Alrighty. Let's get going. So welcome to our webinar. We do try to do a webinar every month. This is our June topic, which is AI at scale. So trying to cover a lot of different things. So we're gonna talk talk about agentic AI, governance, how to build your skills and culture. All of those things sort of indicate whether you're gonna get off to a fast start, a bit of a slower start, or stalled out. First though, I'm Robert Curtis. Most of you know me. I think I recognize a lot of the faces in our webinar. So it's a pleasure to meet you again. And if it's the first time you are joining us, welcome. I am the managing director for Interworks Looking After Asia Pacific. I've been with Interworks somewhere in the neighborhood about twenty years. So I've been here a very long time and I've been looking after this part of the world as theater for Asia Pacific for more than a decade. So I've got an opportunity to work with entities, organizations, not for profits, government agencies within Australia as well as all over Southeast Asia. So it's been a pleasure and it's given me a pretty unique perspective. So hopefully I can give you some of that insight today. A little bit more about Interworks. We do three basic things, data strategy, data solutions, and data support. A little bit more detail on what that means. If you think about the life cycle of an idea as it sort of matriculates through platforms and data and then down into the building of a value for a solution and then how you sustain and grow it, that's kind of how we presented our solutions here. So we have a whole bunch of strategy services that we offer. That could be your data strategy. It could be AI specific. We could be talking about, your architecture or governance policies or frameworks. From a foundational standpoint, we do help you build these things. So it could be looking after and building your cloud footprint, your data itself, so that could be your data application, the data pipelines, the architecture, all the governance around that, which could be quality control, master data management, cataloging, indexing, all of that stuff. And then the fun part, most people want to get to, and this is the stuff that get the executives excited, it's the solving of problems. That could be analytics in the form of operational reports, advanced analytics, prescriptive data science, AI solutions, which of course we're going to be spending the majority of what we're talking about today on. And then once these things are built and we're looking to grow and sustain and maintain, we support you in a couple different ways. We can support your user community, which is invigorating them as a community of practice. We can support your individuals through skill development, mentorship, other ways, coaching, as well as support the applications themselves. So if you want us to help you look after your cloud or your data platform or your specific data pipelines, we can do all or some. We are very, very happy to be a bit of a puzzle piece. And whatever part of the puzzle you need us to finish or perhaps a section of the puzzle, Interworks is happy to help. A little bit more about us. I am going to fix that slide because I keep telling myself every presentation I'm going to fix this slide because it is out of date and it is annoying me. We are not twenty seven years in experience. We are thirty. Done. Fixed. Now I don't have to get irritated myself every time I do this slide. We formed in nineteen ninety six over here in Australia. I've been doing things here since twenty fourteen. So a tremendous amount of experience and pedigree. Seventy five, probably more like seventy seven, I haven't counted lately of the Fortune one hundred are or have been InteWorks customers. What's what was the very funny. I I can't say the specific names because obviously we have nondisclosure agreements. But for our analytics partner, we built their data infrastructure. And for the exact same data vendor, we did their analytics. So it's funny. Two big names, and we're literally doing the cross of the others, as well as a whole bunch of other top ten type companies. You can go into the Torx website and look at our blog and case studies and get a sense of who some of those big names are, but across all imaginable verticals around the world of all different sizes. Have an industry leading blog. And when I say industry leading, I'd be shocked if there were, other companies that had three and a half to four million page views per year coming to read about the semantic layer or Tableau server security permissions. It's all the different stuff that we put on there over the years. So tremendous amount of knowledge share that we put onto there. And and quite frankly, whenever I meet new people, half of the time they're like, oh, Interworks, I know you guys because I read your blog articles or I have printed off your your workflows or whatever and put it in my cubicle to help me do the things I do. I mentioned we have a lot of customers, somewhere around four to five thousand at this point across the world. Some of them are very small, some of them are very big. We are looking for folks that want to partner. So it doesn't really matter to us. Data is data to us. You understand your business. We understand the tech, data, AI, governance, etcetera. Come together. We can make great solutions. One of the little awards that we're very proud of is the Forbes Small Giant. So some a couple years ago, Forbes put together a list of twenty five small companies that greatly punch above their weight, and we were delighted to be one of those twenty five. So we were the one in data and analytics. Small company, maybe even though we're global, maybe we're three hundred people. But we've had the honor and privilege of working with enormous companies doing amazing things for years and years and years. So we maintain these relationships and we've got just a backlog of folks that we have really, really great relationships with. Great. That is enough of the preamble. As we get into the actual webinar content, those folks that have been here before, hopefully, this is all second nature to you. But for those that haven't, we have a little chat. So you can throw any questions you've got into the webinar chat. I'll do my best to have some time at the end of this to answer these questions. It's not always possible, particularly if we've got a pretty deep topic like we do today, and I've got a lot of very rich slides. There's also a q and a if you wanted to chuck your question into there. I do my best to try to have both of them up. Although now that I'm looking, there it is. There's the q and a. I'll chuck that onto my screen. Either one of those is fine. Like I said, I won't I probably won't stop in the middle of the presentation and address a question. I'll try to get back to all of these at the end. So if you do have a specific question, write it out so I have a little bit of context versus maybe if I answered it right on the slide you were talking about, that'll help me get you a better answer. Alrighty. So let's talk about AI. And I feel like every time we have to talk about AI, we