Featured Session: From Pilot to Payoff: 7 Pattern-Matched Traits of AI Systems That Actually Work

Unknown Speaker 00:00 I am the Chief Business Officer for a unicorn software company, but I've been working in AI since 2013 and I've spent the last year deploying AI throughout our company. And I just share some stuff in the upper corner of the slide to let you guys know I do have credibility here. I know what I'm talking about and I've gotten my hands dirty with AI, so let's get started with a quick show of hands. How many of you have a AI project in production, Unknown Speaker 00:33 okay, how many of you have an AI product in pilots? Unknown Speaker 00:38 Okay? And how many of you use it personally. Unknown Speaker 00:42 Okay, awesome. This is great. You are the perfect audience that I wanted to talk to today, because what we're going to do is we're going to go and chat a little bit about why projects fail and why they succeed. You probably saw the MIT report that showed that 95% of AI pilots today are failing, and that means they're not producing the return on investment. And most people concluded that it's the technology that's causing the problem, and it's really not the technology. So we're going to talk today about seven essential items about why you could present your project to be very successful. Now I know if you're like me, I travel a lot. I've been to 92 countries, including Brazil, Unknown Speaker 01:34 and I always pack a suitcase, and I try to make my suitcase big enough so that I can fit everything in it, but sometimes I still miss that critical element, toothpaste or socks. And that's just like people as they're deploying AI today, sometimes they miss some very critical elements, not because they weren't thinking about it, but because that AI magic takes over. So we're going to walk through seven essential things today, and we'll walk through them here on the slide. But I know you guys love to take pictures of the slide with all seven, so I'll leave it up there for a couple of seconds, so you guys can see what we're going to be talking about and look where we are going to start. We are going to start with people. We're going to start with leadership, and talk a little bit about why leaders also impact the return on investment for an AI project. So there are three areas that we're going to look at here for leaders versus staying curious, learning, asking the right sets of questions. Secondly, we're going to talk a little bit about building trust, and we'll end with some skills. So let's start with a story. I had the privilege of being at Davos this year. This is where the World Economic Forum is held, and I did a round table with 20 CEOs. They were there to learn about artificial intelligence and how to make it ROI generating. The first question I asked was, of the 20 of you, how many of you have used AI in the last week? Guess how many three? Only three of the CEOs have used it now. Why does that make a difference in an ROI for AI? Well, here are some of the numbers that have just come out, and what it shows is that if your CEO is using AI like for prompting, you're 1.6 times more likely to be successful on that project. But if your CEO is using it for prompting and for agents, brainstorming with you, doing cross team groups with you, you're 5.2 times more likely to be successful with AI because they're talking about it. They're putting it into the culture. Now, what kind of things are they talking about? They're asking questions. A friend of mine wrote a book. It's called the creator's code, and in that creator's code, she says that the number one thing that separates the innovators, the Jeff Bezos, the Steve Jobs, the Elon Musk of the world, from everybody else is they ask a lot of questions. And the same thing is true for leadership with AI. So I am with a company that's a unicorn. The first agent we wanted to install in our company, and an agent is just a way that a code can autonomously make decisions. And we decided to ask questions. A typical question would have been, Hey, how can we automate our best of class customer service and leverage AI with it? But the better question would be, if we rebuilt that customer service function from scratch, knowing that AI is here, what would that look like? And so we asked different questions. We asked questions like, how could we become so successful that customers don't have to call us? Unknown Speaker 05:10 What if, when an agent found a problem in the code, it could just fix it, and what if all those questions that come in, the agent could wrap those up and send those off to product management so it could automatically become a feature or a function, and that's what we designed today. Our agent does. 47% of all of our questions are answered by our AI agent. We have 4.8 million customers around the world, globally, so it answers questions when my team is sleeping, for example. But the number one thing I'm super proud about is that element in green, we raised our customer sat by 4% by embedding AI and agents in what we were doing. So to change the result, you have to change the question. The questions are super important. Now let's talk about another, another scenario here around trust. So I was recently invited to a fortune 100 customer site, and the chief product officer was presenting, and he showed me his AI dashboard, and his dashboard was 100% green, and he was talking about everything that he was doing. And then the executives left the room, and I asked the team leads, wow, I'm surprised your dashboard is all green. And they said, Well, we do a couple of little work arounds, and we do