Creating an Enterprise Data Governance Strategy for the AI Era

Good morning, good afternoon, and good evening to everyone. I'm coming to you from just south of San Francisco here in California, and I'm actually in the center of Silicon Valley here. It's pretty early. It's 7 AM in the morning. It looks like it's going to be a beautiful day, so I'm very happy to be with you from my home office here. In Silicon Valley to talk to you about what's becoming an urgent and high priority subject. And there's lots of demand for jobs and opportunities. So hopefully I inspire you a little bit today in this topic. Hopefully you walk away with a basic understanding of what data governance is and how it relates to AI and the future. Just a little bit about myself, not too much. As Shilpa so beautifully said, I run a small tech company here in Silicon Valley called Human Future. We advise and consult clients all over the world about the future of technology. But I'm also a professor at the University of San Francisco, at the College of Menlo College, and a university called Pepperdine University down in Southern California. I get to lecture and teach at schools all over the world, which is a great privilege for me. And finally, I'm an author. I've written 8 books, of which I'll talk about one of them today at the very end, very briefly. A couple of them were bestsellers, as you can imagine, largely focused on, on technology. But you can learn a lot about me online. I make a lot of content available for free. I really believe in sharing And empowering people to live their best lives. So, please do check out all my free material that's easily accessible. So again, thrilled to be here and to talk about data. Data. Such an important topic. It is the raw material of the digital age, right? We in fact find ourselves today in 2026 in a data revolution. I mean, you've heard of a kilobyte, right? You remember KB is like 1,000 bytes. And you might have even heard of 1,000 kilobytes, which are a megabyte. And if you were in my classroom now, I'd be having you guess what the rest were. But we have gigabytes and we have terabytes. These are numbers which define volume, quantity of units of data that we capture and store. And you've heard of petabytes, and then there's exabytes. Each one of these is 1,000 times bigger than the last. And now we find ourselves using the word zettabyte. By the way, that's a 1 with 21 zeros. A 1 with 21 zeros. And it's a very big number. In other words, we are in the zettabyte era. Maybe you never thought about it like that before, but because we are handling, processing, storing, you know, archiving, using data at such great volume, you know, variety, and velocity. We call it the zettabyte era. And just how much data do we have today? Well, here's a relatively recent chart. These are estimates. The numbers you see here are in zettabytes, of course, right? So just in 2020, the world had created around, was creating about 64 zettabytes of data. And then you can see as it grows there to 79, 97, and last year about 181 zettabytes of data in the world. And this is continuing. We think it'll double every 2 to 3 years. So, um, it's well on its way into the 200s and soon into the 300s of zettabytes. What I mean, but why? Why are we having this data revolution? Well, we're a hyper-connected digital world. Even right now, you sitting in front of your phone, in front of YouTube, if you're watching this on record, if you're watching this live with GoToMeeting, we're creating data just by sitting here, right? We're creating a lot of data. It's happening over in the cloud. It's happening on your computers, on your devices, every time you use your phone, every time we transact with a bank, every time we have an interaction with a digital resource, we're creating data. It's happening obviously many, many times every day. Again, in a way and a volume that there's no precedent. We can't look back and say, well, what was this like a few decades ago? We We don't have that. All we know is sort of we're having breakthroughs every single day now on the quantity. And then the question is, well, what can we do with it? How do we manage it? I mean, isn't that the core question? It's— there's this Cambrian explosion. Well, it's very useful. Data is useful. I think we'd all agree. And, you know, it helps us make better decisions. Think about Any decision you made today, if you're watching this in the evening in India or in the morning in the United States, the decisions you might make today or yesterday, it was informed by data, right? And the better data you had, the better quality, the better your decision was. We're able to tell stories and find issues in our organizations and then hopefully tackle them as we gather more evidence, solving problems. After all, as leaders, that's what we do. We, we try to solve problems and make our businesses run better, make the world run better. But what we do know for a fact is this AI era that we find ourselves in, the intelligence age, one of the core reasons why it's happened now rather than in the past is we have so much data, and data becomes the raw material that is powering this intelligence age, that makes generative AI work and different types of AI, right? We had the software, we had the, we had the deep learning and the neural nets, you know, for a while, but something happened in the last few years. What was that? Access to massive volumes of