do have to frame the opportunity. And so I've got this slide. And these are some very big names and very big titles. We've got Google CEO, NVIDIA, Microsoft, McKinsey, etcetera, etcetera. And in most of these conversations, these webinars or speaking engagements, whether it's at a conference or whatever, a lot of the if you were to survey my perspective on this or the things that I'm telling all of you, over the last two years, my messaging was more AI is coming, you need to be prepared for it. And the answer was, you need to focus on your data. You need to add governance. You need to have a really clear data strategy that will become an AI strategy. Once you have those things, then you'll be ready for AI when AI is actually materially here. I can tell you it's here. It will only get better, but what we are seeing in the marketplace from the tools and from the solutions and the things that we are doing ourselves, it is a in some instances, a five x multiplier for efficiency, and then some instances, a thirty x multiplier in efficiency. So it is well and truly here. And these quotes speak to that. I won't read all of them. I think I like the one by Microsoft. So we are not in a technology cycle. We are in a restructuring of every industry on Earth. Businesses businesses that act now will define the new incumbents. I think that sounds like it'd be sort of bordering on the edge of hyperbole. I don't think it is. There was another quote, and I don't know if I put it in here, that was saying something along the lines of AI is not like the steam engine. It's not like coal. It's bigger than that. And what we're seeing from when AI kind of became in the modern or common lexicon of everybody, which is about two years ago, two and a half years ago, where everybody started understanding what ChatGPT was. The difference between now and then is just amazing. What you can do right now, and quite frankly, I'll I'll talk to a lot of customers or we're we're working on things together. And what we can do now versus thirty days ago is actually measurably different. So if you think about what things are going to be like in another two years or five years, it's crazy. And so I really want to emphasize that if you don't have an AI strategy, if you don't have a significant amount of interest and investment put into your data, you are behind. That doesn't mean you have to stay behind. You can accelerate. But it is no longer a time to get ready for AI. It is time to operationalize, productionalize AI. I had this slide in a previous presentation and I liked it, so I brought it forward. But there's kind of two big ways that AI fails. You either have the mouse of me or you have the wrecking robot. One of them you go too small and the other one you've gone too big too quickly. Now and I think I did this like maybe a year and a half ago. So the mouse, poor use cases, meaning we don't really know what we're going for. It's not clearly defined. It's not something that's gonna materially impact our business. We're not having buy in from the executives or from the stakeholders. We're off doing our own thing and don't really have anybody to either, one, fund it or two, approve it. Or our data is just poor. And what happens is you end up sort of spinning your wheels, a little mouse runs around, and you don't get any value for it. It's just distraction. On the other hand, there was the wrecking robot. These are the ones that everybody sees. These are the big news headlines that you see on the paper or on Twitter. You've got bad data, which produces bad results. The model has been trained, which produces bad results. You've got no governance or poor governance, which produces bad results. And no security means people probably go and grab your data. The verdict is disaster. You are handing in your CV to everyone else and to the surrounding industry because you're probably not employed anymore. So those were the two pathways that when we looked at how AI can fail, we were thinking, okay, it could either be a clash of symbols saying that or an explosion. The reality as we've gotten into AI at maturity or AI actually starting to develop scale is overwhelmingly it is the mouse. Yes, the wrecking robot, the big calamitous disaster which everyone gets a chuckle when they see this model failed disastrously and publicly. Those don't happen as much, not nearly as much. It's just nothing happening at all. And when you think about the opportunity cost, it's potentially greater than having a bit of a flame out. Now I'm not saying if you had to choose between these two, go for the wrecking robot. I'm saying you don't have to choose between these two. You can choose the middle path. But overwhelmingly, when people fail at AI, they're failing because they don't have big enough horizon. They're not putting the preparation in versus going too far too quickly, which is a disaster. But it's the mouse that's more prevalent. So when we talk about reasons for this sort of nothing happening, not really getting off the ground, stuck in committee, we're talking about it, there's a bunch of stats, and you're going see some more stats in this as we go along. But eighty to ninety five percent of corporate AI pilots fail to deliver measurable ROI. Misalignment, poor planning, lack of initiative. Again, it's that we didn't actually figure out what we were doing. We just started tinkering. And as a result, we didn't get anywhere. And there's a couple different things. There's going to be a lot of ways I'm going show you that people sort of steer themselves in the wrong direction. The first is the tools first trap. And this has been true ever since technology has been something businesses have bought. They buy a tool and they think, we now have analytics or we now have data. We now have governance. The latest version of this, we have AI, we have a tool. You don't have a tool. I mean, you don't have AI, you do have a tool, but the AI tool has to be used, adopted, implemented, solving problems just like all the other tools before before you actually have anything measurable. Software does not equal value. Usage. Solutions equals value. Lack of strategic alignment. This has definitely been a problem for at least five years. A CEO, one of your c level people, one of your directors will go to a conference. They'll talk to a vendor and they'll come back. We're doing AI. Okay. What do you what what what do you wanna solve with it? Don't care. Do AI. Okay. Well, alright. That's hype. That's a buzzword. AI is not a destination. It's not like if we sail in that direction long enough, we'll arrive at artificial intelligence land, and we'll get to ride the roller coasters that have cotton candy. AI is a tool. And just like all the other tools, we have to figure out what we're gonna it for. If I have a hammer in my tool belt does not mean I've successfully become a carpenter. It means I have a tool to do carpentry. We have to make sure we have a strategic alignment versus treating this as a buzzword. Garbage in garbage out. Again, this is something I've been talking about for years. Your AI is only as good as your data. In fact, the very last