Transcript
Please welcome to the stage Chief Revenue Officer, John Jester. All right, welcome back to the VeeamON closing session. I'm delighted to have a little time with you to try to wrap today up. I got to run around, drop in the different breakout sessions that all of you were participating in. It was great to see the energy, the level of discussion, how everyone was really leaning into this topic. And what was interesting is, hey, at the beginning of VeeamON, we asked you to pick which track you wanted to go to. Were you going to the resiliency track? Were you going to the data security track? And while we think there's always going to be importance for expertise on both those domains, you are starting to see this converge. You heard that this morning in the keynote. You've seen the different slides we've put up. I'll put this one up. This is how we see this coming together. Resiliency, data and AI security. And that is why we're creating the Data and AI Trust Company for you. So expertise in both domains, but the platforms coming together, the architecture coming together, everything that you need to collaborate and to work together. But there were three takeaways that kind of I had as I went through the day, had discussions with many of you and listened to some of the sessions. The first is, hey, this convergence is real. Over and over, I had people that I've worked with on the resiliency side talk about the importance of understanding AI, how you need to govern it, how you need to partner with the teams. So many of you working on data security, protecting information, making sure that you are engaged, you understood, hey, what's the backup plan? How do we keep the business going if something comes up? The second takeaway I had was BCG recently did a CISO survey, and there were two data points that really jumped out at me. The first is half of attacks in the last six months have been a non-human identity. So this concept of people using AI to do cyber attacks is taking off. That's only going to increase. And that means you have to defend. You have to protect at AI speed now. Now, the second data point I thought was fascinating. For actual AI projects sponsored by the business, so this is not, you know, employees randomly doing AI, but actual AI projects the business went into, half of those projects ended up with unintended data leakage. So there was data that was in the environment, in the system. It wasn't properly secured or didn't have the right identity and access layer around it, and it got exposed to employees, to partners, to suppliers, leaked outside of the company by accident. So almost half of projects had that. And that, again, tells us why it's so incredibly important that we have this governance in place, the data security, identity and access controls. We understand the AI agents operating in our environment. And then my third takeaway was the framing of this is detect AI, understand all the AI in your environment, what's sponsored by the business, potentially what's been brought in that you need to get on top of, you need to manage. You need to govern that AI, understand those agents, understand what they're doing, the operations they're taking. And when you do that properly and you're tracking every bit of data they've read, they've written, they've edited, they've deleted, that's what ultimately gives you confidence as a business owner that you can roll AI out and you know that you can govern it properly. You know that you have safeguards in place. You know that if something goes wrong, and you heard about some of those examples this morning, it's not always a malicious attack. Sometimes it's just mistakes, but you know that you have precision recovery. That gives you as a business owner confidence that you can roll AI out. You can empower your business because of that confidence. I think that's really, really important. So those were my three takeaways. But to really bring this to life, I think the best thing to do is to bring up two industry experts. And we're going to have a conversation with them about how they think about resiliency, how they think about data security, and, of course, the convergence of the two and how they can empower their business. So first up, let's welcome up Wim Gerden from EY. And Gwen and Steve McIntyre from Fidelity. Yeah, don't fall on. Anand took care of me. He got me some new shoes. Look at that. They're about a size too big, so there could have been a stage dive. And that could still happen. We still have 40 minutes to go. But a little bit awkward. But, hey, first up, Steve, you're senior vice president, data security, you know, at Fidelity. Fidelity is a big company, 80,000 employees, 7.1 trillion assets managed. And you have to think about this concept of, hey, how do you empower the business with AI but secure the data, protect the data? And I think what was fascinating is your background. You also led the public cloud journey at Fidelity. And I think there are some parallels between those challenges as we went from running everything on prem in a data center to moving to public cloud and thinking about security that I think are going to apply to some of the challenges that we're starting to see now with AI. Yeah, 100%. You know, it's really interesting when, you know, the explosion of AI happened and suddenly it doesn't just affect security and the attackers. What's really interesting, and it was very