different types of data that then could be trained And is then enabling the great magic, the magic of artificial intelligence. So data and AI, you know, they're connected at the hip. They go together. Okay, so we— I've sort of made this quick case that data is important. There's a lot of it. It kind of runs our world today. It helps us make decisions. It helps us run our businesses, our cities, our airports, our factories. It helps us deliver the world in the 21st century. And there's so much of it. The question that we have to ask ourselves as employees, as staff, as leaders of all these types of different organizations is, are we actually managing it? You know, are we— do we have oversight for the data we have? Do we know what we have? I mean, these are the important questions. It's not just enough to have data. After all, the data could be bad. It could be wrong. There could be gaps. There could be lots of issues. Maybe we're not taking advantage of it. Maybe it's feeding AI models with really bad data and that's having bad consequences. So at the core, at the very core of this topic is the question, is your organization's data being managed? Ask yourself that. Maybe when you go to work tomorrow, ask, ask that question. Are we really good at managing data? So enterprise data governance as a definition, if we're to embrace this topic, it's about managing the value, right? All the things that data can do for us and also managing the risks of data. Well, you've heard of the risks, right? Cybersecurity, right? Hacks happen every day. Data gets stolen. People's personal data gets stolen. There's lots of fraud and crime that happens online if we don't protect our data. There's privacy questions, right? There's compliance issues. You know, today we live in a regulated compliance business world. And if we don't, well, we get slapped on the wrist. We get fined. Sometimes people can go to prison, right? It can be really bad. Are we managing the risks related to it? And of course, where most people are most interested, especially folks that are driving business value and trying to grow businesses. Perhaps you're an entrepreneur and you're trying to build your startup. You're interested in the value of data. So we look at both of those, the value and the risk, right? So what's in scope when we talk about value and risk? Yeah, it seems— I think maybe it seems okay. We can kind of get our head wrapped around it. But let's go a little bit deeper. Like, what does that actually mean? What does it mean? So, well, some of you who have studied or will study cybersecurity know that there's sort of 3 core elements to it. We talk about CIA, right? Confidentiality, integrity, and availability of data. So is it available? Is the data that's valuable to you at a point in time available? Right data at the right time to the right person? Not always, right? And can it be used? Is it connectable via an API or some sort of integration? Or can we rely on it? Is it secure? Do we have standards, right? Do we have standards? How many ways can you, for example, enter the name of a state into a form online? Right? Is it here in, here in the United States? You know, if I take something like Florida, right? The state of Florida. How many ways can we say Florida? Well, we can say Florida. We can say FL. But we can also say FLA. That's 3 ways that we can capture the state name Florida. Perhaps we should have a standard in our system so that we aren't capturing many different ways. That's just a very lightweight example. Maybe our standard is FL, right? So every state is just 2 letters. And then do we have written policies that help guide what we're supposed to do with our data? So here's just a brief list of the ways that we can think about it. Now, these are all complicated. Every one of these needs attention. In fact, as we step back, how do we ensure value and the management of risk relative to those areas I just shared with you? Well, data governance is a big, broad, deep, complex topic. It's rich. It's a rich topic. It produces a lot of value if it's done really well. This is why there's such a massive demand right now for data professionals, because it's a complicated, very skilled type of work. And it covers things like cybersecurity. It covers things like support for artificial intelligence. It covers things like being able to have a strategy for innovation. Because after all, at the end of, end of all of that, data is the empowering component of all of it. Okay, so I've spoken a little bit about what it is, what it's made up of, and perhaps why it matters. I hope that at least at this point in my short talk, you're getting a sense of, okay, I see where Jonathan's coming from. I see what this topic is about, and I can see how organizations are increasingly making it a priority because they are. Data governance a few years ago was an afterthought. Organizations weren't really paying attention to it. And while there's a lot of work still to do, organizations finally are putting priority on this topic. And so how do they put in place a program? Or let's say you decided after being inspired by my presentation today, you wanted to be a data governance analyst or data governance manager or to have a role related to it. What might you do? Well, one of the things you might do if you went into an organization that didn't have any type of program is start to design it. And, you know, there's sort of 3 areas that