webinar we spent a lot of time talking about how the semantic layer is critical. It is the king. And you could even take it a step further. Context is king. I came from an online marketing background, we always say content is king because search engine marketing, Google, and all the natural algorithmic searches was how you're going to get all your business. In the new world, this is true in search, it's true in data, it's true in everything AI related, context is king. And if you don't have good context, which means you probably don't have good semantics, which means you probably don't have good data underneath it, all of those things snowball. You're not going to have good AI. Ignoring the human element. Again, AI does things. Sometimes it needs humans to do it, sometimes it doesn't, but it needs humans for framing of what it's going to do and why and how it's going to be valuable. So people determine how AI is successful because they're the ones figuring out what AI is going to do. They're the ones that are going to contextualize it. They're the ones that are going to train it and they're the ones that are going to be the recipients of it. So if we don't bring people along on the journey, then it's not gonna be successful. It's in fact impossible for it to be successful. The other thing I wanna talk about with I guess these are these are outcomes of having a little bit of a, we'll put AI over here in the corner and we'll have some people play around with it. There there are reasons, that you might have for doing that. But being reactive when it comes to AI adoption has its own problems. You can hit the mouse of meh, by underselling or underplanning or under delivering what you're trying to do. But I wanna go back to the root cause of why you guys may those folks that did get the mouse there, why they did more of a reactive approach versus, hey. We're gonna make this a central strategic initiative, and we're gonna be bringing in several c levels. We're going to hire AI specific people to lead it and really make this into something that's central because that's honestly the level of intent and level of scrutiny it requires. So if you are reactive, it's obviously there's competitive displacement. If remember on those quotes we had on the very first slide, everybody was saying if you are not investing now, other people are and you are not going to get the benefits of it. That 10x, 15x, 20x multiplier of capability that other companies are either realizing now in certain sectors of their business and will realize across the business in very short time or very quickly will. Talent drift. Every single person, that is related to technology is thinking about AI. They're either worried it's going to replace them and desperate to try to get across it, ahead of it, and into it. I have a lot of children and they all love computers and I'm telling all of them, you're not going to be software developers. You have to go into the AI. You have to go into the data side of things. Even though they're really good at programming although it's not mine, Roblox. Minecraft is a generation older. Even though they're really good at doing all that stuff, that's the stuff they want do. I'm like, there's not going to be a job for that. I can't confidently tell you to go invest your time in that. If you do not have AI forward thinking business you're going to lose people. You're going to lose talented people because everybody knows that's the future and they want to be on the front of it. It's exciting and they want to future proof their career. And of course there's the compounding debt. Think about this as lost opportunity cost. If I told you you had a hundred million dollars that you could invest in the bank and you did that on day one, you'd be getting benefit from that and it would be compounding as more and more interest was added to it. But if you waited five years before you put that, you had it buried in your yard and then you put it into the bank, you're not only going to not get that money making earning for you, but the compounding interest just like AI, just all of these accelerators adds up. So this is a little fun slide. The gap between enthusiasm execution when it comes to AI. What do organizations say versus what do they actually do? So organizations say AI is our strategic number one priority. But in reality, it's pilots, it's demos, there's testing. But there's not this is a critical use case for us. If we solve gives us this return that's measurable. It is the most difficult thing we deal with and we're going to go solve it right off the bat. Another thing they say is we're investing heavily in AI capabilities. Shadow AI spreading all over the place without governance. We're going to talk a lot about shadow AI this presentation. So we'll go into exactly what that means. But it means that the organization is doing it, not strategy. It means the tail is wagging the dog. We have a data driven culture. No clear ownership, no COE, no framework. Even after self-service analytics has been such a pervasive thing across our industry, there's still people that do not have a formulated or rather a structured center of excellence or a focused and nurtured, watered community of practice. And there are folks that have either half written or potentially not written at all governance frameworks. Some people think buying the tool again gives them governance and there is more to that than just we've put these rules into this tool. We are moving fast on this leadership waiting for someone else to do something. Again, everybody is busy, particularly as you go up the organization, but somebody has to say, know, this is transformative and we're going to jump on this. Which gives me question that I'm going to ask you. And secretly there's a follow-up question that kind of answers it for you. Here's the question. If we're not doing anything, if we've decided to hold off on making a decision, which in itself is a decision, will our community of users wait until leadership is ready before they start doing AI stuff? Here's my follow-up question. When have our community of users ever waited when something cool has come before they start playing with it? So there's a key takeaway here. We're way past the point on these w questions. Whether we should do it, when should we do it, why should we do it, who should do it, all that stuff. It's really the only question that remains that matters is how. How are we going to do this? It is unavoidable. The longer we wait, the further behind we're going to be. So whether deliberately or not, everyone, every single one of your organizations, my organization, everyone that's not on this call, everyone is making a decision on how they're going to approach AI. Some are sprinting forward, making mistakes and learning as they go. Some are taking baby steps and they'll probably end up at the the mouse of meh and some are waiting. All of these are decisions but I will assure you of this regardless of what the organizational positions is everyone is using AI today. I think there was a stat in a presentation I did last year, something like eighty five to eighty seven percent of