similar in the cloud journey, was it also impacted the business and it increased their velocity and their ability to innovate and go faster. And so then what happens is we have to be there and to be able to safely enable that for them. And, oh, by the way, we have to really learn and understand what's ahead of us and be able to take some calculated risks, right? Security is about risk minimization. We never get rid of it all, right? Or we wouldn't have a job. And so by really being able to minimize it and take some calculated risks, not the big risks, there's big breaks you just can't do. And sometimes we have to say no. But definitely, there's a lot of parallels. Yeah, you're right. You're right. So WIM, chief architect at EY, again, incredibly complex company, 400,000 employees, $53 billion in revenue. You operate in 150 different countries with all kinds of regulatory compliance challenges. And it's not just the responsibility of EY's data, but you have your clients' data that you have to protect, secure, and make sure you've got the right resiliency behind it. So we're thrilled to have you here as well to kind of give us that context of how you run such a complex organization. And, you know, we met originally on the resiliency side of the business. Yeah, and resiliency for us has multiple aspects to it. So the 150 countries are really their own profit centers. So every one of those countries, we call them member firms, really has their own requirements about what resilient data means, how we store it, where we store it, how long we need to store it. There are countries where it's an indefinite period of storage, which is always interesting as the formats change. But more recently, what we end up having is really having to deal with the resiliency through our geopolitical things. There was Ukraine attack, get everything out of Ukraine into Germany, Russia split, Russia divestiture. How do we know what we need to send there, what we need to keep, what do we need to keep for later? We have October 6th in Israel, get everything out of Israel and Lebanon. And a couple of weeks ago, we finished getting everything out of the Emirates. And these are all different platforms. Some of them are on-prem. We have about a hundred and something. I'm looking at Jeff, about a hundred on-prem from really small data rooms to big colos to our cloud stuff, all different infrastructure, different platforms on it. Sometimes we know file shares, we need to get it all out, make sure it gets restored correctly and back operational with minimal downtime. That's where the resiliency comes from. I think, Wim, you outlined it, but operating in 150 countries, it's an amazing scale for the business, but you also don't dodge a single crisis. Everyone that's hit in the last 10 years has come right across your desk. But just moving to the questions, and Wim, I'll start with you. If we go back to a simpler time, let's say five years ago, when we used to talk about secure data, recoverability, what did that mean to you five years ago? It mainly meant that from a regulatory perspective, so in the regulatory aspects of our business, that we could prove to the regulators that our data was restorable, indirectly prove to our clients. But if you do audit, if you do tax data, you base that on history, how do we know that if there's a dispute, if there's, I think, wire card for us, not our best moment. But you need to go back and get all these files restored. Speed is less important as precision and accuracy and knowing that it's all there. Because if as part of one of these disputes, it's discovered that a file isn't there, did you do it on purpose? Did you not do it on purpose? So really making sure that the data sets that all these findings are based on are there and complete. And as I said, speed less important because these regulatory investigations take years to resolve anyway. But we need to make sure that we're complete. We cannot deal with incomplete. So that's what I wouldn't say five years ago, because I think the mess started about seven years ago. But before that, and for us, it started with China starting to separate. But that really was before that. It was really making sure that we could prove we didn't lose anything. And Steve, kind of same question. Five years ago, when you were thinking about data security, you were probably in the middle of the move to the public cloud. And how were you thinking about that? Well, it's interesting. When talked about regulations and controls, right? And it was really about controls. And did we have the right least privileged controls in place? Were we doing the right access reviews? Were we making sure we had attribution around service accounts, right? Did you own this set? And what did they have access to? And you attest to that every couple of years. And so we looked at it. Also, what was interesting, we tend to look at it through the eyes of an application. And so it was based on the rating of the application. And data didn't stand on its own as a first-class asset. And that's fundamentally changing with what's happening now. Yeah. And I guess I'll come back to you, Steve. But AI entered the equation the last year or two. You were leading into this. But how did that change the concept of data security for you? So it was really interesting. And it hit home for me when we immediately realized that all of this data that we didn't manage very well or didn't