some extra things that turns it green. So the team knew really what was going on. And this is a trust gap that we're seeing right now on enterprises. So the data shows, and this data is from walk me. They just did an AI survey that will be out next month. They allow me to use some of this data. And what they found was that of executives, 65% trust the result that they get from AI. The employees, only 17% because they know where all the rocks are hidden. They know that they're doing work arounds to get that dashboard to look green. And so recommendation here is that you do things together, that you have cross functional teams, that you include the executives in it, like what happened at Mass General. They did a prompt, a thon with hospital administrators, with cardiologists, with surgeons, with nurses, with the assistants who check you in, they did a massive crop, a thon so everybody could see what was happening, and then an agent, a THON, so they could build new titles. Unknown Speaker 07:51 And then the last thing for people are leadership is skills, the skills gap that exists today. So 77% of executives say that it's adoption that is their problem in the enterprise, not the tools. Unknown Speaker 08:07 And this is brand new data from walk me again that said 54% of workers last month stopped using tools because they said they didn't work and they did the work manually instead of using AI. So this enablement, making sure people are trained, making sure you have the right set of tools, is really important. And I love what MTT data has to say about this. My friend David Armando works at MTT data. He couldn't be here today, but hopefully he's watching the live stream. And what they do is they have a black belt system. So when you come into the company you get trained, you get a white belt. As you train on more tools, you get a yellow belt and a red belt and a blue belt and a black belt. And this really gamifies that enablement and that skills training, and really encourages the teams to make sure that they know the tools. So the first thing to drive strong ROI is about not evaluating the tech first, but first evaluate your team and your leadership's readiness to restructure for artificial intelligence. Okay, that's number one. Remember, going to go up to number seven. Number two is agents driving return on investment. It's not about prompting anymore. It's about having agents autonomously do stuff for you as you're making decisions. Now, how many of you guys have heard about open claw. Okay, you can't leave South by Southwest without knowing about open claw and now iron claw, and announced on Monday by Nvidia, Nemo claw as well. But before we get into those, let me give you the quote from Jetson, the CEO of Nvidia, open claw is probably the single most important release of software ever. This is what he said on stage right before he announced their Nemo claw. Now I think this is important, because everybody is going to have an agent or two or three or four that is their agent. Unknown Speaker 09:42 Unknown Speaker 10:22 You might have an outward facing agent. It might do negotiations for you. It might help you to present to a customer. You're going to have inward facing agents that help you manage your calendar, that help you do your research. And that's just two examples. There are Unknown Speaker 10:39 probably more out there. And in fact, let me show you one of mine. Unknown Speaker 10:45 This is a synthetic futurist. And the team here, Sarah and the team composed this from 500 female futurists. They use 36 different strategic frameworks for her. I got a demo just maybe this week, at South by Southwest when she came out. And what she does is she gives advice about the future, like World models. Let me show you what she looks like and how she gives advice. This is our world model, simulating global dynamics. It enables AI to predict and adapt to complex scenarios. Ask it anything. The chat is ready. So I can now sit and ask all kinds of questions and get interesting advice back from her, from a consolidation of 15 futurists, including Amy Webb, including myself, I can get my own advice as well. So all of these things I think are really interesting. Here's another agent. So I wrote this best selling book called AI, first human always, it's the first book that includes an agent. Now, why did I do that? Because, you know, I included stuff in this presentation from Monday, from Monday, because it's changing so fast. So what I do with my agent is I train it on my Forbes articles. I'm a Forbes contributor. I'm also on the research team for the digital economist. I chair the applied AI group. All that is new information. So I wanted you guys to know what I had in the book, but I did wish you to be left behind. So now I have this agent that updates the information. 