employees use AI. Most of those tools are not the enterprise tool. That number is undoubtedly higher twelve months later. Everyone is using it, which brings us to this point of shadow AI. So just like shadow IT, the idea here is is that people are going to find the value that AI promises, whether you give them the tools or not, or if you give them a small amount of credits and they run out of it, they have tools that they're going to bring in. When your governance lags behind the tools, meaning we're not gonna let anybody do anything until we figure out exactly how we wanna control it, the tools are already there. You have to move with agility because there's such promise of value. People aren't going to wait even if you specifically tell them not to. It's in their interest not to. It's in their interest to go forward and find their version of the whether it's Gemini or ChatGPT or whatever. The value is personal. Like I said, sometimes it's a two x, five x, in other instances it is a thirty x multiplier for efficiency. I use AI in my role. We've got people that develop stuff for us. Our software engineers at IntoWorks and it is a 30x accelerator for them. It is incredible. And if we told them that they couldn't use it until we found out exactly what we wanted to do with it, they would leave. I'm positive of it. And we would suffer the consequences because we'd be doing things thirty times longer if we were trying to develop new tools to give to you guys. Failures are invisible. When when something bad happens with a shadow AI tool, meaning I put the wrong data and make a bad decision or I put data that gets compromised, it's invisible for the most part. All of those little bad things don't announce themselves. They don't end up in a register or or an error log. They're happening in secret and behind closed doors and that is what's that's probably the thing that keeps your eye your your legal folks when it comes to AI usage up at night. And they can say no one's using it until we figure out how to govern it, but it's inevitable and it's impossible. And here's some stats to prove it. The vast majority of AI usage at the enterprise level is in shadow AI tools. These are surveys and research done. Ninety percent up to ninety percent of enterprise AI activity operates outside of the control and visibility of centralized teams. I think it's like eighty to ninety percent. So there's a range there, but it's overwhelmingly the majority. IT teams are typically only aware about four or five enterprise AI tools, but on average, there are fourteen of them in an organization. Seventy eight percent of employees report bringing their own AI tools, BYO AI, to the workplace to increase efficiency often intentionally because they know they're not supposed to bypassing restricted networks or policies. And one third, only one third of AI tool interactions happen to approve enterprise accounts because organizations have been slow to move on this, individuals are not. So the risks of shadow AI, obviously, you can lose data. You can put data in the wrong places. Data can get leaked, compromised. There is legal and regulatory liability if you don't do AI correctly or the outputs are used in an improper way or because you're not governing it as closely maybe there's PII information that gets put into a model. Obviously there's a capability fragmentation. You've got probably a whole bunch of agents throughout your business that some business user has created. There's no documentation. These agents might be great. They might be performing well. They might be well governed. But because there's no centralization or if that person leaves the business, all of that IP goes with them. And of course, all of these things run off of data platforms that that almost all use credits or tokens consumption. And those can explode because maybe you've got models, agents running all over the place and some of these are probably doing the same thing. There's a bunch of risks. So if we think about all of the different sort of challenges to AI that we've just talked about. So it's being reactionary. It's lack of having a plan when you plan to fail. No, when you fail to plan, you plan to fail. One of those little word plays. It's the extra challenge of governance. And of course, all of those things are represented in shadow AI and more. Big challenges. But as we said at the very start of this, which is why I like to define the opportunity, massive, massive upside. So it's worth navigating the challenges. So let's look forward towards success. We're going to talk about five things. And again, I could build a webinar on one of these bullet points and fill you up for ninety minutes. So we're gonna have to go a little bit light on these topics to try to get through them. I tried to put these in somewhat of a logical order. You could probably tackle these in a different state if you really wanted to. But I think building the AICUE, I think is this is a foundational strategy component. And from your AICUE, you'll have the living strategy that will combine data strategy and governance into an AI strategy and path forward that will be organic and will be self sustaining. Selecting your tools, assessing your current state, and I'm not really talking about culture, I'm really talking about your data a bit more here. How you start to govern agentic agents, me say that way, agentic AI, which are agents. And undoubtedly, you've already got them in your organization. I'll talk more about that when we get to this section. And then how do you build the AI culture? How do you build the capabilities within your user community? So let's start with the AI center of excellence. I think just like an analytics center of excellence or data, there's four tiers here. And they look very similar because, again, AI is growing out of analytics and data. But now we're having AI consuming all of these things versus just human beings. So it starts with executive leadership. And I think there's it's more important that executives understand what AI is and buy into it and understand the risks, the challenges, well as the upside than it is for them to understand, say, what analytics could and could not be. Because there's so much more value but so much more opportunity for things to get yucky. So multiple folks across your C level and probably a new C level role dedicated to AI or preferably a CD AI, a chief officer of data and AI as a combined portfolio would be very, very, very useful. It's unrealistic, I think, for your executive leadership to craft an AI strategy. They don't have, for the most part, they probably don't have the technical acumen and maybe being able to see over the next till of what's coming as somebody that is an SME in this space. But they do need to approve it, fund it, and give mandate and accountability and some autonomy to the folks that are going to go build that strategy and then execute it. So there has to be a partnership here. You give me a good plan, I will endorse that plan, I will give you money and opportunity to go do it. The people that are going be building that plan is your Center of Excellence Committee. And