need focus, right? Is first of all the strategy. If you look at my pyramid on the screen right now, what is it that we want to do and why, right? Answering the question, what are we trying to do with this program? Are we trying to actually reduce the amount of compliance issues we've been having over the last 2 years? Maybe your question is, how do we ensure that we eliminate bias in our, in our own large language models, or how do we innovate around AI, right? So strategy is important. You gotta have, you gotta have the reason and, and then the plan. Then the tactics is how do you lead it, what is the management that's involved in the policies and the guidance and the oversight, right? That's that middle tier. And then finally, at the bottom tier is basically running it, right? The operations. We have to address all 3 in our organization. Let me just say this right now. Data governance belongs to everybody in an organization, right? There is going to be a chief data officer. There is a, you know, a chief information security officer. There are, you know, compliance officers, and there may even be a data governance director or data governance manager. Those are very important roles. as it relates to the overall implementation and leadership around this. But everybody who touches a computer or data, which is everybody today, right? I think you'd agree, right? Has a role. Everybody has a role. They have to know what their role is. So again, going down a little deeper into how you structure this. Well, we got to make decisions. We got to make decisions like what policies are we going to have? Like, how— and how do— how will those policies be designed to get the results that we want? If our goal again is to be very progressive on artificial intelligence, first of all, to have good quality, you know, internal models to be able to deliver internal processes and deliver new services into the marketplace, we've got to decide how we do that, how we use data, how we prepare ensure that it's good quality data. So, we have our council here. We have some working groups. And then, you can see how they— how it drives policies and processes in an organization. So, what I thought I'd like— what I thought I'd do with the— in the short amount of time I have is sort of give some really actionable advice here. Like, you've decided to spend an hour with me, which is great. Thank you. I'm very honored. And I don't want you to walk away and say, well, that was a wasted hour. I want you to walk away with real skills. So first of all, we've defined enterprise data governance. We've talked about a little bit of the structure that it applies to everybody. Now let's go into even more detail. So data itself, every organization today, almost every organization is a technology business. Think about that for a second. You probably are watching this and you represent all types of industries and organizations and entity types. You might even be a student. I mean, you come from all sorts of backgrounds, and depending on the organization you have, I'm telling you, you use data at some point, right? It would be hard. I would— I think all of us would be hard-pressed to find an organization today where there's no data or it's not really used that much. Okay, and so there are many sets of data like, uh, product information, customer information, market information, financials, right? Wow, there's just, just lots and lots. There's, there's, uh, as we know in the world, zettabytes of this stuff, right? Zettabytes of data. Now let's go into your organization. You're You're one business. Does all the data need a data owner, right? What is a data owner? Okay, I actually think this is one of the most important roles organizations can have today. A data owner is somebody who is responsible for a single dataset or multiple datasets, and they come to work and they worry about those datasets every day, right? That's an interesting Sentence, right? A data owner is somebody who is responsible for a dataset and worries about it. Well, what do they worry? They worry that it's secure, right? Is it protected? Is there cybersecurity in place? Is it is privacy protected? Is it available to those that want it? Is it high quality data? Is it current, right? Think about the questions you might ask, and if it's an important dataset, somebody needs to worry. About it. But the thing is, there are datasets that don't necessarily have as much priority, and you, you've got to rank those and prioritize them appropriately. Not every dataset requires a data owner, but the ones that are important, you know, the ones that are really important, they need somebody who comes to work every day and worries about it because they're the ones who are going to be making sure that there are policies around it, there are, you know, checks In order to ensure that the right data is available at the right time for the right person. Okay, so think about whether you are a data owner or who the data owners are in your business. Now, it wouldn't surprise me, by the way, that you do a little bit of poking around, or you maybe have lunch with the chief data officer, and he or she is like, Well, I don't think we have a lot of data owners, and that's a point for discussion. That's a really good point of, well, maybe we should. Maybe we should have a small set of very motivated, talented individuals. It's not their full-time job, but it's part of a job to make sure