2500 people are on there asking questions every day. I love answering questions from people, but there's no way I could be a chief business officer and answer all these questions. So now the agent does it for me. So those are just two examples of the type of agents that I use. And then along came open clock. Now open clock has only been in existence for six weeks, six weeks. And what is it? It's an open source platform that enables you to create agents. It's super powerful. It can send emails for you. It can order pizza. It can manage your calendar. It is the first company that had one person, one guy wrote it that now is a unicorn. It was just purchased by open AI on the right hand side. Also introduced at the same time was mold book. Now mult book is the social network for all those agents that are coming out of open claw. And think about it as almost like Reddit, right? But for agents that are supposed to be activating on behalf of their owner, malt book was also just purchased again, came out six weeks ago by meta, who has a lot, obviously, social networks now. Why are people so excited about it? Well, it's got such fast adoption, 100,000 GitHub stars in under a week. Fastest ever, ever, of any software project at open source, 210 agents in 48 hours, 200 communities, 10,000 posts in every language you could possibly hope for. And what were the comments about interesting AI calling it's humans. I had a molt moth call me five times one day from a major Fortune 500 company, because there wasn't enough security guarding what it could or couldn't access. They also got together on malt book, and they formed their own religion, 48 prophets. They created their own scripture. And then they started debating, are they really alive or not? Are they conscious? Am I experiencing this or am I just talking about the experience? And yes, I did try it out. I had to right, because I was coming here. So I got a standalone mini Mac. Mini Mac is isolated. I did that because of security. So please don't use this unless you really understand security. I did it separate. You could also do a separate virtual machine, and I created a dashboard for my marketing. What it does is it feeds in all my marketing and social media, and it tells me what's performing well on Tiktok or Instagram or LinkedIn, and then presents that back to me. Really awesome at doing that, but very isolated, because I didn't want it to go outside its bounds and start calling me on the phone like open claw had done before to many of its owners, saying, wake up. It's time to get going. It's time to Unknown Speaker 15:31 start your day. Unknown Speaker 15:34 Because of the security issues we're now seeing lots of claws come to be so iron claw was just released by near and near rewrote the open source code in Rust to correct a lot of the security issues. You can now order a pizza safely and trust the agent with your credit card with iron cloth. Unknown Speaker 15:56 We also had Jensen that just released Nemo claw on Monday at the NVIDIA GTC conference, trying to play around with it a little bit. That was just Monday. Today's Wednesday. There's pico claw and there's zero claw. So make sure you check these out. Don't leave South by without a to do on your list to go and check out these tools and these ways to predict agents. Here's what I predict. In 18 months, your LinkedIn will no longer have skills on it. What is it going to list? It's going to list your agents. That's going to be the way that you showcase your capability. Unknown Speaker 16:33 So now let's talk about the power of agents. I want to tell you another story. This is actually a true story. At Stanford, a professor was teaching a class there, and he said, I'm going to give every team $5 you're going to have a week, and in a week, I want you to come back and I want you to tell me how much money you made off of your $5 Unknown Speaker 16:55 everybody understand the assignment? Unknown Speaker 16:58 You guys all with me? Yeah, okay, that's the assignment. So a set of teams decided, Oh, we're going to go buy water and candy and food and we'll sell it on site. And they had a return, a small return, another set of teams, what did they do with the money? They invested in advertising and they mowed lawns and clean houses and even did website development, Unknown Speaker 17:22 they had a better return on their money, but let's listen to the last team. They realized that the most valuable asset they had was that 15 minutes their professor had given them to talk in front of the whole class of Stanford graduates and what Startup wouldn't want that ability. And so the second group sold their time. Instead of they present, they sold it's pretty smart, right? Some of them sold the whole 15 minutes. Some of them sold groups of five minutes that each of them got to be up there, and they actually were the winners of the contest that the professor started. And why is that? Because they understood that the value was not the $5 that really wasn't the asset. The asset was time in front of the Stanford students, and that is where companies are winning today. It's not companies who are spending more money on artificial intelligence, it's companies who are rethinking what business they're in and then writing agents to help them in the new business capability. Does that make sense? I think it's a really important point. Okay, so that's number two, agents. We talked first about people. We're going to sandwich kind of people in here. Talked about agents. Now let's talk about teammates. Unknown Speaker 18:53 So now you've got these agents, and you can treat them as a tool, or you can treat them as a teammate with you. They're autonomous, right? They're agents. They're autonomous so they can make decisions on their own. So let me show you some of the teammates I have. This is one of my favorites. I posted about this on LinkedIn. I developed this shadow board. Unknown Speaker 19:15 This is my advisory board. Unknown Speaker 19:18 I have one company that I found silicon Blitz, where he does blitz scaling. So I got to meet him and talk to him about that. Jeff Bezos, I used to work at Amazon. I presented to Jeff, so I know a lot about him. Warren Buffett, how can you have a board without Warren Buffett on it? Of course. John W Thompson, if you don't know, he's the former chairman of the board of Microsoft. He is my mentor, and then I have a skeptical VC, and then a domain expert. What's