Transcript
I'm your host, Rick Vanover, the Rickitron. And with fresh perspectives to wake up to today, I'm joined by Emily Tellez. Emily, thanks for joining us. Thank you for having me. Tell me a little bit about your role and what you do here at Veeam. Absolutely. So, my name is Emily Tellez. I am a field CTO here at Veeam. I've been with Veeam for almost close to 12 years. Various sets of roles from, you know, being part of our sales organization to an inside systems engineer, which was doing back-to-back demos for different accounts and different clients to, you know, actually deploying and integrating and architecting our software for different organizations across seven different states here on the West Coast. So from that, I've been able to join the strategy team. And now I work with analysts and customers globally, as well as working with some of our alliances and partners to build some of our security integrations that you see today in the marketplaces. That's right. And the Veeam of today is the conversation that we're going to really focus on in this episode of the Wake Up Podcast. And I'm going to start with a compelling question. Emily, feel free to push back. Trust me, we didn't practice. It's authentic. But I've got an assumption, and I want your take on it, that organizations today on an AI journey are really facing a risk that that data is not trustworthy. What's your take on that? So, agree. I think every organization right now is on their own AI journey. It is probably mandated to them from either a board of directors or their C-suite executives or maybe just internally employees wanting to do better, increase operations, increase the way that they do day-to-day tasks, overall take down time. And the way that they're looking at it is they're looking at it based off of tools that are readily available to them. So when we think about Gen AI becoming a huge explosion in the last two to three years of now, every single person can walk around and they have an AI tool or an AI companion now in their pocket, right, sitting on their cell phone. And so they're used to being able to leverage that tool to get quick answers to whatever it is that pops in their head throughout that day. And I think it's the same when it comes to organizations wanting to take that power and utilize it for better outcomes. The problem is is that we are putting in data based off of assumptions that we have. And I think we all kind of know that based off of if you've played with chat GPT or any one of those different types of models out there, right, you ask it to do something and maybe the output is good. Maybe you go to have it create some type of photo and maybe all of a sudden you have some random floating artifacts that you didn't ask for, but now it's in the photo, right? So it's just looking at different ways in which we actually leverage those tools that have been given to us. And it's the same for organizations. They have to be able to trust the data. If they're making assumptions that the data that they currently have in house is going to work for them and that they could just make or take actions based off of it, well, they're going to fail and they're going to fail pretty fast. I think there's really something to dig into about that. And I'll take my own copilotry, if that's a verb, take my own GPT-ing, if that's a verb as well. What I've learned is the better you prompt, the better the response, right? In fact, we were working on an initiative and we had like a 14-mile prompt for an agent that we shared. And I think that speaks to a real change from going from like passive to real active AI decision-making. And I think there's an opportunity for potentially a misstep of organizations maybe being quietly wrong with some of these decisions, bad prompts, bad data. Do you think that risk is real or just a dream? So I could say even going back to like 2022 when we started having these AI conversations, you know, we would go and we would meet with different organizations and they would talk about some of their AI projects or their AI journeys and, you know, where they would actually create very specific use cases. So there's an organization that I worked with that dealt with it from a water initiative. Hey, we want to be able to understand from a water sensor perspective what types of counties are actually feeding or receiving water based off of, you know, different utilizations throughout the year. And so they were using this to kind of metric or meter the system. But the problem was is that in order for them to check those sensors, a human had to be in person on site at data center. And for them, they were thinking about, okay, well, the next closest person to get on site to this data center is 35 to 45 minute drive away. So what are some different ways that we can leverage AI and leverage different types of tools so that way we could save that time from somebody having to drive a 35 to 45 minute distance to be there in person to actually be able to check on these sensors. So that's what kind of started their AI initiative. It was actually designed so that way they could actually stop or save some time from their actual personas. But then on the other side, what they ended up being able to do was leverage this technology to understand, okay, you know, what's the accuracy and the flow rate? You know, what are some ways that we can see different counties that are going to be receiving more or less water? What is the use cases or the utilization of it throughout different periods when it's summer versus winter? So because they tested it and they built it for a very specific use case, they were able to get additional outputs. But then you go back and you look at it from, you know, this was a project that they ran two to three years ago. You designed it for a very specific use case to save time and now you're just trusting it that it's going to be able to give you this accurate information. Is that truly the case that you need or, you know, was there other data that we kind of funneled into that so that way we can get some more accurate results based off of what it is estimating in the actual flow rate? So those were some of the conversations that we were having, you know, three years ago And now we fast forward and we see different types of projects, you know, from different organizations