that data is being looked after. It can make a big difference. Let me say that again. It makes a very big difference. Okay, so you have a lot of data in your business, a lot of datasets, a lot of different systems, right? Imagine all the systems you have in the cloud, all the apps you have, all the on-premises data. And we need to think about, well, okay, in order for data to have value and in order for us to protect it, we need to know what we have. We need to know where it is and how we access it. This is again all often in an organization quite difficult to answer. How does an organization know what data it has, where it is, and how it's being used? Well, we have an answer for that. In enterprise data governance, we have a product, a tool called a data catalog that, if done well, has the answers to these questions, right? And it looks a little bit like this. The data catalog is typically a system, right? There's lots of good providers out there. Data catalogs usually sits, you know, in, in the cloud. It's software as a service. Lots of great vendors doing great work there. It ingests across your enterprise different data sources and it collects metadata, right? Data about data. The questions we want to get answered: where is it? What is it? What's in it? Who owns it? Who can access it? You know, all It should have all that metadata, and it's constantly updated as new datasets get introduced, as some datasets get deleted, archived, et cetera. And then the right end users can access that data catalog just like doing a search. They can type in, you know what, where is the— I would like the current list of customers in— well, just the current list of customers. Let's keep it simple, right? Where do I find that? Well, you could ask your colleagues. That just seems kind of inefficient, like, where is it? Or you could go to the data catalog and type in that search query. You know, the query is, what dataset are you looking for? You just type in current employees. And it then gives you a, uh, produces the, uh, uh, you know, the, the results that has all the metadata, and then, and then you can, you can, uh, start to use it. Well, you know, maybe you don't know how to use it, maybe you don't have access to— hopefully has a phone number, a phone number or an email address that points you to the data owner, right? Because the data owner is going to know if you can use it and how you can access it and all that good stuff, right? Okay, 2 elements of it. Now we keep going. And we want to go a little bit deeper into sort of how do we make it real? You know, okay, yeah, we can assign people to positions. We can give people responsibilities. We can put in place things like standards, data owners. We can do the things we've spoken about, but we want different behaviors based on different datasets. And what I mean by that is Well, how do we deal with PII, you know, personally identifiable information? Personally identifiable information, PII, right? That stuff that, you know, that's a particular category of data, right? That stuff that maybe has a person's name in it, maybe their address, their phone number, maybe their— here in the US— Social Security number or whatever is equivalent in your country. And we want to have rules about how do you handle that. That's a particular type of data. How do we handle that data? Well, we need to have policies, right? We need to establish as an organization data policies. And these are documents. Obviously, they're documented. Guides and procedures when handling. They should be— the policies are agreed by the data governance council that I showed you previously, that hierarchy. They are owned often by data owners, sometimes by another role called the data steward. The data steward is a very active role here, makes things happen in the data governance world. It's not— it's sometimes in a big organization, it's a full-time role. In small organizations, it's part of somebody's role. It can also overlap with a data owner for sure. And so, we want to make sure these policies are current, you know, they make sense, they're approved, and people are trained on them. People know about them, right? That's only fair if somebody is going to be held accountable for a policy, they ought to know what the policy is. Okay, so what I want to share now is, uh, again, at a very high level, um, how, um, how do we make sure the program runs and, and, and ultimately is successful? Because we don't want to waste our time, right? The worst thing we can do in the world is a program that has no value, right? We're building this because it has enormous value, right, for delivering, leveraging the, the great capabilities of data, but also managing the risks around data. So I wanted to share with you 5 essentials for maintaining the enterprise data governance program. Um, well, we start with just high-quality documentation. And again, you know, Well, it's worth saying it here. I don't know if I said it already, but if I didn't, here it is. If I did, sorry, I'm going to say it again. This is not, you know, quantity is not the measure here. Quality is, right? We don't want to have hundreds of policies that, and each policy is 20 pages long. Because, you know, the truth is, you might have seen this, you might be smiling as I say this, you might have seen this, but it's people aren't going to know about it. They're not going to follow it. It's not going to work. You'll have created