really think we cared about, meaning unstructured data was just kind of there, whether it was on SharePoint or OneDrive or NAS storage, didn't really matter. Along comes LLMs, and suddenly all of that data becomes gold. If you think about the entire history of research notes at Fidelity, it's all PDFs. Think about the value of being able to evaluate that, draw insights from decades of research at Fidelity. And so what really happened that hit home for me was we gave out a few licenses for Copilot. And immediately, like two days in, one of the legal team that I work with came to me and said, hey, we have an AI problem. And they said, one of the women on my team, I asked her to go find something. She did a search, and she came back and said, oh my gosh, AI found all the PowerPoints that were on SharePoint that I used like four jobs ago. And it hit me really hard. I'm like, this is not an AI problem. But AI is just a tremendous search engine, and they said it in the keynote this morning. It runs at speed. And suddenly it's searching everything that you have access to and surfacing that to you in a meaningful way, and suddenly everybody thinks we have an AI problem. What it showed is, back to the comment of securing data, well, that's the area that has to improve. So Wim, kind of same question to you. AI arrives, and we've talked about this as well, like just data getting exposed that you didn't anticipate prior to AI agents, and the speed and capacity they have to find information. I would say we found it about six months before Copilot. When Bing Enterprise Search was launched, all kinds of stuff started to surface. That people went, huh? And so for us, we as EY Global, where I belong, we don't own any of the data. Back to that member firm. Every member firm owns their data. So all of a sudden you have people on Luxembourg finding stuff in Australia, or people in Papua New Guinea finding stuff in Suriname next door. But that is really where the first questions were raised. What's all this? How many SharePoint sites? We have multiple petabytes, I learned this morning, of data. And really what it is, is that it was a wild west. There was no lifecycle management on these SharePoint sites. Half of them had no owners. We didn't know when they were last accessed. So the first thing we did with AI is really starting to try and find out who owns the data. That we had data owners of the repositories. Not the individual ones. The second thing we did is we shut everything off. So really you can't do anything if it's not co-pilot, if you're not licensed. So that's the two things we did. And what we're now doing is really starting to figure out how do we use AI to basically label the data how do we get the data to the repositories? that's in the repositories, to make sure that if a document has draft in it, why do we need people to say this is draft? Open it up and have the, if it says confidential, if it says financial services, these are all labels. And so that is what we're now trying to do at scale for all of our knowledge repositories and unstructured data is how do we protect by using AI on itself, if you want? Because it's unfathomable if we have to have, especially the turnover, about 20, 25% turnover every year. So every four years you have 100% new workforce. You can't keep training and you can't make sure that compliance, you hammer it in, the people are gone by the time they understand. And in the scale of it, right? 400,000 employees is just massive there. And you know, what's fascinating, so many customers I speak to have the same challenge of data that was on a SharePoint site or in the environment. It had been there all along. It just didn't have the right security, identity and access controls on it. And then AI exposed it and it wasn't an AI problem. It was just literally the productivity and the ability for AI to find things quickly is what surfaced and it happens everywhere. I spent 20 years of my career at Microsoft. And I remember years ago, Microsoft had, of course had SharePoint deployed everywhere, but they linked every SharePoint site together. And there was a global search index and we were quite excited about it as the leadership team, people were gonna be able to collaborate, work together in a better way. And about 24 hours after we rolled that out, we had to unplug it and we had to unplug it because employees were searching for, you know, salary information, reorganizations, mergers and acquisitions. All that data had been sitting on these SharePoint sites all across the environment for years. It was just the discoverability. And Microsoft Search is nowhere near as efficient as what you can do with open AI and Anthropic and Copilot, Gemini for the enterprise, et cetera. So I think we've really just run into this issue that was already there. It's been, you know, surfaced to everyone via AI. But Wim, kind of coming back to you, you talked about resiliency, you talked about AI. How do you now think about governing AI agents and how do you think about that risk curve of empowering the business units or your franchises to use AI? So the first thing is we have to know what was there when the AI ran. But that's not any different than we have to know what was there when the person ran it. So yeah, we need to have the historical picture, the versions. The second thing I kind of talked about, which is really we need to have, go way beyond labeling of