of how they've seen AI use gone bad, whether that was with a particular, you know, user that was feeding data that shouldn't have been fed into public systems, whether it was them allowing actions to be taken on behalf of the organization based off of information it thought it had and that thought it knew. So we're starting to see a lot of these different types of outcomes kind of start to spin out. But the good news is, is that we're at least recognizing them and we're recognizing that these are these are problems that, you know, we can try to start solving for. And that is something to wake up to. Now, that leads me to a really good example, Emily, of a of a situation I had with a bank recently. I was talking to a risk manager that was the function, a risk manager, and the individual had put probably a year and a half's worth of team budget and extra purchasing into an AI fraud solution, fraud detection solution should clarify that as a big difference. Yeah, but it was an AI fraud detection solution and. The individual was mortified about the audit and the reason that, you know, they were so mortified about the audit is the organization had not done anything in scope of auditing. And what they told me is that the team worked really hard to do was to focus on explainability, showing and documenting the flow of the data, where it was touched and how it was touched and the permissions and all of that. And that really conveyed a massive amount of trust. And I think each of these examples, whether, you know, the water example, the fraud detection solution, they make a big difference. So think about these learnings. And so I think that organizations, yes, the rules are always there. But what about the velocity? Sometimes we're in this challenge of the innovation may outpace the controls. A hundred percent. What's your perspective on that? It's it's a very it's it's very different when we have to kind of think about it in terms of, you know, the way that we want to accelerate innovation, but then also try to put guardrails on it. Because when you think about guardrails, immediately everybody thinks slow down. Right. Because now you're trying to tell me that this is how we need to do things. And maybe it's not a way that's going to be feasible in the way that we want to actually build out workflows, take action, ingest data, whatever it may be. But the good news is, was when you put guardrails around a project or around an AI initiative, right, then you have a less chance of it actually failing because you've actually took the time to understand what it is that we want to actually build. What's the outcome? What's the explainability? And so then once we have those guardrails put in place, we know that we have mitigated our potential risk that could unfold onto the company. So, you know, again, being a filled CTO, I go and I meet with lots of different types of companies. And, you know, probably about two weeks ago, we did an event in in San Jose and, you know, I had the opportunity to sit at a table for about an hour and a half and talk to head leaders from all different technology based companies. So, you know, right next to me was like a social media executive. And then we also had somebody that was from a background of enterprise storage and then also software and whatever it may be. Right. And we went around the table and we talked about their different types of AI initiatives. And, you know, we have one organization that says that they've already rolled out 10 to 20,000 AI agents within their internal infrastructure. Right. And they have a committee, an AI committee that sits through and every quarter they're going through and they're validating and they're asking questions and they're having teams produce the project and tell them this is the project that we're working on. This is what we're what we're planning on leveraging it for. This is the outcome. This is the use cases. This is the data it's going to have access to and the people that are going to be able to prompt and or utilize it. So they had a full system that was dedicated to it. And then on the other side, you had somebody else that said, we're doing the exact same thing, but we also hire a third party company that comes in and does pen testing against our different AI projects. Right. So we're wanting to understand, is there any vulnerabilities that we have in there? Is there any prompt injections? Is there any ways that, you know, somebody can take this information that we're capturing and gathering and utilize it for bad use? Meanwhile, right next to me, I had a gentleman said, well, we just rolled out cloud code to every one of our employees here and we want them to engineer, to design, to build, to do whatever they want. And they could have never designed or coded a day in their life. Right. But we're going to give access to every one of our employees to build agents for their specific use case and what it is that they do day to day to help them from an operational standpoint, which it's great to be able to do. But then when you sit down, you think about it, you're rolling it out to all these different individuals. But have you put any guardrails around it? Have you explained to them what access of data that they're going to have and what types of action it's going to take? So I think everybody's going to have different levels in which they're going to want to go and drive towards an innovation strategy. But the guardrails is very important because, again, if you're not putting those in place, what risk are you going to introduce to your business? And then if we start thinking about it a long way down the road, there's already news articles of different companies that had some type of material breach or material exposure that was due to some type of leakage of data or some type of A.I. initiative that went wrong. And then it puts it back into perspective. Well, then who's going to be held accountable? Right. And generally, most points of the time, it's going to be your C-level executive team. It is going to be the board. Yeah, the responsibility falls up. But I think that's a fantastic example of three different perspectives of risk tolerances for this agentic era. And I can look at even my own personal practice where I've made agents, I've shared agents. Am I doing the right thing? I don't know. I'm doing all