a lot of documentation, but it won't be effective. Good quality documentation is current. It's really clear and it is actionable. It's applicable, right? Everything else should be, for the most part, archived. It's not important. So, Having high-quality documentation here is really key. Putting in place a process for change. And there's lots of ways you can think about this. Change could be how do you go from a world where enterprise data governance doesn't exist to where you want it to exist and you want to have a mature program. So that's part of change management. And then of course, all the— everything that happens within that. But change management too, in terms of how do we onboard a new dataset for a new application in a way that is seamless and it gets the results and outcomes you expect. Anything related to change and data, maybe even anything related to data and its consequences like artificial intelligence. One of the things that happens a lot with data in organization is there's problems, right? Maybe you can't connect to a dataset, or perhaps you access a dataset, you realize that it's old. Well, another very painful aspect of data is there are multiple sources of truth, right? You know what I'm talking about? You've probably experienced that. What Which is the current dataset, you know, or why is there 3? Which is the one that's correct, right? Oh, that's, that's a, that's version control, and that's, that's painful if you don't have that under control. So we want to log a ticket. Maybe the ticket gets logged with a data steward eventually, or a data owner, depending how you have organized, and it needs to be resolved, right? We need to resolve data issues. Yeah. And we have to have a process around that. How do people know what's happening? And by the way, the real thing I want to share relative to communications is communicate when you are being successful. Okay? This is something that is probably not lost on you in the last few minutes is there's some bureaucracy here. Sure. You know, this is sort of not the sexiest of topics for an organization. It is one of the most important, but it's one of those things you kind of have to do if we're going to get good results. And people will reluctantly come on board. It's hard to be successful at first. It comes with time. The failure rate on new data governance programs, unfortunately, is exceptionally high. So, when people say, okay, I'm prepared to invest in this. I'm prepared to go along with the plan. They want to see that it's working. They want to know that what they're doing is not wasting their time. So, communicate when you have good outcomes. Obviously, communicate changes. Communicate new policies. Have a strong communications plan. And then, I think Well, look at that. It says, report value. Report the value through your communications plan. Be frequent and consistent around that. Okay, as we kind of— as I think about where we're at, we're coming to sort of the end of the arc here, right? Which is we're doing all this. We've bought into it. We like it. Makes sense. We want to have really good artificial intelligence results, right? And outcomes. We want to have great cybersecurity. Absolutely. These are top priorities in 2026 and will be for a long time. How do we know if what we're doing is getting results? Good question, right? How do we know that what we're doing is actually achieving the strategy we have around data? Well, I hope you'll ask this question often. I think it's a question you ask at the very beginning. How are we going to measure it? And then how will we measure it? As in what mechanisms? And then finally, what are the results? What is the measurements? What are the metrics that we're getting? How do we know the enterprise data governance program is getting results? So there are many different ways to slice and dice this. I'll say that straight up, but I want to give you some ideas. I want to give you some actionable stuff that's real and you can take with you. or something you can build on? Well, the first thing, because at the heart of all of this is data quality, right? Think about that. What's the most important measure of data governance? Well, yeah, it is about more innovation. We're going to see that in a moment. It is about compliance. It is about security. But there's a big umbrella sort of metric, which is quality. Quality. Why? Well, if data is bad quality, everything after that is bad, right? If the data, for example, has big gaps, or the data is wrong, or the data is old, right? It's not usable, and it's problematic, and it can create legal issues. And it can create problems with AI, and it can create problems with privacy. So quality becomes the big topic, right? Have we improved it? And you've got to have metrics on that. You know, if there's a dataset, let's say with, with a, I don't know, I'm going to make it up, a million records in the dataset, and you notice that there are duplicates, and you say, well, we actually want to reduce duplicates. You know, right now, out of the million transactions, 8% are duplicates. And you say, well, our performance goal is to get that down to 3% duplicates. Well, you're measuring that. Where are you relative to that? You were at 8% at the beginning. Is it now at 7%? You can measure these things, right? And now you've got sort of guidance on Are things improving? You know, we want to have. We want to empower people. We want data to