confidential, non-confidential. We need to have geo-restrictions, geo-labeling. We need to have our line of business labeling. We're linking it to our contracts because we get an enormous amount of client data. In every contract now, we're specifying what can we do, what cannot we do. Can we use it, anonymize it? Because what the other thing we're doing is, is we're looking at AI to make artificial data products. We can't just do that out of thin air. We need real data to make synthetic data so that we can run tests and benchmarks and all of that. So really it also has a big kind of on the commercial side impact on what can we do with our data? Who can see it? How far can we go? Do we need to stay in a line of business, in an industry, in a country? All of that is now being codified in contracts. That's the easy part. Then we have to codify it in whatever. Some technology structure with some system and that is for now still very, very cumbersome and creates a lot of fear because we really don't know what all will happen if one of these things gets out that wasn't supposed to get out. It's interesting what Wim said about contracts and we actually are seeing this now where partners will, in the contract, explicitly say what data they need to perform the function and we'll try to redline in that they're not accountable for a breach of data that we send that's extra. And they're literally trying to indemnify their self because of the speed and the vulnerability that is now in front of us with AI. And then the last one on my side is really exfiltration of data through engines. That's why all the public engines are shut off. We're pretty locked into the Microsoft ecosystem which we kind of generally understand what happens with it. Stays in our tenant, stays somewhat in our geo boundaries but really we can't afford somebody's earning statement to slip out. Yeah, no, and I think that's so important. I mentioned the BCG survey but even for business sponsored projects, right? AI projects, still almost half of them having data leakage. And then, you know, I think the proliferation and we were talking about this, you can run Gemini. It's actually included in the latest build of Chrome, right? So you can run that right in the web browser just blocking installing chat GPT or cloud isn't enough. And that's something that we're all challenged with is that you don't have that right data security, identity access controls that employees bring in their own AI tools. You're gonna have leakage issues that are quite serious. Yeah, and there's a real conversation we're having about the telemetry, not that we do. I mean, we have enough telemetry, right? I mean, we create more and more telemetry every day but now we're thinking we need to get more telemetry off of the end points because people can do so many things on their local machine and employ their own agents. And, you know, you catch a lot, you try but the controls are the controls and they're not foolproof. And so now we're having this conversation about do we have to take control of IDEs at a really fine grain level? And that really is impactful to the development process. And so there's a lot of struggle in terms of balance. Again, I go back to what I said about that parallel to the cloud journey in terms of balancing the risks of that individuality and the style that the development teams wanna use versus, you know, whether we wanna lock down to be able to protect, it's getting harder. Yeah, so Steve, coming to you on data security. So that's your domain, your expertise. Obviously it's changed radically with the implementation of AI. How do you think about governing AI across Fidelity? So we really look at it, we're trying to look at it kind of in three pillars. One, we have to know what's being used. And so that really brings into play the idea of shadow AI, shadow IT, all those kinds of things. And it goes back to a little bit of what I was saying about the endpoint data. We have to know that the inventory is accurate. And the inventory is not just the assets that are being used and deployed, but are they aligned to the use cases that are registered and approved? That way, at least we know that, you know, if John's working on something, he should be using cloud because it's tied to a particular project that was approved for that. And so you have to do that inventory, you have to have the assets. Then we have to think about what's the safe environment where we want these agents to run? And how do we want them to interact with the foundational models? And so what architecture do we put in place to funnel all of that equity into a place that gives us the right visibility and telemetry so we can see that agents and applications using AI are behaving in a way that was intended? Or, you know, on the backside, just misbehaving on the way back because the model made a mistake. And so, you know, and as Wim said earlier, that model making a mistake, nothing, there was no intent, but it can happen. The impact is just as devastating. And so you have to really balance it. And then the third piece that's hard for this is we're struggling with agent identity. And how do you give an agent identity and do you persist it? And, you know, you go through the research and we've talked about it, do you give everybody, you know, a fidelity ID, right? Do they become employees? Well, what if my agent only lives for seconds? Like, it's a really interesting