right. But, you know, I think Veeam's in the market, make no mistake, to help folks get that visibility around the agentic era. So, you know, of course, go to Veeam.com for more information on that. But Emily, one question for you here. What might be a signal of an organization falling behind or maybe something's not working as expected in some of these different journeys? You take those three examples, like where's maybe the writing on the wall of something not going well? Well, there's tons of examples, right? If you're ending up in the news, it's probably you've already gone too far. You failed pretty fast and pretty publicly. But for some organizations, it could just be a sense of just, you know, even from just the very beginning stages, if you don't have that explainability or the outcome in mind of what you're trying to drive to, then I would say pause and stop. OK, what is the actual use case that we want to be utilizing it for? Is it we're just going to enable copilot and we're going to be able to give everybody broad access rights to go create whatever they want and then see what happens? That shouldn't be the use, right? It should be what can we do in terms of helping our organization either broaden when it comes to overall operational costs? Is it taking mundane tasks that can be easily repeatable and be utilized for something else? We're here at RSAC, we're here to talk about security, right? There's a lot of really great initiatives and a lot of great innovation that has happened in terms of the actual security vendors that are leveraging AI to help them in doing mitigation, being able to respond to potential threats, be able to build out incident response playbooks to deliver them a better and more repeatable outcome. So for some organizations, I would say it's not about failing fast. It's more about just understanding what is that actual outcome that you're going to drive through and then making sure you actually create a committee of of different stakeholders. So that way you can have those open conversations before you start building and testing and, you know, potentially failing. And from that perspective, you walked right into my next topic, Emily, and that's around the stakeholders. And I think the the committee, the the real central theme is leadership, is executive support, executive sponsors and things like that. What kind of questions would you ask or what kind of advice would you give to organizations and business leaders that are dealing with these challenges yet have their business to run? What would you say to them? That's a pretty heavy question. That's what we do on this show. That is not a question I want to wake up to. There you go. I would ask them, number one, like what what is the expectation? What is the expectation for your teams to be leveraging AI first? Right. Even asking that simple question can help to pull out a lot of different different outcomes that maybe you weren't even aware of. You know, maybe somebody is driving to leverage AI to help them mitigate better response tactics when it comes to a cybersecurity practice. Or maybe it is leveraging AI to help with, you know, doing ticketing systems internally. Right. So so I think just even starting with that very basic question of what it is like, what are you trying to achieve? You know, and what is your expectation of the organization and leveraging it? And its users, because once you start to identify those topics, then you can be better set up for success. It's almost like we have to go back to that original, you know, the original piece of conversation where we always talk about, which is it's people, it's process, then it's technology. The technology is always going to be there. You're always going to find a piece of tech that's going to be able to do what it is you need it to do. But if you don't have the right process and if you don't have the right people and you don't have the right guardrails, well, then the technology doesn't mean anything because it's either going to fail fast or it's going to put you in a situation where you're heading up in the headlines for the news. Fantastic perspective, Emily. And it hits on one thing I like to say a lot of business benefit first, compliance always. And so my perspective, the question I would ask to decision makers today is to really think about the cost model, right? Because this stuff, you know, especially to do it correctly, isn't necessarily free. And the thought here on that is if you don't have the business benefit right and the cost model isn't right and then you've implemented it, guess what? It's too late. We've learned that through cloud economic models before. Right. So I think those are just just a few things to get started with. And last thing for you, Emily, is just what's the risk of doing nothing? Falling behind. That's number one. Right. I mean, I think there's a lot of really great use cases and, you know, positive outcomes that can be driven from those that have started experimenting with with AI, you know, whether that's, you know, helping you to become more creative in terms of how you want to brand or leverage a specific message or, you know, again, going into it from a risk perspective when we're working with these security analysts and these SOC teams. OK, what are what are things that maybe we didn't think about prior, but we can funnel in all these different types of signals and be able to make a better decision based off of the evidence we are able to collect. So for a lot of organizations, it's being able to take that information coordinated out and then be able to deliver an effective response. But if you're not leveraging AI to do any of those types of outcomes, then your business is going to fall behind. And there's a competitor out there that's definitely going to jump on it and they're going to take quite quite a bit of the share. So so definitely recommend like have the question, have the conversation to just understand what it is that you can be utilizing AI for. Truth Soup for free on this podcast. No joke. Emily, thank you so much for joining us here on the podcast. Thank you for having me today. All right. That wraps this episode of The Wake Up Podcast powered by Veeam. Find more episodes on a podcast platform near you and more information at Veeam.com. The Wake Up Podcast is brought to you by Veeam.