this operated just very similar, obviously with potentially a larger remit and more faces that might have to come into this as an analytics center of excellence. But this is the operational core of the AI program. Strategy, standards, platforms, tooling, governance policy, how all of these intersect with the things related to them. They're the hub and a hub and spoke model. Measurement, evaluating use cases, putting priority on those use cases, all of that sort of stuff. And then again, how they operate with legal and security and cyber and all those other all the other parts of your organization that have very, very strong stakeholder interests, they have to be a center point, a communication organizer and message, the evangelist of this stuff. Now, tier three is a little bit specialized. If you're smaller, you can probably bring the subcounsel stuff into your COE committee. If you are larger, I would recommend you actually spend the time getting out your subcommittees. That would be things like architecture or governance. I have some examples on the very next slide of a sample AI COE. You'll see some more stuff there. But these are the folks that have genuine specialist knowledge and they need to have very deep conversations that are probably outside of the scope of just a general COE conversation. Those folks need to be represented, their points of view and their specialty and expertise need to be represented in the COE committee. But they probably would be best served to nerd out in their sub console area and then come back and say, this is what we think. This is our position. This is our decision on that particular part of the strategy. Then we come down to the user community. And again, the top three are great, but if no one's using these, then we're not really getting maximum ROI. But when we get down to the user cohorts, there are the builders, the leaders, the practitioners. The builders are the people that are building and implementing the solutions. People roles, the leaders of the folks that are managing adoption and you'll see a clearer way of breaking all of this down on our very next slide. And then the practitioners, these are the folks whose roles and processes are being reshaped. Maybe they're being maybe you're able to repurpose those folks to other things, like partnership or innovation as sort of the manual parts of their jobs are automated with AI, or maybe they're being accelerated. I think it's important to look at your user community as a series of cohorts that have different goals and different expectations, and you're going to empower them and train them and work with them in different ways. And if you don't, you're going to miss the opportunity to speak to them more directly to get them on board and using AI in the right way and with acceleration. So let's take a look at an example. I put a RACI and I'm not here to debate which RACI should be here or there or which one is what, but just to give you a sense of how this might look. But the tier one, that's your leadership level. Again, executive sponsor, the AI program champion, the CFO, legal and risk, and then the business leaders, your GMs or your heads of those types of folks. All of these people need to be engaged here. Some of them can just be informed and consulted, some of the specific leadership like legal, they need to be very, very across how this is gonna work and how models are allowed to use data and what sorts of outputs agents are allowed to create because they can send emails, they can create documents, they can do stuff in the real world. All of these folks need to work together as that leadership tier. Just underneath that is your center of excellence. The committee of all these people bringing together. And again, you might have folks playing multiple roles just depending on your organization. As you get bigger or more complex, you'll probably find value in separating these out. So you'll have somebody that is leading the COE. In historical times, it was a director of analytics or head of analytics. It probably helps to have a little bit more of an AI bend here. You need your data and your architects, your AI folks, governance, people talking about change management and how we empower and enable, the platforms that these things are built on. All of those things are a part of that. And again, the COE level might be fairly fluid. You might bring in some folks who were tackling some big challenges or some big initiatives, and then as that thing gets done and circumstantially we don't need them as much anymore, we can let those folks sort of drift out. Next down is your sub council. So again, data governance, AI ethics and risk, your platforms and tools, enablement, your architects could be a sub council, your analytics, your data scientists. All of those folks could be specialty groups that come in and weigh in and offer positions and opinions when your COE needs SMEs to spend a little bit more time and enrich the conversation. And then underneath you've got your users. So you got your builders and you can see all the different things that they're doing. The leaders, the change leadership, the adoption. These are the team leads, department heads, project expansions, the people that their teams are going to be directly impacted. And then the process cohort, these are the folks whose processes are being automated. So you need them to do the knowledge transfer and hopefully be the beneficiary of this stuff. So the goal would be once you have your COE to architect something like this onto a whiteboard. Put it onto a confluence page or whatever for everybody to understand so everyone has accountability, everyone knows where they need to go and who they need to talk to. Visibility on when people are meeting so people can put their hand up and say, I'd like to be a part of it or I'd like to be on the agenda for this one. Transparency is very helpful here. Let's go to the next point, selecting the tools. I see a lot of really cool stuff on LinkedIn and on various social media pages that say here's an index of all the different models and tools and here's the ones that good for research and here's the one that are good right now for coding or analytics or or blah blah blah blah blah, image creation. I think those are cool, and I think they're useful. But I wanna give you some potentially deeper ways to think about how you might select the tools or tool or stack that might be best for you. So what what what most people do is they look at the biggest need that they've got, sort of the the clarifying point of why we're talking about AI specifically, and they solve for that use case and how they choose their tools. And I would say you it it may work, but you might be missing something, which is why I think it's better to think strategically. Let's think about, longer term needs and is a single vendor going to match and grow in the direction we need them to be? There's gonna be winners and there's gonna be losers. And right now, it looks like Claude's a big winner, and we'll remain that, but these these models and the market shifts so quickly. Do you wanna conversation, if not just