problem. And I don't know that anybody's solved it really well yet. And in a regulatory environment, I need to be able to go backwards. I need to be able to say, I'm looking at Drew sitting out here now. If Drew used the mobile app and he executed a trade, I need, and there's agents involved in that, I have to be able to tie attribution all the way back that says that agent took that action on that data set that's now in our record-keeping environment because Drew asked it to do it. The governance around that is really difficult. And the accountability model that you have to put in place to roll on all those hops, that's probably 13 applications of fidelity that are involved in that single transaction. So that's really hard. So Wim, how are you thinking about that agent identity concept? I think this is something everyone does. So for us, as one says, there's very rarely anything net new under the sun. So we ran into the agent problem with RPA bots. Now what are they? Yeah, because some of them triggered an extra SAP license, really. I mean, and they're all doing something on behalf of somebody else. So that we, from an economic perspective, we saw a lot of runaway licensing costs, the same now. When we reassign identity to them, it kind of proliferates to it. It's the logging, the auditing of what they do, the disclosures we need to do. Yeah, we live in the world of the European AI Act, the California AI Act is like kindergarten compared to that one. And so we really need to really start to massive all of our disciplines around application, portfolio management, inventory, CMDB, they're all coming back. But now sometimes at the scale that is so small, there's debates, is a prompt an agent? Should we take prompts? Does a prompt have an identity? Yeah, normally it does, but it's the start. And so really that is where the challenge is, where do we strike the balance between the business being able to go? At what point does something need to become on our radar? And at what point does it become really, really relevant to the business? And it's somewhere on that continuum. And we're using, as I said, we're using a lot of lessons learned of the RPA world, because inherently it feels a lot the same at the most basic aspects of control and of data access and licensing. The one extra thing what we saw is, especially when you have agents calling agents, calling agents, if A calls B, B calls C and C calls A, there goes your accounting, there goes your budget for the day, the month, the year. And there is very little in the infrastructure to detect circular agents, especially in networks of agents. And then the identity sometimes go crazy in there too, because they're swapping identities as they go along. Anyway, that's another aspect, which, yeah, SAP was very happy with us there. And what's interesting about the controls, et cetera, right? And I don't know how you all feel, we can chat about it after, but I just took on risk and controls recently and my controls lead said to me, he goes, Stevie said, we're struggling with this because GRC is historically a slow human driven process. And so they're really trying to figure out how do we build automation? How do we use AI potentially to help the GRC function get aligned to this? Because this is moving at light speed and they're trying to still drive the Volkswagen from the 60s, right? Yeah. Hey, and again, first question, when we started out five years ago, we weren't talking about any of this, it's how fast it's moved along. But Steve, one question for you on that spectrum of you want to empower the business, let them take advantage of AI, but you need to manage risk and security. How do you think about that spectrum? And then how does being able to govern AI change that? You know, I think one of the things we've learned very quickly with this at Fidelity is over the course of time, our technologists have not intentionally, but we've really usurped ownership of data from the business. This is making us realize we gotta give it all back, right? Like the people that own the data have to make decisions about the data. And so there really has to be commitment from the business to help us govern. And so, you know, go back to what I said earlier about a use case. So if I say I want to take equity research and I want to take it, I want to run it through a set of models and I want to gain insight from that, asset management has to be aware, the business of asset management, not their CIO, their business has to be aware and be okay with that. And that's where the governance has to start. Then, you know, and one of the things that I've been embarking on for a number of years at Fidelity is just creating an asset and inventory of data as its own asset class. You have to understand your data. Otherwise, how can you make an informed decision about where it should be used? And so that's another really big part of this that, you know, we're up against it because we're chasing the game. Yeah. Like the cat's already out of the bat. It's true. It's true. So Wim, your thoughts on same question, you think about the spectrum of risk versus business empowerment, obviously the complexity that we talked about before you have to deal with as well, but how do you make those decisions on that spectrum? It's very carefully, very slowly, very methodologically. Yeah. Everyone in the general, what we call enablement functions is hanging on the brakes. So