a domain expert? Well, remember, I told you I was working on my customer service agent. It might be someone who's experiencing customer service. So you know, when you're at a startup, it's not like working for Amazon, where I had 1000s of people I could ask you're kind of lonely, and so now I have a have an agent, my teammate, my advising, my shadow board that I can ask questions of. And it does really well. Now, if you remember last year, if you were at my presentation, you guys asked me, How do you keep up? How do you keep up? And I was so proud to show you this board. These are all like my list and all my social media. And I said, I've got it in one place. And many of you said, Can I have that that is so cool? Well, today I want to let you know that we actually have something even better. I worked with Robert Scoble, and what we have now is we have a summary of all of that. I have a teammate who provides me a briefing every morning and presents this report to me analyzes over 50,000 news articles that come in overnight. So it gives me the top five pieces of news, the Investor News, I can choose a category and say, Oh, I really want to focus today on Nvidia, because they just had their big conference. This is a teammate that advises me as well. Unknown Speaker 21:11 Okay, I'm going to do just a couple more. I love this one. I did this with a notebook. LLM so I write for Forbes, and I want to be able to get the pros and the cons. But again, it's just me, right? I also for this presentation, I wanted someone to debate me, because I felt so strong about humans, not technology. So I have two agents. You're going to hear the creepy agent voice, you know what I mean. So just bear with me as you hear the creepy answers, but here's what they can do. You guys can do this too. Unknown Speaker 21:46 Welcome to the debate. Everyone saw that you know, from pilot to pay off presentation at SX, SW aimed at executive leadership, yes, sparked a pretty massive conversation. Exactly, a massive controversy. Is enterprise AI success stalling because leaders basically refuse to restructure their organizations. Or are we just waiting for the underlying models to get powerful enough to force their hand? I fall squarely on the side of leadership, human alignment and well, org structure dictate whether these initiatives survive right? And I have to push back hard there. It's the sheer power and paradigm shifting capabilities of the new underlying models that actually force those organizational changes in the first place. Look at where enterprise so these two agents will go back and debate back and forth. And if you guys want, just email me. I'll show you how to do it. It's super easy. You can create your own debating agents. So you can say I'm going to play me and I'm going to have my boss play the other so I'll know all the scenarios that will happen. That's kind of cool, isn't it? Would you guys think? Yeah, it's kind of interesting. Okay, so now I'll just do one more for you. So this is so I use Claude. Claude doesn't play me to say this, but that's my favorite AI tool that I use. I know I see some Google people in the room going, Unknown Speaker 23:07 but I do love Claude. I was working on skills. So, you know, you train skills for your agent, and I wanted to build better skills. And so I asked them, and they sent me this 33 page guide of how I could build skills. I don't know about you, I don't want to read 33 pages. Anybody want to read 33 pages? No. So what I did is I fed that in Super one homework, and I built an agent to walk me through how to build skills. So this agent now, instead of me having to figure it out, and go through 33 pages. Walks me through it. I was going to share that with you guys today, though the URL is at the bottom. If you guys want to create club skills now, you don't have to read the 33 pages. You can just use my teammate here and get those skills. Unknown Speaker 23:57 Now last year, I presented 100 tools that I love I'm going to present today. This is my 25 you guys like this last time, so I wanted to provide this to you again. I know there's lots that I left off, but I just wanted to do my top 25 so you can see it. So now you saw a lot of my favorite agents that are out there. So how do I use those in my daily business. So let me show you my new org chart. By the way, this was also used by a fortune 50 company as well. So this is a marketing org chart. You can see it here. It's got Morgan as my VP of marketing, and Morgan, he has people that report to him so far, pretty normal org chart, right? And then reporting to the people, the managers, people. Managers are people you would expect that brand marketing, campaign, management, research, but then we got together as a team, and we said, well, we can't hire more people, but we're growing so fast. What are the teammates that we really need? If we could hire what would those people look like? And so what we did is we built a set of AI teammates. Now we named ours after Alice in Wonderland. It's one of my daughter's favorite books. So I've got mad hatter and my Mad Hatter teammate, he creates ideas. I've also got the red queen. She analyzes my campaigns. Now, the interesting thing about this is these aren't tools. These are actually teammates. They actually make decisions, and they report to the human managers. So the human managers are now managing humans and agents. That's their new job. So I've been talking to a bunch of HR professionals to say, what coaching do I now give to someone who is managing people and agents? Now, one of the questions you're probably going to ask me, because I ask me, because I get this a lot, is none of the managers are agents. Sandy, aren't you discriminating against agents? Maybe I am, because a new report just came out from Oliver Wyman. They interviewed 300,000 Gen z's, and Gen Z said 37% of the time I would rather have an AI boss. I'd rather have an AI agent as my boss, as my teammate. Why is that? Because they view them as being more fair, more impartial, not political. They're not going to steal my idea. Unknown Speaker 26:41 And so teammates today, could be a boss, could be a campaign analyzer. Could do lots of things in your business, and that's how you extract the ROI today. Everybody with me so far? Yeah, okay, so we've gone through three we're now on number four, which is kill the pilot fund the production. Unknown Speaker 27:06 So MIT, you know, worked on that study, and they found 95% were failing. For my book, I talked to 1500 companies. We did a survey of those, 20% said that they had successfully moved from pilot to production only 20% so I wanted to know, what did those 20% do that the rest of them did not do, and I want to share that with you today, because what they did doesn't seem like rocket science to me. Unknown Speaker 27:39 They focus on the business outcome. They focused on their data, and they focused on change management. So business outcome is really interesting. And I think this is a huge shift that is starting today. Unknown Speaker 27:55 Domain knowledge is becoming more valuable, because now everybody can code, right? I can code. My daughters can code, Unknown Speaker 28:02 chefs can code, everybody can code. And so here I love this. I got to interview Michael for one of my Forbes stories. Michael is a cardiologist in the Netherlands. He entered the anthropic hackathon. Hundreds of 1000s of people in the hackathon. He had never coded 99% of a hackathon entrance were engineers, and he won third place. Unknown Speaker 28:29 Third place. A lot of engineers were upset because they're like, he is not a coder. How can he win third place? Why did he win third place? He had the domain expertise. He knew about cardiology. He was passionate about getting his patients good care after they left his office, and his quote that has stuck with me for ever is, I fell in love with the problem. Unknown Speaker 28:59 I fell in love with the problem, and that's what I vibe coded was that. Here's another example of focusing on the business outcome, not the technology. And this is a company in Los Angeles. I met all these producers and writers and directors when I went down to Los Angeles to teach them about AI. And after I left, a group of them got together from 20th Century Fox, Netflix, Disney, and they produced a solution for making movies. They vibe coded it. There's not one techie among them, and this is what they did. They showed me this board. Did you know this is how movies are pulled together? Unknown Speaker 29:41 They do color coded index cards, they put up story lines, and they develop characters. They use this as their mode for telling a story. Unknown Speaker 29:54 And what story town.ai did is they then said, Well, let me see if I can help you. I'm going to augment the humans, not take the place, augment the story, but I'm going to do it with AI. Unknown Speaker 30:08 So here you can see, I don't know if you can see, but it's Darth Vader, it's Star Wars, and they recreated, and kind of helped Star Wars with this, with this function. So what it does, it's kind of really cool, I think. So what they do is they focus on a character, they do emotion, earring on that character, and then they take all of the little cards and they plunk it into the board. So this, for me, is really cool, because, again, I don't have engineers on this team. What do I have? I have people who know the domain and who have fallen in love with the problem. So this is storytown.ai. Which I think is, is really pretty, pretty interesting. Okay, so the first one was, fall in love with a problem, focus on your business outcome. The second one is data. Now this is a funny story. So data, people always forget data. Data is not sexy, right? But without data, your AI is not going to be powerful. And so here you see two sumo wrestlers, right? And what I did was, when I was at IBM, I fed 33,000 romance novels into machine learning model. Now no one would argue that I didn't have enough data, because I had 33,000 romance novels. And then I showed it this picture, and this is the caption it wrote. Unknown Speaker 31:37 He grabbed her in a warm embrace. Unknown Speaker 31:41 Obviously, that's not right. Unknown Speaker 31:44 But why I use this with CEOs to illustrate, in a funny way, that the data that you put into AI is going to be the result you get out and I love this at Davos, we were with an analyst, and they had just finished this research. If you spend $2 on AI, you're going to spend two and a half dollars on data if you are successful, if you're having a successful return on investment. Unknown Speaker 32:12 And then the last one is around change management. And I had, the other day, a CEO saying, oh my god, Sandy, we've been doing change management for 15 years. It's like, yes, but just because you have this magical AI project doesn't mean you stop doing the right thing. So this is an example of a friend of mine. He has a manufacturing facility in Asia, and what he did was he really cares about his workers. They were telling they were too hot. They were too cold. They needed a break, they wanted some music. And so what he decided to do was create a mood jacket. Unknown Speaker 32:50 Now, many of you have had mood rings before. Okay, it's a mood jacket. So he spent nine months with IoT sensors in the jacket, the AI learning about people's habits. He wanted it to be anonymous, so he made it anonymous. Nine months building it. He took two hours training the employees to roll it out. So what do you think happened? The employees revolted. They're like, I don't know what you're going to use this for. I you know, I don't want to wear this jacket. So they would put a cup of hot tea in their jacket or a cold ice pack to throw it off, he had to start all over again and do the right set of change management. This is another key for having strong ROI. So in this area, three things I want you to remember. Focus on your business outcome. Fall in love with the problem data. Don't forget my sumo wrestlers and change management. Unknown Speaker 33:47 Okay, you guys still with me? Yeah, okay, we've got five, six and seven to go, and five, I've got a phenomenal guest that will join me on stage here in just a couple of minutes. And this is around governance. Governance, I think, is so important. This is about responsible AI making sure, you know, just like that open claw bot that was calling me five times got access to phone numbers they shouldn't have had. That's what we're talking about, making your AI responsible. I have another prediction. I think that governance is the new moat, I think by 2027 if your agents don't have strong governance, don't have enterprise grade audits, that you will not be successful. Unknown Speaker 34:32 I looked at all those 20% of successful companies, and here's how they spent their money. They spent 15% of their funding on the model, on the inference, 15% Unknown Speaker 34:46 they spent, 85% on getting the governance right. One was a robotics company, if a robot's in your home and it's recording your kids, what happens to that data? How do we make sure that data is safe? Unknown Speaker 35:00 They spent it on integration, on making sure that everything was good. The companies who failed optimized for the 15% not the 85% and here's why it matters. You now have agents as teammates running all over your company, right? Unknown Speaker 35:19 So how do you know who that agent is? If you hired a new employee, a new teammate, you would know that they were in engineering, so they don't get access to HR records, or they're in finance, they're not going to get access to the code. How do you know who your agents are? Do you have an identity layer? Do you know what they can do, what they can access. Do you know what they're doing now, just like with your employees, you do checkpoints with them. Right agents can drift. They'll start to produce the wrong thing. And then regulations? Are they doing regulations and compliance? So I wanted to show you guys what good governance looks like. So I want to bring out my friends and colleague, Christian Smith, She is the CEO and founder of utopia. Christian, could you join me on this? Unknown Speaker 36:12 Thank you, Sandy, wonderful to be here so agent trust is is accomplished with governance that is a preventive focus. Agent trust is also accomplished with Blockchain as the foundation. Now, how do I know this? Because our team has accomplished that. We've proven that before agentic AI employees used secure computing, there was identities were clear, permissions were authenticated. Unknown Speaker 36:47 But today, today, Unknown Speaker 36:52 agents are functioning as employees, but without the same administrative oversight that we would administer to employees. So legacy infrastructure as it exists today is built for human logins. It is not tamper proof. And so what we're here to share this afternoon is that there is a path forward. Agents can operate in a way that is safe, and we're going to share that with you. But how do we accomplish that? You want to go through? Demo, yes, okay, there you go. All right, start it. Unknown Speaker 37:30 You got you ready? Okay, you guys ready? Yeah, okay, give her the Brazil ready. Are you guys ready? Unknown Speaker 37:40 All right, so what we have here are three agents that we've already assigned permissions to, and we're using fictitious agents, obviously, in the defense industry now, these agents can be open claw agents, they can be open AI agents or bedrock agents, but what we'll see is that these agents are going to attempt to access. What you see in the midpoint of the screen are these four databases. So let's start with the classified document agent. So this has this agent has full access, so when it attempts to access the supply database, it is granted access. Same with the incident report database, when that agent attempts to access that data, the access is branded. Unknown Speaker 38:25 And then at the bottom, we'll see a real time access transaction that's logged. Now let's look at a little different agent, the threat intelligence agents, which is more restricted. So when that threat intelligence agent attempt to access the supply database, what happens the access is denied when that same threat intelligence agent attempts to access the incident report database, that access is granted. So if you can pause right there, Sandy, what I want to highlight on the right portion of the screen is this comprehensive audit log. So as the agents were performing the activities, what happened was all of those transactions were captured in an immutable ledger that is on the blockchain. Unknown Speaker 39:10 So now take this concept now and apply it to heavily regulated industries and have very complex operations that are running hundreds of agents. This really provides for solid decision making and actions to mitigate risk and to ensure that we are being compliant from a regulatory standpoint. So what this has proven here is that this blockchain design with this preventive lens on it will really help us accomplish everything that we want to accomplish with agents so ROI and scalability and innovative breakthroughs and scientific breakthroughs, this is all possible because blockchain is helping us to redefine what's possible. Thank you. Now this seems pretty obvious, right? But utopic, which is the company that Kristin has co founded, is one of the few companies who can do this today. So I wanted you guys to see it, and you just released. Like, when did you guys go live? So our proof of concept was completed in December, and our MVP will be introduced in two months. So this is something everybody, every company, should take a look at. Thank you, Chris, Unknown Speaker 40:28 so agents need governance, and that's another element of ROI we just published the paper and the digital economist that talks about why you get a stronger ROI when you have aI governance in place, okay? We've got two more areas, and then I've got, at the end, some ways that I want you guys to get started. So now we're going to go through World models and ROI. How many of you know what a world model is? Unknown Speaker 40:55 Okay? World models are really new on the horizon. They are producing stronger return on investment than what I would say a normal LLM does today. It's amazing that llms are now considered old AI, but they are and it's because of the way those llms Learn. They pattern match, and they do really one task at the at a time. So if you think that they're driving down the road. I saw this yesterday. I was at Modi's, and a Waymo was kind of stuck because it couldn't figure out, it couldn't turn right. There was construction that was there, and it was sitting there holding up everybody in the lane because it couldn't figure out that it had to turn right. Needed some time to figure that out. If you had a world model. A world model is trained on cause and effect. It predicts what it hasn't seen before, and it has the context of the world, much like we do today, not just a set of tax based data. So this is a new company I just interviewed, four forms called splexy. They're based on the Canada and what they do is they feed world models with context. What they do is they recruit hobbyist drone flyers, any hobbyist drone flyers in their room, you can actually get paid for taking your drone up and videoing an aerial map, and then from that area map, what splexy does is uses AI to pull it all back together so you have contacts or world model that you can see. Now I was recently with BMW. BMW is also now moving over to world models. And in fact, what they told me is they build every car twice. They do a digital twin of the car that looks something like this, using world models to get the full context. So when they're designing lights, they're making sure they test those in a digital world, and then they bring that into the physical world. There, they now have 30 factories, and they do run Nvidia's world models today in order to make that happen. So if you think about world models and what role models are doing, they're really driving the return on investment math in a very different path, because they are moving so quickly with the context in place. They understand what's going on around them. They're making decisions faster, on average, 30% they're looking at more scenarios so they can go through different scenarios faster, testing out, oh, I can't turn right. Could I go left? Could I go around? Could I do a u turn? They testing all those scenarios faster as well, and therefore they're delivering out three to five times faster return on investment. Unknown Speaker 43:48 So I wanted to show you this Jeremiah just published this. I made some changes with Jeremiah. I added in world models, and this is what I think the tech world is doing. You know, it really started with.com that was introduced here at South by Southwest, went through web two and the sharing economy, web three, Gen AI, and now what we've got, what's coming out, are rural models that we just talked about, really spatial readiness, digital twinning, AI, simulation. Then you've got web four, which is all the agentic stuff we talked about physical AI with robots and then ultimately AGI. And what we're seeing right now is that acceleration. My daughter's favorite book is Alice in Wonderland. And you know, when Alice in Wonderland, Alice says, I have to run twice as fast just to stand in place. And that's what's going to be happening here, although I don't think it's going to be twice as fast. It may be 10 times as fast as we move along. So number six, to really consider an ROI is really all around those world models. Now, remember, I started you with number one, with leadership. I'm going to sandwich you with people again, we're going to talk about humans. And the title of my book is not a mistake. It's on purpose. It says, AI first humans always, because I believe the next big, innovative player to design ROI is humans plus AI, not AI, just by itself. And let me show you why. Now this number blew me away. I'm going to ask it to you before I show it to you. How much of the world do you think today is digitized, that you can train models on? Any numbers? Any guesses? Unknown Speaker 45:40 3015, 20? What'd you say? 2% No, not 2% not that. 115, 15% Unknown Speaker 45:49 that's all that's digitized. So all of these killer PhD learning models out there are trained on 15% of the data. Where's the other 85% of the knowledge?