hedging? You wanna figure out who gets to make these decisions. It it shouldn't just be IT. The business should weigh in here. The folks that are experts in data and the leadership and architecture should weigh in here. There's a financial component to this. And it's not just how much do we pay the vendor, it's how do we integrate these into other tools, what's the consumption going to look like, how does it cost, do we have other things that we have to migrate so that we could support this? Opportunity cost. Is it gonna take us a lot longer to use this tool, or is it slower to get value out of it than another tool? Is that worth us? All of those things need to be factored in. Cultural. There's probably really easy end of the spectrum here that I could use to best illustrate this. But, basically, how how self sufficient is your team if you're going to go highly technical versus something that's a little bit easier to use? So for instance, on the rapid acceleration ROI today, Snowflake is probably on that end of the spectrum. Very easy. You don't have to worry about all the little compute clusters and all that stuff. You've got automated controls on the size of your warehouse, etcetera, etcetera. You can get off and going within a few hours. And you can get all your security and stuff set up in a couple of days. On the other end, the other way, you probably have something like Google, and BigQuery, and all of the massive sort of build your own solution in the Google ecosystem. If you've got a lot of folks that like to tinker and they don't mind going through the one million Google widgets to piece things together, you can probably build something pretty bespoke. Which organization are you? Our our general position is let's get going today and let's get value today because it's opportunity cost and it's that compounding interest. But if you've got big goals and big needs and it really speaks to you to have full flexibility in tinkering and building, then maybe another tool is better for you. But you got to have that conversation. Technical. Where are you starting at from today? Is everything in SQL Server? Are you on Redshift? Are you on Teradata? Understanding of what other AI tools are you using? What's your shadow AI footprint look like. All of those things are important so that you can make a decision to get yourself value today, as well as figuring out all the financial impacts of what your technical debt might look like. Specifically for ending shadow AI, there's a bunch of stuff here. So again, you have to do a bit of an audit and look. You can look at API traffic, there's lot of different things. You can kind of get a sense of who is calling our data and where is it coming from. I would recommend running an amnesty window. So again, we don't want to punish anybody. Everyone was excited about AI and maybe we were slow in getting them what they needed. So everybody declare what you've got. No punishment. We just want to see so we can inventory and help build fast lanes so that people can use it. I think the little quote there underneath my title, I think it's a great way of thinking about this. Goals not to stop AI use, but we want to make them using the governed tools easier and more accelerated than having to go do it themselves. That way they're incentivized to do what we want them to do versus the threat of a stick. It's the carrot that we want them to be excited about. Give folks a very clear, easily read acceptable use policy, an AUP. This is how we use AI here. One page, very straightforward, very simple. The longer it is, the less people will read it. The other thing, and I'm gonna bring this other part up here as I say this, treat this as insight into your users' behaviors. If you get a sense of what ShadowAI tools there are, you automatically have a massive acceleration in terms of how people are using these tools and what solutions you can start building that you could have a high confidence is gonna have an option. Shadow AI is a signal of unmet need. It is not a problem to stop. It is an energy that you want to make constructive and controlled versus chaotic and destructive. Very, very important. These are people that want to use AI and they want to be better and more accelerated and more efficient. Help them. But do it in a way that help them in a way that is governed. We'll go through this next one pretty quick, but common traps that people fall into when they're thinking about their tools. So again, let's solve today's problem. And in eighteen to twenty four months, oh my gosh, we picked the one thing that was very unique to start with and everything else kinda is more in this area. Man, I wish we'd picked a different tool. Have a little bit of a pause and think about, hey, what are our twenty use cases that we're gonna probably go solve? And how do they align to the tool that we're selecting? Don't let a single department drive the decision. This is often an outcropping of the one we just talked about. Finance has this one thing they want to solve and because they've got the money and the drive and the initiative and the executive listening, they picked the tool. But holy cow, all of our AI business, all of the great stuff we're gonna do with AI is really on the product side. And the tool we picked isn't great for it. This has to be a group decision. And the sooner you get to a COE with that AI COE at all four levels, the sooner you'll have the full conversation and a and a full dialogue. Underestimating integration and maintenance cost. There are a lot of secret shadow costs in every tool related to data and analytics. You are probably gonna struggle for most vendors to tell you all of them. So that's why people like us help because we actually get to see multiple iterations and implementations all the way to maturity again and again and again. And we can be like, okay. So there's the virtual servers cost for Power BI. Nobody plans for that until you get handed a check. There are the compute costs, and you have to have controls over who has the lot to set those things up, all these virtual warehouses, because they can run over Christmas break and, oh my gosh, we come back, we got a forty five thousand dollars extra large warehouse that was running. All these little extra stuff. Don't make a decision off of a demo. Make a decision off of a proof of value. Demos are engineered to be successful. They are cookie cutter, very small digestible snapshots of the tool. And that's good. It's valuable. You wanna understand the value proposition of the tool in a forty five minute demo. But you need to actually start playing with the tool with your data and some of your use cases before you make your big investment. So if you are doing an evaluation cycle, plan for thirty to sixty days of some of your smartest people with biggest data sets of challenge to go and solve some things and play and practice and then come back afterwards and do a big conversation. How did it perform? How did yours perform? Which one's better? Plan that in and assign costs so people can actually make can get hands on with the tool so when you make a big investment, you're doing so with