work for us. We want to have data work for our company. So, can the right people get it at the right time? And then there's sort of just issues like incidents that are reported. Remember through the resolution process, how many issues are are being reported about data? What extent of cybersecurity events are there? You know, data governance is a cybersecurity topic. I think you get this, right? Data governance— you're governing data means you're protecting data. You're keeping it secure when it makes sense. The right people can access it and the wrong people can't access it, right? You're getting better compliance results So after all, this was a presentation about how data governance enables success in the intelligence age, right? Every single day the topic is AI, AI, AI, right? It's all the time. It's overwhelming. Well, this is a very big deal. This is a big— this is the topic of our time, artificial intelligence. Well, let's Let me be really clear about this. Data governance is the foundation of successful artificial intelligence. Let me say that again. Data governance is at the foundation, is the reason we can have success with AI. And that's because data governance feeds into the very governance of artificial intelligence, making sure that AI produces value, making sure that AI has access to the right data, making sure that AI is safe, making sure that AI is secure, right? Having policies and procedures. Data governance is foundational to AI success. It's because data governance drives— it's the underlying components of your AI governance guidance and approach in your business. These cannot be separated. You have to have data governance underlying it, right? So everything I've shared with you in the last sort of 35 minutes or so is to get to, to get to this point, right? To get to this point. These are topics that are related. And in 2026, why do we see so many openings? Why do we see the opportunities for enterprise data governance? Because of the intelligence age because we have embraced AI and AI is defining the era in which we live. So where are we headed with this? Let's take a little dive into the future and then I'll wrap up and we'll get some questions, etc. Right? Let's look at the future. Well, you're probably a student if you're on this. There's probably— I should say that differently. There's a high likelihood that there's plenty of students watching this either live right now in September of 2026, or you're looking at a video, this is posted on the EC Council University website or on YouTube or somewhere like that, know that there are many job opportunities for students. There are many job opportunities for those who want to pivot, or you have a data role, you want to change that data role. Let me just say, if you go to one of your favorite job sites and type in data governance, data analyst, even AI governance, you know, some variation of these words, and you have to try different words. I think you're going to see a lot of openings. This is a great, well-paid, highly talented, highly skilled opportunity for lots of people around the world. When data governance is done well, it's helping organizations be more successful. I mean, you couldn't make a stronger argument for this topic. When data governance is done well, it increases the value of data, which in turn drives business value. It's essential for AI integration. When we do data governance well, we have more success with AI. We can do more interesting things with AI. We can integrate AI systems. We can build new AI systems. Above all, with that included, it can drive better services and products from our business, both, by the way, delivered externally to the marketplace and also internal. And it meets that need for leaders who say they want to have data-driven enterprises. The surveys say that 83%, 83% of leaders want to have a data-driven enterprise. How do they get there? Really good enterprise data governance. So as I come to my time here, right on schedule, I first of all want to thank you for listening. If you've made it this far and you found it interesting, I'm very humbled by that. But if you want to go deeper, learn a lot more, well, the EC Council University is going to share with you many ways to do that. But I want to also share with you my own book. I've written many. I am the writer of Data Governance for Dummies, which isn't for dummies. It's for smart people who want to be really successful at this. You can get this anywhere, any bookstore, anywhere online, and I would be so thrilled and honored if you considered purchasing that book or accessing it in your local library. Again, it's been an absolute privilege. Why do we think it'll break down? I'm not sure that I understand. Why it's going to improve, right? Yeah, I see, I see actually generative AI being an absolute advantage to the data tools we have currently available to us. So we'll be able to be asked more complicated questions, have AI do a lot more of the work. Identify sources. It also— what generative AI does in the data governance world or the data world in general is it makes it more accessible to more people. So in organizations today, they often have large data teams, and you have sort of this relationship between business users who hand over data requests Maybe for reports or otherwise to data analysts. Data analysts do the work and then they send it back perhaps to a business analyst who then delivers it to the business. And often it could have required a developer too. We have a space called DataOps, right? Which was developed in order to, in fact, improve that and create efficiencies between the business user the data group and the developers, right? So there's more efficiency. Well, what data AI does in the world of analytics is it elevates the capabilities of people closer to the problem. So the business user can have more capabilities. Now, it doesn't eliminate the data analyst or the developer, certainly not in the short to medium term in the way they roll today. They means they work on more sophisticated things. They work more collaborative collaboratively together. Of course, when we when we look out into long term, everything will you know there is going to be a job evolution, job changes. So so that that's something just we we need to prepare for and and watch closely as our as the marketplace evolves. So in many ways, if I understood the question correctly, I actually think it's not breaking down, it's improving. AI is bringing phenomenal capabilities to data governance tools, things like, you know, categorization tools. So it's automated categorization, routing of data through the organization in a more optimized, intuitive way, and flagging issues. So just a small set of possibilities. So really a very strong positive coming from AI in the short to medium term. Well thought out questions here, it looks like. Let me just process the question. Okay, that can be— I'm seeing the question here. Should data classification agents authorization? If so, what should the controls sit in data governance? Well, if I understand the question correctly, clearly as we move to an agentic world, right? Software that has a degree of autonomy, it's been empowered, right? It's given access. The guidelines, the oversight, many of the things that are in place today in a more manual world have to evolve. Sure. Okay. So let's establish that as sort of the baseline for the question. So we have to even look at this sort of on a case-by-case basis, right? To what degree does your organization empower autonomy in its agents, right? That's the first question. To what— and sort of the supporting question is, to what degree are there humans in the loop, right? All of these questions, by the way, you would think when you read media and sort of listen to vendors, we're way ahead. For many organizations, we're still in the beginning sort of phase of all of this. Most organizations are not using agents significantly yet, certainly not in very sophisticated workflows, but that's changing, of course. I think 2027 is an agentic year. The agentic organization really comes into its own. And if I was, you know, if I was working with— if I was a data leader working with, you know, the chief technology officer and maybe the chief AI officer, I'd start to look at the processes and determine where the governance needs to sit, what controls should be in place. There wouldn't be sort of a universal answer to that question. It would depend on the workflow. You know, if it's rather benign, you know, maybe just doing retrieving, you know, current price lists from vendors or something, well, that's one level of control. But another one would be where it's extremely sensitive and there needs to be a higher, a much higher degree of oversight and control. We might even decide that it's not ready for agents or we're not giving the agent You know, the login and password for a system because of the level of risk. So there is a case-by-case approach. I think there is a risk-based approach that is very much core to cybersecurity and overall, you know, business control. So looking at the end part of this question, yeah, it's, it's gonna be a mix. There's gonna be, there's gonna be organizational confidence, there's gonna be data governance, there's gonna be AI governance, and then there's gonna be risk-based approach. So all of those things compounded per workflow will get you the answer you need. Everything begins and ends with people, right? Otherwise, what's the point? Think about that, right? We're clearly initiating it because there's some value, and the ultimate delivery of that value is to humans. So, in the big picture, in the big picture, I get it. So, obviously, within a workflow, within A process. There could be systems that talk to each other, and well, there will be, and there is today, of course. But in this question, where does it begin? I mean, my company is called Human Future, so I believe humans are at the center, that the future is about humans. And, you know, if we're building for the robots and building for a robot future, I think we got a bigger problem to deal with, right? So yeah, everything starts with a vision. What do we want? What do we want our organizations to be? Why do we want our organizations to exist? Yeah, maybe the person who wrote this, they're being very clever in that there could be entirely autonomous organizations. And there are a few that exist. But when you kind of peel back the surface, you know, blockchain was one of the ways in which we were sort of looking at this notion of the autonomous organization. Well, it was initiated by humans, right? It didn't just spring out of nowhere. I can't speak to the world in, you know, 20 years, who knows. But to give this question the the credit it deserves. I believe it starts with people. People leads to data. Data leads to systems and solutions and then human outcomes. I'm impressed. Okay. What is the appropriate data retention and deletion policy for data used to train or operate AI systems? It is case by case, company by company, right? We don't generalize on things like this. No. Look, one way to think about this question is if data is used to train and the data doesn't change, it doesn't need to be retrained, right? If the dataset is changing, well, you're going to have retraining on a periodic basis. So the training part of this is a little simpler to answer. Uh, my, my, based on my experience, you know, datasets and organizations, typically ones that are important, change all the time, right? Your, your financials change, your human resources, your HRIS systems change all the time, your, um, you know, your, your marketing systems, um, you know, your information systems change because they're being used and they collect and they store, they edit, they update. Data changes all the time. So for your internal models, you're going to be retraining on a periodic basis because you're going to have new content, or at least the new content is going to be retrievable by the AI system. But because the question asks about training specifically, I would think those systems would need to have regular retraining. Okay, so yeah, if— yeah, the question is a little confused or confusing because it depends on the system and how the system uses data, right? Yeah. If it's a live system, of course, it's not going to be deleted. Okay. And then what does it prohibit? Data retention. Well, I think maybe the answer is the same. Yeah, your retention policies aren't guided by AI. They're guided by compliance requirements for your industry. Financial industry has certain compliance, and depending on the country you're coming from, Things like health data. Health data has very specific compliance retention schedules. So maybe if I kind of reflect a little bit further, I would say, you know, data retention and deletion policy is not going to be guided ultimately by AI, but by other factors. Maybe that's the better answer. I don't, I don't, uh, I, I don't take back what I said at the beginning because I still believe the retraining or the training part is still, uh, makes sense. But I think, uh, I think the latter part of my answer is, is actually the right answer. I'm not here to advocate for any solution, so I'm not gonna— I'm not going to, uh, uh, you know, identify and list a product, a vendor, right? There are many vendors that do this. And in fact, this is obviously a growth area. What we're interested in here is DLP or data leak prevention, right? There's many different forms that that takes. And you, for example, training could be one mechanism, right? If you have an organization, if it's appropriate for your organization, you can train them on what not to upload and what not to share. And that clearly is a very important part of any business, a policy around that. And that is a data governance issue. And then there are tools. Yes, there are a whole lot of excellent vendors who can look for things like patterns, can look for particular You know, types of data. And it can do everything from prompt you saying, you know, a popup that says, are you sure you want to share this? To you're not allowed to share this, to even reporting it to your security officers or compliance team. So I think there's layers to it, right? Training is essential because look, The truth is you could have really good tools, but you know, you're going to have smart people who use their home computers. You're going to have smart people who use VPN. People are going to find a way to circumvent your tools if they need to. They're also going to make mistakes. So you've got to have a combination of training, enforcement, and software and hardware, because sometimes these prevention systems, these intrusion systems, or not intrusion, but filters. I can't think of the right term right now, you know, that monitors data coming in and coming out. They can be very, very effective, but they also can also be not foolproof. Right? So I think that's the right answer for this question is you decide what the right mix is, you decide how much to invest, you decide how sensitive your data is. There's lots of questions to ask on deploying this, but there are good solutions today. Thank you very much. Well, thanks, thanks, Shilpa, again for hosting me. You know, it's been a lovely engagement over the last few weeks and months to put this together. Thrilled to see that you'll put this on your website so people can watch it later and share it. I would just end by saying, you know, there are some topics in the tech field that are maybe not as valuable today. People are really starting to question what they're, you know, what will they do in the future and what will happen to their work and their jobs. There's plenty of that going on, and rightfully so. I'm very optimistic about the future of data governance. If you liked this topic, if you like the topic of data, if you like the topic of risk management, compliance, cybersecurity, you know, value generation, this is probably a place to play. And I think there's going to be lots of opportunities. It's an exciting space if it appeals to you. I talk to organizations all the time and there's high demand for it. So that's my message is seriously look at this as a profession, take it seriously and make it a priority for your business.