that's risk, illegal, IT, HR. Yeah. Because the business wants to go way beyond, way too fast. And I think an interesting lesson, luckily our audit business learned in the very, very beginning, that if you think back three years ago, few years ago, any, most of the LLMs would pass a CPA exam. most of the LLMs would pass. That was not the problem. The problem was that every question had an equal likelihood of a wrong answer. At which point you cannot build a control structure on the quality of the output, because you have to check everything the thing did. If you think about your college graduate, they'll have, you have 5% of the CPA question and 95% of the people will get wrong. So you can build your traditional pyramid of supervisors and the like and going up. And that was a very early on discovery, which at least on the audit side, they slowed a lot of things down and thought about the technology in a different way to basically assist instead of do. And I think in our tax practice, they kind of looked at that and waited a bit and are now more comfortable about starting to do. But I think if we wouldn't have had that, that experience three years ago, we'd be in a very different kind of push and pull mode, but something really, it broke, it broke the whole model. And so you cannot run a business where if it mentioned, you have to check every, every answer. And that push and pull, right? Like you said, between what the business wants to do and how fast they want to go. And then what we're comfortable with. I mean, there's not a lot of precedent setting cases out there yet with this stuff, right? And so, you know, our legal team has a lot of concerns and rightfully so. And so, you know, I spend a lot of time with our LRC community, just educating them on this and educating them on what it means and where we can make an informed decision. And again, that takes time to your point. And we're trying to get faster, but it's hard. And the cynic in me said about two years ago, I think our best investment was to sue ourselves on the use of AI to see what case law would be made, because then at least we had a framework in which to run. I like it. That was not a very high, that didn't make it on the investment. But it is something we all will have to eventually deal with. What will be the court's decisions? And then how geographically will it be different? What will be allowed in France? Will it be allowed in Germany? No idea. Yeah. And we'll have to work through that. I do love that example, right? The LLMs can already easily pass a CPA exam. That's been going for quite some time. But there's a chasm between passing a CPA exam and having confidence in any type of AI driven audit, like they're just not remotely the same thing. And I think that's the thing that we're all challenged sometimes, explaining this to the other business units, all of you just mentioned HR, legal, the different units, they come to you. They think you're the expert on AI. They think you're the expert on data security, business continuity. So you're constantly having to educate them on what you can and can't do. 100%. You know, it's interesting, the ability to get someone to understand is interesting because they don't want the accountability. And so you're trying to educate and get them to agree to something. And it's so fast moving and ever changing. And that's why sometimes, you know, you talk about all the regulatory frameworks, having a platform that can actually do a lot of that for you helps you frame the image a little better for people and give some comfort that we're trying to keep up or someone's working on our behalf to keep us up with what's changing in the industry, because it's changing really fast as well. And, you know, I invited you both to this panel because I think you're on the leading edge of data security, resilience, taking advantage of AI. So, you know, as we close the panel out, Steve, I'll start with you. But what words of advice do you have for our customers, our partners here? Is there, you know, in this journey or embarking on this journey, any quick lessons learned that you would share with them? I think the biggest thing I would say is just, is get a handle on what's actually important for the business and the use cases, and then understand, get a handle on your data. And then if you can marry those two, you can make some, you know, risk-based decisions on where to apply the work and where you're just not quite comfortable. And so here's an example, right? If we, let's say we created a chatbot, conversational chatbot on fidelity.com, it would take minutes for the attackers to come. And basically, they would try to prompt engineer that chatbot to tell them to take all their money out of fidelity and put it at Charles Schwab and then post that on Reddit and chaos would ensue, right? So for us to make that decision about making that and letting AI represent fidelity to you as potential customers, are we ready for that, you know? And are you ready for that as customers, right? That's a different experience for you all. And you know, that attribution and that control, the intent of the control is one thing. The examination, the control has to be a hundred percent accurate. So to your point about using AI to do controls, sampling is going to find where AI missed and now your control fails. Yeah. Yeah. Same question. To build on that, but I'll spin it more of an internal use and