a lot of confidence. And then, of course, ignoring the people part of this. We've mentioned this a couple times. Buying a tool does not mean you have AI. People are intrinsically important to the success of AI. So you have to understand the talent and skills you currently have and what you are missing and you need to bring in. Your current state assessment. I mentioned in the very last webinar just how important the semantic layer is. All of these things get powered off of the semantic layer without your context, which is a combination of some other things and your semantic layer coming from your data. AI has no idea what it's doing. If I ask AI a question, tell me who the best soccer player of all time is, it will give me an answer even if it doesn't know it. Or if I say it's the best I think the best soccer player is person from Morocco. It'll like, you're right, here he is. Doesn't mean it's right. AI is designed to give you an answer, not say, don't know. And so for that reason, you have to make sure AI has the best information. So there's a bunch of I'll throw these both up at the same time. There's a bunch of questions I think you should be doing as you get into AI. There are a lot of businesses that tell me, well, our semantic layer is actually really good. We spent a lot of time and energy on it over the last five years. So I think we're ready to start thinking about context and AI at scale. I'm like, great. Where does your semantic layer sit? And they're like, well, I don't know. What do you mean you don't know? Well, we have reports and we know the reports are really good. Okay. Well, let's go look at what the reports are looking at. Some of them are from your data warehouse, from nice little map tables. Great. Some of them are pulling in data from one off Excel spreadsheet sitting on SharePoint. It's not good. Some of this business logic is sitting as custom SQL in your Tableau workbook. Maybe three, four hundred lines of it. That's not good. Some of them are unique calculations for profitability inside of the worksheet, that's not good. That does not mean a semantic layer. But because that worked for AI or rather because that worked for reporting and analytics, people thought, well, I have a good semantic layer because my reports are automated and they're accurate, but it's dispersed. For AI to work, we have to collect that semantic layer in one spot. And that spot needs to be discoverable. So if you've got it locked in a tool, which is some of the reasons I don't like some of these AI, the trends that analytics tools are going, you can have all these agents that they provide you, but it's in a proprietary language, it's in a little bit of a black box, You're then mapping your future to that individual company's ability to create great agents. When in fact, if you had something that was a bit more accessible, like say the iceberg tables and snowflake or whatever, then you could point whatever model you wanted at it. And as the rest of the marketplace completes, you can say, great. I'll have some Claude. Oh, now, like I said, I'll have some chat GPT. So the semantic layer has to be sort of its own little entity that you can then hang all this other stuff. On the context layer is part semantic layer, part a bunch of other stuff. So you can kind of get a sense of this right here. Now I've got a white paper coming out summarizing the last webinar that we did that will come out probably in the next week or so. So if that's something you're interested in, reach out, we'll share it with you and it goes in a lot more detail than we can do in the last webinar or this one. But all of this stuff is, do we have good data? So if you think about the Mittitan architecture, gold, silver, bronze, all the way up. Do we have a a platinum layer that would represent the semantic layer? And then do we have the context around it? All of the lineage, all of the classification, all the metadata, all the mapping so that AI can look at it and make sense of it. And then on top of that, it's all the information that you might have in the larger collection of information that you have, so in Box or SharePoint or whatever. All of that stuff is things AI can use and could if it was well curated. This is a little simple way of thinking about it. So you've got your semantic layer and things sort of help your business glossary data dictionary, metric definitions, the semantic model itself, bringing in lineage and freshness and all of the index to dial the information management data quality scores, all of those things come in together to help power your AI model. And then you could see if it's not ready, if it's partially ready, if it's AI ready, sort of how you might measure that. So AI ready, governed semantic layer, full lineage, trusted, documented, contextualized data. We can point an AI model on that with supervision, but have confidence that we're going to get usefulness out of it. At the other end, we are hoping and hope is not a plan. You're going to have hallucinations. You're going to have values or results that aren't going to be supported and potentially risky. AgenTic AI governance is another big challenge. We had a roundtable in Melbourne and Perth and we're specifically talking about AI. And rather than people saying I have five thousand dashboards, ten thousand dashboards, man, it's such a problem. In the context of AI, people are saying, we have fifteen hundred agents. And I don't really know what any of them do because they're just popping up all over the place. We have to work because agents can do stuff, I. E, what happens when AI acts without asking? They can do stuff. They can send emails. They can change files. They can do all sorts of stuff. There's a lot of risk here. Way more risk than a dashboard with bad data. A human being might make a bad decision off of it or the human being might say, I know that data looks wrong, and they may contextualize it. Agents act. The risk is continuous because AI doesn't sleep. The other thing that's probably worth thinking about is government, state, federal, national, international, whatever, it is catching up. And so depending on your business, your vertical, whether you're foreign owned or whatever, there may be a significant need for you guys to regulate and pull this in as quick as possible because there might be penalties if you don't. So then you start to think about, well, if there's autonomous action, if robots are out there doing stuff, who's responsible for it? Who owns the agent? Who owns the data that it's using? Who gave it the scope of action that it's allowed to do X, Y, and Z and not this thing over here? What happens and who's responsible if it does what it's not supposed to do or if it does something outside of what it's predicted to do? And who can turn it off? All of these things need to be considered and documented. Here is a framework that might help you. Starting from the top down, this is sort of your base. So from a compliance layer, legal decides, hey, for our business and our data and the way we think about customers, these are the parameters that we have to put in place. Then the folks that are