41 It's really how do you deliver value to enterprises and nations and consumers through AI without breaking the bank. So there's an ROI aspect to it and there's a security aspect to it and those organizations who are delivering those services and capabilities are the ones that will be successful for the long haul. I think Alex said extremely well the word was outcomes. So undoubtedly today the models as as we see commoditization and acceleration of model development the constraint that we're seeing is the bottleneck about access to data. So if I think of the model as essentially the brain and the data is the memory or the context the intelligence that the brain is imbued with. next we need a nervous system.

9:26 We need control because simply giving AI access to more data whilst in theory it it reduces the upperbound limit we have today on the intelligence and ultimately therefore the business outcome that AI can deliver. It also brings tremendous risk and we see that every day at octa as enterprises and developers innovate in AI and build agents. We're finding that identity security is the foundational layer for agentic security. You need to provide both access to data for these agents but you also need to do it whilst mitigating the risk and the blast radius. And ultimately if we can if we can solve that seeming paradox then we can provide just the right access for the intent and the task of the agent in order to ultimately drive a business outcome. And I think given where we are in the maturity cycle of AI it's very easy to talk about compute and infrastructure and technology but remember all of this is ultimately a means to an end. And an end is either driving workforce productivity or you know a retailer for example delivering a frictionless experience and meeting the consumer where they are as they increasingly search and discover for products through ChatGBT or Gemini. So if we can connect both the the model to the data and to a measurable business outcome whilst mitigating the re the risk that that's where I'm most excited about the explosion of innovation and growth and ROI we can drive from AI.

10:51 Garrett, that's a perfect setup to where I'd like to take the conversation and double click on a theme that a lot of you have mentioned in different ways. You know, if we look at it today, we're we're at an incredibly unique point in time where the public at least has seen a lot of value in the buildout of AI, whether that's silicon, whether that's infrastructure beginnings beginning to think about the data impact. But as you think about this AI ecosystem between models, between infrastructure, data, application, search, security, cloud services, how do you see this value evolving over the next couple of years?

11:27 >> So I think the first caveat is that you have to get a strong foundational base in place and we're seeing today a recent report came out 48% of enterprises deploying AI in production are not securing these agents. And so we see developers hard- coding API keys and providing overprivileged static credentials to agents that create tremendous risk for bad actors. And perhaps if you're doing this in a totally sandboxed environment, you you can kind of mitigate some of that. But the minute you build AI in production, which is where the rubber meets the road, and you deliver business value, you need to ensure that that foundation is secure. And we're not in a world of automation. We got to remember that agents are non-deterministic actors.

12:10 And so I think what we're seeing emerge a is the need for identity and security to be at the foundation of all agentic development. Whether you're building a developers building on lang chain, you're building on bedrock or vertex, you're turning on agents in thirdparty software that you deploy within your enterprise, you have to ensure that those agents are tightly scoped. They can only access the critical information necessary for the task at hand. But again, how do you do that without also limiting the potential power of the agent? And that's where I think we'll see an evolution from today fine grain authorization where you can effectively define very granular access policy for an agent in order to solve the problem and do so safely to a world where we have more of a a semantic or an intent runtime layer. And that means you have to understand the intent of the user's prompt. You have to understand the behavior of the agent at runtime. How and of course you can't see the reasoning. So that may be what tool calls are they making to internal or third party MCP servers. How are they trying to access internal resources? And then ultimately you need to be able to see how that translates into an action at the end of that reasoning chain. And if we can help both govern that layer as well as give developers and businesses the tools to to better govern it, I think we'll see an evolution where we build more secure systems from the ground up, but we also have much more performance systems and we see the business ROI come in.

13:40 >> So I think three simple things which are easy to state and difficult to implement. I think the reason why DDN is being adopted as the data engine so many organizations in so many areas is these three things number one the technology the capability that you're deploying and delivering to your customers has to integrate into their environments easily if you're an enterprise well you have a set of capabilities you have an infrastructure the CIO of that enterprise eyes is worried about yes AI has lots of promise lots of potential but is it going to be disruptive to what I'm doing I have to keep the lights on my CEO is telling me keep the lights on I see the promise of AI is it going to integrate seamlessly same thing if you're a new cloud provider and you're delivering services to your customers is the capability that is being brought that is enabling data to operate at better tokens per watt or better tokens per unit of time does that integrate seamlessly into my data centers if you're an AI native if you're a sovereign so ease of integration into an existing infrastructure is the first one the second one is the economics if you're delivering a capability that revolves around data and data I think is shifting in terms of the value that is being delivered to enterprises and nations and consumers from the model world pre-postraining into inference and agents. Well, where a chatbot makes one call, an agent makes 30 calls. 30 calls is 30 times the power, 30 times the compute, 30 times the data, 30 times the cost. So you need an implementation on the data layer side on the data engine side that ensures that the economics are working out so that these AI investments that organizations are making are actually penciling out from an ROI standpoint. So that's the second one the economics of it. The third one is security and security is measured differently. if you're an NCP it's multi-tenency. If you're sovereign it's air gapping. if you're an enterprise is what needs to stay inside the enterprise and what can move out of the enterprise.

16:16 The problem gets a little bit trickier when you look at agents because in order for agents to operate successfully, they have to move out of the organization and back into the organization. If the gatekeeper is a human, relatively easy to do. But we're very quickly moving into a world where agents are interacting with other agents that are on the other side of the air gap. So how do you manage that? But basically solving these three problems, ease of integration into customers environments, economics penciling out and the security attributes. Solving that delivers the business outcomes and the financial outcomes that enterprises, NCPs, agentic AI and so on and so forth require. So we solve that problem from the data engine side. I think everybody else who lives in the ecosystem of AI needs to look at the same things and that's what will accelerate the adoption of AI and the value creation of AI for the world's economy.

17:21 So much to build on there. Maybe take it a slightly different direction. So the question is like where is this going? And I think it to me it is very clear I think it's becoming widely accepted that that one of the places this is going is to extract value from AI as for your business to differentiate and give you an advantage and also to provide value for consumers for consumers to get value. It's going to be about things getting much more specific, specific to the business proposition you have, specific to the problem you're going to solve. And the the kind of big AI labs and also, you know, whether they're open or closed models are are not specific.

18:02 They're not specific by design. They're general purpose. they're not trained on AI from or data from anywhere. They're trained on just the data that happens to be everywhere. And so that kind of era of just scraping the web and tra training general purpose models while giving us a lot of capabilities isn't going to get us to where we need to be or where people are trying to go in five years. People want specific stuff and so how do you get that? How do you get that as a you know the Bank of America or an auto manufacturer or a government? I there's three things I would be watching or that I think are going to grow in the next five years.

18:40 One is obviously like using the data that is yours that that belongs to your company that belongs to your government. It's a a big piece of sovereignty and building on that. And so that's where you know doing your own custom post training doing fine-tuning using rag like finding ways to take what you have and make more specific implementations more specific models. You're going to see more and more of that. You're already seeing more and more of that.

19:06 And and then I think the other thing is going to be people who have got specific data that is hard to get to, you know, may be willing to share it with you, but they want to share with it you on terms that are, you know, are fair to them. So whether that's very arcane languages or people who have got language data sets that involve things like sign language or, you know, code switching inside of of languages, you may want to access customers through that. And this is somewhere where Misilla Data Collective has specialized out of the gate through those more specific pieces of language. It's hard to synthesize that and there are people out there who want to offer that to you. And so I think you're going to see new approaches to kind of data relationship negotiation which is what our data collective is about. that that are going to unlock that value for you and unlock value for the people who own those data sets. And those are not the data sets that are everywhere. And then the third thing is collaboration and building real value that is specific between people who may not even be trusted parties but but even if they are you know they can't share the data between them and pool it. So you know it's been a long time we've talked about privacy enhanced technologies things like differential privacy and federated learning and you will see those in the next 5 years come in to their kind of maturity. I think especially federated learning where I can you know learn from this and learn from that and learn from that without having to disclose the the base data sets. you'll see those technologies really start to grow in in their use as people are trying to find specific ways to deploy AI for their customers and their their business.

20:52 >> Whoa. When when you're like a force in a row with smart guys, you're like, "Okay, what should I say?" So, look. Yeah. I mean, so first of all, of course, >> something about shovels and the gold rush >> and Mr. Bren. so yeah first of all I mean the value doesn't sit with the models right whether you and and and the open source models they they prove that these this this approach gets cheap and cheaper so we think that the value sets of course with the data right and the what I mean so actually with with the data that you can legally use right so where you have the right to use the data or your agents they have the right to use the data And then there is a second layer. So how this data is governed. So basically where do you host this data? How you get access to this data? How you move this data and how how many frictions you actually do to to get this this data or to make this data accessible for your agents, customers etc. So there are like different layers because early stage companies they see it differently, they do it differently. However, at the enterprise layer or the government layer. So these are the the the different things and they actually the the the cost of a failure here is significantly higher. So again the data and the governance and of course the security like those three pillars I do believe will will stay with us for for at least the next five years and maybe beyond. So will over to you now it's five people.

22:31 >> I'm the fifth. Okay. I'll try to be different. Okay. yeah, then the question is like where does value acrue you know in the the chips or the security or the search or really everything and I think the happy answer is that there's value everywhere, right? I mean this is a trillion dollar multi- trillion dollar economy that's being created and so it's hard when there's trillions of dollars to not have some piece of that that's worth billions of dollars, right? So so there's values value everywhere. and you know, I've been doing this for 5 years at Exa and I've seen like companies years ago that I thought were so silly like that's such a tiny thing like oh compute for a certain type of workflow like how could that be a big company and now it is a big company or you know I there like dozens of companies like that I was like that can't be big and the answer is I know everything can be big which is awesome. that doesn't mean that everything will be equally big right and so again like I don't I don't like a world where things are centralized where where power is centralized to a few companies I would few companies I would love to see you know a polar proliferation of tons of companies that all are very valuable. And I think what's interesting is like okay, so where does the value acrue I think we're we're now entering this world like it's super hyper competitive. I mean I'm not young I'm not old enough to know like what things were like 30 years ago but I imagine building a company now is extremely like it's extremely competitive because like even if you you know gather some of the smartest people in the world into your company well now intelligence is being automated. Okay, that doesn't matter. So now the things that matter are very weird. It's like physical resources.

23:52 there's like data, right? Like like the things that matter to to make your company not a commodity are increasingly those things. just the the the ferocity with which the company moves is the most important. I I'd say like like watching a lot of these companies in Silicon Valley succeed. It's always the ones that just have this intensity. they're working non-stop obsessed with the company. They can't just be like chilling on weekends. so that matters too. but yeah, I mean I I see a lot of forces for breaking up a lot of the the the centralization be because like even these labs like you know what is the network effect they have arguably it's not even clear right I could see a world where you know they they don't have as much power but also you could also imagine worlds where because they're working so intensely compared to all of us or because of whatever regulation like they can't acrew power. So, I guess what I'm saying is it's unclear. There's a huge amount of value. I think companies that figure out the the ferocity thing, some sort of data network effect, can can somehow accumulate resources that other people don't have are the companies that going to win.

24:52 >> I love it. Direct from Will. No, no chilling on weekends, folks. So, >> we're in Paris. You have to chill on weekends. I mean, come on. >> This is this is a city of beauty and the city of love. And there's fashion week going on this week together with race. Diversity. >> Let it be noted from the best dressed man at this conference focused on fashion week. Yes. Dimmitri, I wanted to come to you. First off, I don't want you to be fourth on this in the row, but you know, we're talking I think there's a we're all in agreement data has tremendous value.

25:32 >> Yes. >> But where where are we today? Are we seeing enterprises actually unlock the value of their proprietary data? If not, why not? What are we waiting for? Are we seeing sovereigns or governments act any differently? What what are we actually seeing today in terms of utilizing that data? It's a great question. Thanks. Thanks a million. so I think enterprises they are not ready to be honest because there is still a lot of unstructured unlabeled data that exists and again there are a lot of especially here in Europe a lot of regulations who can get access to to what kind of a data right where this data can be stored etc etc and the data itself it's like you know I'll try to to create an analogy here it's like an uranium actually. Soanium without enrichment doesn't make a lot of sense right and you need actually reactors to make this enrichment happen. So the same applies to to the data. So you need to prepare this data enrich this data and then you you can start using this data but again where what kind of engines what kind of a reactors you get. So I think to your question no enterprises they're not ready to get to a point where they they ready we should all together I mean every one of us who who who who are active in the space we need to build up a lot of a lot of great things and scale our companies to help them to get to a point where at least there is a governance there is a certain security there is a certain assurance that enterprise they can safely interact with the data and allow let's say other organizations other agents let's say people to to to interact with this data get access to. So that's that's my point. I think >> I think I mean we we are we are seeing something somewhat different. So I will beg to differ. we are seeing enterprises adopt AI and benefit from AI but those who do so are the ones who are thinking carefully about the business outcomes that they're trying to achieve.

27:42 What is it that they're trying to do? Salesforce is using DDN as a data engine. It's increasing the productivity or their training and by doing so their investments in AI are penciling out. The ROI is fantastic. RO is using DDN in pharmaceuticals and genomics. It is helping them develop better drugs faster, bring them to market in a more economical way, save lives and reduce the risks associated with getting FDA approval and compliance. we have dozens of enterprises in financial services which is your industry who are using DDN technology for AI in order to ensure that compliance is working out and they are compliant across the globe same day instantly or are able to generate and create more complex models which deliver better returns to their investors. We're seeing nations were deployed in actual countries where the country is using AI as a framework and is tying into every department within the country from commerce to agriculture to defense to immigration in addition to having a super app that is delivering value to the citizens of that nation and making it happen. So it is happening. It is happening for those who are thoughtful about it are working with the right partners are involving the right technologies and are tracking that process carefully. I think those who are trying to knee-jerk react to well AI is happening what should I do how should I do it and are not thinking it through it's much more challenging but it is well underway and I would say over the last 18 to 24 months it has been accelerating at a pace which is unimaginable I mean the growth that we're seeing in enterprise adoption and actual monetization and productive monetization across all these industries from financial services to manufacturing to autonomous driving to retail is is phenomenal. It is happening. It truly is happening now. And if I may just to build on that very quickly, I think we see that at octa again both we have you know we launched on orzero the developer side of customer identity November of last year a GA product to help developers build and secure agents and we now have thousands of developers actively building from startups and AI model labs all the way up to large enterprises B2B software companies that that have a mandate frankly to innovate.

30:25 obviously there's real opportunity but also a credible risk of disintermediation as AI transforms the experience layer of traditional B2B software. We're moving to a headless world. You you've seen it with Salesforce and the way both consumers, business users and workforce employees are engaging and interacting with software is fundamentally changing. So I think what I would say is that back to that earlier stat of 48% of AI being unsecured. It is less a question for me of is real AI being built. I think last year was a year of sandboxes. We're seeing real production systems being deployed both within and across enterprises. But the the critical question is which of those agents are being built securely because in the rush to create value and to innovate also again the this this paradigm or this requirement for security becomes essential. So the most thoughtful customers we're working with are architecting and we we've actually built an enterprise blueprint to help articulate what is a reference architecture for the agentic enterprise.

31:30 what is a reference architecture for a customer agentic experience and I think that's where we're starting to see the maturity and we're still somewhat early on the maturity curve but you have again where where real capital flows comes real accountability to outcome otherwise we truly are in a bubble here I and I don't believe we are but that's the chasm every business every developer has to traverse and again if we can do that securely and again security isn't the only layer it's just a foundational component but that that is what enables access to data and ultimately to outcomes. And I think that's that's the exciting thing to see and perhaps evidence of of this conference is that shift from sandbox to production.

32:12 >> Mark, I'd love to get your view because I think you have a an interesting lens on this both from open source closed the data collective but also you're looking at the consumer side as well as this concept of enterprise. So sort of how are you seeing data being leveraged in in this evolution? Well, I I kind of want to differ a little bit here in that I think your original premise is actually a good one that there there isn't there's a lot of value in data that is not unlocked. There are a lot of people who are afraid to unlock that data. Yes, there's amazing stuff happening and people are doing smart things, but you know, the amount of data on the web is sort of like this much.

32:52 Amount of data in the world is this much. And maybe we're a little bit outside of that, you know, kind of public data in terms of how much is being used, but there's so much potential not unlocked yet. And one of the big reasons, whether that's from the consumer side or from, you know, people that we work with in the data collector who've got really unique data sets or inside the enterprise is trust. Like people like your data is like your memory or your context and like people are afraid to give it up. And you know, one of the interesting things, it's funny, I guess the French in the policy spectrum have this, you know, computer says no, concept. And Firefox is a little bit like that on AI in that one of the first AI features that we release is the ability to turn it all off, right? people have a real caution about AI because they're afraid of themselves which means their data being exposed and and you know kind of used to train something that they don't understand and might get used against them and so that is a reality out there and then from the Firefox consumer perspective what we're trying to do is figure out how do you build that balance with people how do you make the you know not the binary privacy promise but a promise of having choice and control choice and control over what models you have choicing control over what data is disclosed getting some value back from that and I think that's the same thing that businesses are looking for I mean one of the reasons I hear from businesses about an interest in open source or an interest in onrem I mean you're probably having these discussions in the bank is like I don't want to have my business's memory I don't want to have all that I know going to to train claude or open AAI or Gemini and do I trust trust that it's not. And so that's the the kind of business version of the same anxiety.

34:43 And so that's all about trust. And I do think a bunch of the things I talked about earlier are people seeking that trust. Whether that is using federated learning to collaborate around data and trusting that you're not stealing my essence or you know as somebody who has got a unique data set that there's somebody who can be their broker who's on their side or whether in my own business I can do this as onrem or airgapped or whatever that's all people seeking a place that they can trust to unlock this value and still own themselves.

35:17 >> I I just want to build on that. Okay. So well first of all I love the the visual analogy of like the size of the amount of data in the world. So the web is really like this in compar. So the web if you do the math it's actually like roughly in the exabyte range which is a happy coincidence that we're called exabyte. it's mostly video actually but so yeah xabyte range and then the world's the world's data I haven't done the math here but I would imagine it's in at least the zetabytes which is like a thousand x more so like huge and and there's an interesting thought experiment which is like as an example let's say all that data was available to be used you know freely obviously no one wants that their business data to be used but as as a thought experiment if all that those zetabytes of data were were available for agents those agents would perform way better right so you can imagine you're like a hedge fund AI agent and you're making a bet on where you know you're trying to make an a billion dollar investment if you have access to those databes you're going to make a much better investment right so there's a huge amount of value in this data but it's not currently unlocked now okay so that thought experiment is not real because we're not all going to give away our data but like I do think a lot about how can you you know can you create a market such that people with data that's very valuable can make money off of that data and so because there are agents that want to consume that data right and so we actually happened to launch a thing called exconnect recently that is a step towards this world where basically people can you know you know upload their data or or or have an API over their data and then get paid if any agent uses it which is pretty cool and I think this agent economy though very early and you know there's not that much revenue flowing through it yet will literally a thousandx over the next few years and then you can start making a huge amount of money from from your data. So that's pretty cool. So that's like one way to unlock that data.

36:53 But I I do get very excited about like can we get can we draw insights from from all the insane amount of data in the world. yeah, I'll just add one more thing. There's a totally separate point which is like the value of like data that a company has. I don't know at Exo like we're very much like we want to because in if you want your company to move as fast as possible which is what I constantly think about like you want agents to have full access to all the company's information. Now it's kind of hard because things are in Slack and notion and you know there's like you have all all sorts of data and all sorts of applications but I think if if you want your company to move fast you have to like create systems such that your agent your internal agents can access all of that. I think this actually benefits startups because they have less red tape. which is kind of nice.

37:31 but yeah. >> Well, folks, we were the six of us were sitting backstage wondering how we were going to take up 40 minutes. And it turns out we got through three of the questions we we had planned for, but I do want to thank these panelists for an absolutely fantastic discussion. obviously you guys are still around the conference. Hopefully folks can find you, but but thank you for sharing your views today. >> Just one one thought on that. I I think we talked about the topic of security and governance and sovereignty.

37:59 I think these are very very important topics. we've DDN has collaborated with some of our customers to develop airgapped secure solutions that operate at the level of nations today. So if anybody is interested in how to do it, how to secure for your organization the data, ask us and we will provide advice because we've done it. It works at the level of countries and obviously enterprises. >> Thank you Alex. Thanks everybody.

38:30 >> Thank you.