yeah, of your internal IT, it's all about cost cutting. So there's two things there. One is an education about what does non-deterministic mean? They're used to, yeah, most people are used to, if you ask computer one plus one, it's two. Yeah. If you ask an LLM, if you ask it enough, it might say it's 17. So what does not, and where does that matter? Because what we really realized is that there's a lot of processes, especially knowledge worker processes where we actually don't know what good is. If a person does a task in the aggregate, the group of people in 90% accuracy, why do we want an LLM to do it at a hundred percent? If it does it at 92% accuracy, it's better than the humans, but we don't know what good is. We don't know. Is it 90? Is it 80? Is it 70? And so we're navigating this landscape from an IT perspective where for 70 years, it's been one plus one is two. And now we're in a different world with this technology where the businesses that really still think this thing is going to each time give the same answer. And when they question why it's not, I've asked, what do the humans do? How good are the humans? How many mistakes do they know? Silence. No one really knows. And so until we find some way through that, that we say, okay, how do we compare an LLM output to a human output? Not just at the cost, because that's like, yeah, how much does your colder cost? That's the kind of irrelevant or lines of cold, but really what is a truly relevant quality metric and how does that compare? Till we're there, it's always going to be a bit of a uncertainty and we're going to have to go probably a lot slower than we would like. And a lot slower, not till that gets resolved. Well, it has to get resolved because knowledge workers and quality is, I don't know if there's anyone wanting to go get to do a PhD in operational research, but that would be a good topic. What is quality for a knowledge worker? I don't know. And when a human makes that mistake at 90%, we have someone we can talk to and educate them. Ventire and punish. I can't blame the model, right? Like the model is the model. Like I don't have any recourse there. Well, perfect. Well, gentlemen, thank you so much for joining the panel, sharing with the audience. I learned a lot. I'm sure they learned a lot. Let's give them a hand. Thank you, Steve. Thank you very much. Thank you. Thank you. There you go. Thank you very much. Okay. Thank you. So I learned a ton from that conversation. I think it's really some intriguing things we have to think through. An AI agent. Isn't it identity? Is it a person? How do you treat it? And it's so fascinating. Five years ago, we never would have been talking about that at VeeamON. In this morning, Anand, Rehan, they took you through the vision of our platform. Where we're headed. How we're going to help these topics converge. How we're going to help resiliency, data security, AI governance come together. I think the key thing I learned in that conversation is when you have confidence in your ability to govern AI, when you have confidence in your ability to restore your data, you can then empower the business. You can let them take advantage of AI, get those rewards, but still have the business continuity that you need to do your jobs. Now, that's where we're headed. I do want to come back to, in the next three months, we have three products coming out that enable and empower part of that vision. The first is Veeam Intelli... Oops. Nope. First is Intelligent RevOps for Microsoft 365. So we have 25 million Microsoft 365 users that we protect. We want to help you take advantage of this to get the insights from your data. The second, Veeam Data Platform version 13.1, 70 new features. Many of you use that product today. That's coming out in the next three months. We want to help you adopt that, get value from it. And then, of course, resilience on the Data AI Command Platform. I think that was the loudest clap we got when Rehan talked about, you connect to VDP, into this platform, suddenly you get more insights. But we're doing it in a way that's easy to migrate, easy for you to take advantage of and get the value there. Now, we came out with the Data Resilience Maturity Model a few years ago. Many of you went through that, took advantage of it. It's a way to make sure you had cyber resiliency, you had business continuity. Anand announced the new Data and AI Trust Maturity Model. So we've taken these elements of resilience, of AI governance, we've brought that together. And I think this gives you a map. Many of these issues that Wim and Steve just talked about, this will help you assess your business, assess your controls, assess your data security and resiliency. So if this is something you're interested in, reach out to your Veeam sales rep. We'll run you through this process, give you the scoring, give you the insights. But I want to work with you. I want to help you. If you came to VeeamON, that gets you a ticket for us engaging and helping you through this process. So please take advantage of that. And with that, we've reached the end. I want to thank you again for coming to VeeamON. I know you're incredibly busy. We value your time. I hope you're as excited as I am about this concept of resiliency, data security coming together and how you can empower your businesses or your customers, if you're one of our partners, to take advantage of AI. So with that, thank you very much.