providing the data come in next. This is the gold plus layer. This is the platinum layer. This is the diamond layer that these models are gonna draw off of. And we've got one for this one. We're not ready for that one. Then the people that actually set up the platform that might be Databricks, might be Snowflake where these agents are being built. They say, okay, you're using our framework, you're using our tools to build these agents, we're then next in line in terms of the things that we own. And then who actually builds the agent themselves, this is what they own, these are their controls, and these are the types of people. But you need not a department, you need a person to say, I own that agent. And if that person leaves, part of their off boarding should be who inherits this agent or do we kill it. You can't just let an agent roam around. It's not like a report sitting in a folder that no one looks at. Just generates a PDF every week. These things do stuff and you gotta be very mindful of that. Building an AI culture. I like this change management philosophy at ADKAR, which is awareness, desire, knowledge, ability, reinforcement. Because I think that kind of gets you everything that you need and allows you to structure it in a way that is pretty good. There's a lot here and I've got a couple more slides, we're buttressing up on time. But I think the thing that's important, and I've seen this a lot of times, people have big ideas. They'll be they'll they'll they'll write a great strategy document. They'll tuck it to the wall. Everybody can see it. We're done. High five. Now we can go off and do AI. Change management is like gardening. And if you haven't done any of it, you have to sow the ground. You have to plant the seeds. You have to water it and make stuff comes out of the ground. And it's not done then. Then you maintain it. You pull the weeds, trim the leaves, you pull the fruit. All of those things have to be done. Change management is an ongoing responsibility. It is not a project phase that ends. There is a significant more that you have to do at that part of the project for sure, but that doesn't mean we're done. High five, change management's over. It's the evangelism. It's the reinforcement of policy. It's the sustaining of business processes and culture. Very important. The other thing that I think is important when you think about change management is building the skills within the people that need to do the things. I'm a big believer in the seventytwentyten rule. Seventy percent is hands on, twenty percent is coaching and mentorship, and ten percent is formalized training. That is particularly true in AI because there's not seventy percent of classroom training we could probably go do for AI. All of this stuff is theory and experiential at this point. It has not disseminated down to university level yet or to classrooms. Snowflake or Databricks or whomever can pull you into a room and say, this is how you use this specific tool. This is how you use Coco or this is how you use Cortex. But in terms of the best practices, on top of that, you're gonna largely have to discover that for yourself for now while you are doing. You're gonna make mistakes, but that's how you learn even faster. Different roles need different skills. They need different focuses. You need to talk to them in a different way because the value they're gonna get out of the tool is different. One time training does not match how skills are built. If I show you this is how you drill, put together a cabinet, and then I'm like, alright. In six months, maybe I'll need you to build a cabinet. You will never be able to replicate that. You have to build cabinets to get good at that. Same thing is true with tech, with data, with AI. So if you were think about cohorts, we talked about project roles, people roles, process roles on that AI, COE framework. So the people that are doing the implementation, the people roles, these are the managers and leaders, guiding teams, getting the ideas, ensuring adoption, ensuring compliance. The people that are in the process roles that are gonna be beneficiaries of these are transferring that knowledge that they can be automated for AI. All of these things are super important. And you've gotta think of these individually from a change management perspective, individually from a skills enablement perspective too. Three things you can do in the next thirty days. Do the inventory of your semantic layer. See exactly what you've got nicely built out in DBT or in a snowflake view or wherever it might be. Assess just where you are in terms of is it bronze, silver, gold, gold plus platinum? Like how good is it? And what tables would we say is ready today? Think about your use cases over the next two years. And then pressure test to see, okay, with this tool that we're thinking about Solvit, with these resources and skills that we have be able to accomplish it, what are we missing? What do we need to grow into? And again, remember, vendor demos are meant to be successful. You gotta go a step beyond that and make sure you do actual proof of concepts. One, you'll learn a lot faster. And two, you'll actually understand the tool a lot better. Pick a cohort, pick a use case, and go. Do the seventy, twenty, ten on them. It might say, let's say it's finance. They were the ones that wanted to pick the tool. We stopped. We all picked it together. Now we've got the use case. We've got the tool. Now let's go accelerate. Now that the thing of testing and tearing it down at the end, we actually wanna solve a problem, throw it into production so that we can win the AI battle hill by hill, use case by use case. The good news is, is we can help. There are a lot of stuff that we do, to help you get started, to start to frame the opportunity, to start to identify the problems and come up with solutions. Probably the best way to start is with a very simple exercise that we call a DART, data and analytics review and tactics. And when we look at five different domains, which would be culture analytics, data, governance, and infrastructure, we can then assess your AI readiness. And we wouldn't if we were to then tack that onto a bit of an assessment on your semantic layer, where is your data? Where's your business logic? What does your governance look like? Then you can get a really, really, good view of what to do next. And we can help you in terms of setting up that AI COE. Here's a little bit more about that. And if you wanted to reach out to us, you certainly can respond back to any email that you got to us. So you can scan this and, schedule a time to talk to us. We've been doing this a lot. We are very excited about what AI does. We're very excited to help people do it. We're on our own discovery journey, for Interworks as well, so we're learning as we're doing. We'd love to help you too. Thank you for jumping on. We went right up to time, so I apologize. It didn't seem like there are any questions in the chat anyway. The recording will be out in about a week. And if there's anything that you need from us, certainly reach out. We're very, very excited to help. Doctor. Halena Gazelka: