The 8 Steps to AI Adoption – Graham Donoghue, CEO, Forge Holiday Group

[MUSIC PLAYING] Well, I mean, I guess hands up. Who would like a fair to percent competitive advantage? Who's not got the hands up? And who would like to increase the multiple exit? So we're going to teach you how to do it, because I firmly believe that is what AI can do for you. [MUSIC PLAYING] Hello, everyone, and welcome to the first ever episode of the Pep Talks podcast to feature a live audience. We recorded this in front of 250 private XC back CEOs at our annual conference, and we're joined by four holiday group CEO, Graham Donahue. Graham shares his eight steps for AI implementation under PE ownership. This isn't just theory. This is a usable step-by-step guide for a CEO who wants to use AI to deliver tangible, competitive advantage in their business. It's delivered by Graham, who has achieved this by staying at the cutting edge of the newest tech, enabling his team and putting real structure behind his AI strategy. Throughout the podcast, you'll hear Graham refer back to some slides and some other useful documents. I've put a link to these in the description of the podcast, so you can download them to either run through as you listen or absorb in your own time after you've heard the podcast. I hope you enjoy. Welcome to this episode of Pep Talks podcast. We are here at our conference. [APPLAUSE] You know, we have a live audience. This is how much noise 3,000 people can make. Absolutely. So welcome back, Graham. Great to have you back here at the conference. And we are running a session here on AI adoption, implementation, and value creation. That's what we're going to do, isn't it? Seven, eight steps. You're going to have those in your app to take away with you and those who are listening on the podcast episode who aren't here, you can download them from our website. So let's start off, Graham. Let's kick off. It's a fancy pair of glasses you were in, everyone. It is as much fancy in the ones that had on when you took that picture ten years ago. These are my-- because we're talking about AI, I thought I'd wear my AI glasses. So these are meta AI glasses. So it's just an example of how the hardware is sort of advancing and AI is built into it. So they have bone connectors, interfaces, into a large language model. I can ask you questions. In fact, should we take a picture? Well, I will. I'll get to it with anyway. Let's be quiet. Hey, meta. Can you take a picture? Now, I don't know if it flashed or not, but I've just taken a picture, and that's gone directly to my phone. I could record a video. But actually right now, I've got Judy Denge talking on my ear, reading my wife's whatsapps. And for some reason, I picked Judy Denge. She's the AI voice. I quite like it, but I'll switch them off. But just thought it would be quite fun to wear that. I'm going to talk a little bit about hardware and computing power and how things have changed. And I will make no apologies. I'm quite geeky, and I love AI. I'm like a former. I'm all in here, type of thing. And we're placing some big bets as an organization. And I'll try not to be too technical, but I think it helps give an idea of-- Definitely. What's going on in the world? So before we kick off, I'll switch Judy off. Because she's annoying me. Yeah. They all think you're bonkers. You've got AI glasses on. But that's before we kick off on the seven steps. Let's just talk a little bit about the opportunity here. We've had a morning of reality, haven't we? It's been a morning of, oh god, private X is changing. There's some tough challenges. It's difficult to sell. But there is this incredible opportunity. What is it, Graham? Well, I mean, I guess hands up. Who would like a fair to percent competitive advantage? Who's not got a heads up? Or why are you in? And who would like to increase the multiple exit? So we're going to teach you how to do it. Because I firmly believe that is what AI can do for you. When I talk to-- we've got to the stage now where we're larger. So therefore, I'm talking to more bigger funds. You're KKRs, you're black stones, you're car lives, those sort of guys, the ones that can write bigger checks. And they're all over this sort of stuff. So I genuinely even give you competitive advantage. And the numbers you cannot deny. Because I'm old enough to have worked in the internet in e-commerce. My first job was working for a two-way travel in the late '90s. And I was told the internet would never work. And I left seven years later. And we were doing five and a half billion on line seven years later. I was at my super market for a number of years and we just dropped his using technology. So there's going to be 500 billion invested this year in AI technologies, investing in capability, investing in infrastructure, investing basically in these bets that are sort of a place to go. And I know there's not a talk about a bubble. You know, no correction. And there may be. And I know some people shorting. You're in the video stack, et cetera. But this is faster than anything I've seen ever before in my relatively short career. So I'm super excited about it. And the adoption curves are just stratospheric. Yeah. And I think for skeptics sitting in the audience, what you should be thinking about is the risk of not adopting and not leaning into this. You will be judged. And we heard from Tim this morning in terms of how buyers are looking at businesses from an AI perspective. But how they're looking at the capability of the CEO and the management team from an AI perspective. And if you can't answer the question of what are you doing? How are you doing it? And what value is it going to create? If you can't answer those three questions when it comes to a buyer conversation, forget it. You're back in the realms of this as well. And there's maybe slightly controversial. I like to think it's the CEO's job as well. I think you guys have to lead from the top in terms of the message. This is not a side project for your CTO, your CPU, or whatever it's like maybe, or it's like technology probably, I think you have to lead from the front. If you want to take advantage. And again, even more controversial, there will be some people in this room whose businesses will not be here in a few years' time because of AI. There will be some businesses in this room that you've got a massive competitive advantage because you've leaned into AI. And I'm happy to think that bet with anybody. I have a vested interest. I want you all to be in business in two and a half years. I don't want you all at all. We're going to give you seven steps. So I thought it was eight, actually, but I don't know. It's seven with us. I think eight steps is when you reach the agenda AI, in Nevada, but yeah, we're going to try and help. OK, so last question before we start. You are about two and a half years into this AI adoption, value creation, discovery, driving it really through your business. Our outcome that we want is we need to be catching up with Graham. We all need to be doing what Graham has done in the last two and a half years, irrespective of your sector, business size, or where you are in the cycle. We all have got to be doing this. And these are our seven steps. So I was at conference yesterday. Someone reminded me three years ago. I did a presentation. And it was in Madrid. And it was using Synthesia, who do AI avatars. And I was talking about, this is what the future of conversations could look like. I thought, we were quite early adopters. But I think we're on a curve as a business. And we're ahead of many businesses. But I think we're miles behind tons of other businesses. And that's my formal coming through. So when I think, I think in horizons, so horizon one, horizon two, horizon three. And what I'm trying to do is to place bets in each one of those horizons. I don't really know the time frame. I mean, I put some time frames around here. They're going from enabled to rewiring your business, to rewriting. But every time I do this slide, I keep making the time frame shorter. Because something news come out, and the technology's evolved effectively. So I would encourage everybody to think about one of those different horizons. What that really means is you've got to be placing some bets that might not be mature into horizon three. But you have to place it in those bets now. Because other people are, and many other companies are now. And I think if you look at the 500 billion going into open AI, and Amazon, and Microsoft, and all the other companies as well, they are planning to disrupt a lot of businesses, and they're placing some big bets. Yeah. Just on this sort of road map, really, you can't get to rewired. And there's no way you can get to re-write unless you've done enable. And we're going to cover all of this in the seven steps. So step number one-- oh, we've missed a bit there. Yeah, do you want to just-- Yeah. So I loved Eliron. I think it was Paul who was talking about lead indicators, lag indicators, building it. This is how we sort of run our business. We run this model using a principal goal, the 4DX, the four disciplines of execution. Because I always think my job is to make sure-- clearly, we understand we're going strategically. But also, we've got one eye on how do we get there? What's the framework for executing that strategy? So we use this sort of model, the 4DX model. And really, all it's saying is on focus on one or two wildly important goals, and terrible, I like 15, but I'm not allowed. I can only have two. Act on those lead measures, not the lag measures. So act on those lead measures, and then create this cadence of accountability in your organization, and keep talking about it with a scorecard. But when it came to AI, we said, well, there's probably three things we want to do in the short term. One is adoption, and that's debt for the adoption through the whole organization. So that's about people using AI, and I'll come on to it in a second, but also what they're using. Readiness, so what are we-- what are the bets we're placing to build out this AI infrastructure, and this big bet for the future as well? And then the last one is a value creation, which is the one my investors keep saying, like, so what? Where's the value in this? Give me real examples of where you're saving time, money, becoming more efficient, et cetera. So that's the framework that we use. Fabulous. Right. Let's get into it then. Step number one, build your data brain. We don't need data, do we? Yeah, we've discussed this morning. It's not necessary for exits or anything like that. But is there any difference really between what we need in terms of exit versus what we need in terms of data for AI? I mean, what we discover in our business is we're data everywhere. And the beauty of AI is it can interpret information, it can interpret data, and it can do things with what we couldn't do before in very layman's terms. So we started building out this idea of a data brain, this sort of a said, like, how do we get all of the data, have a single version of a truth. Build out all these little data marks to use a slightly technical term. So you're building up these little sort of pods that sit within the brain effect of it. They can be accessed. Information, a good example can give you, is we have a million contacts a year in a contact sensor. There's lots of calls coming in. We'll record all those calls. Let's get all that information, and let's put it into the data brain. Because actually, there's going to be value in that. And AI, we're really good at interpreting it really quickly and making sense out of it and learning from it. So building out the data brain for us was really about sort of a saying. And we're on this journey where we are walking with partners and in the some of the room to sort of help us a bit. It's really about trying to get this collecting, unification, storing everything to water. Because if you get the data right, it can make a significant advantage. And the beauty is, there's tools now you can use. That are really clever that can sit on top of the sort of the data that can give you information that just massive of the accelerator of business as well. But it's not just financial data. It's the key point in there. It's like anything is data. The image of a property is a data. A meta tag in a property for us is a data. A review is a data. Build out all those marks. Yeah. So there's quantitative data. And then there's qualitative data. And what I've learned in the last three months or so is just the value of capturing the qualitative data. And I don't know how many of you are recording every call between functions in your business and the customer. Not, for the podcast, definitely not a majority, a minority. I mean, that's just a massive opportunity of capturing information. And if you capture it in the right way, what you can do with that in terms of training AI to sift out value. Yeah, so I mean, I'll give you just one very simple value creation point. And because I know there's probably some people who don't say, well, yeah, it's all what you do with it. So I've made a contact with you here coming in. We record them. We use the thing called whisperer. It's like an AI tool that goes over all that data that allows us to summarize all the calls coming in to create a little clever summary for all the agents. And then actually also to do some analysis, predictive analysis, when a customer is calling us or an owner is calling us, why are the calling us based on that sort of data? Save the knowledge of 50 minutes per individual contact center employee, and we have a lot of them to give you an idea of a value creation from it. Absolutely. OK. Step number two, choose your platform and architecture. So we have a brain. We have our brain. We are capturing our data. I mean, let's just start with the basics. Choose the platform. I've had lots of chats with you over dinners. And some people don't actually have a platform yet of choice. Do you know what we mean by that? And I'm asking a rhetorical question. What do we mean by that? Great. Well, I mean, that's quite a big question, actually, because in the narrowest sense, it could be as simple as, like, who's your chosen large-language model that you want? No pilot or-- Yeah, you know, your enterprise level large-language model that you want is sort of use of. And a wider sense, it's like actually how are you building your architecture to be ready to enable you to take advantage? And that's about your data brain, your componentization, walking with partners, you know, be that Microsoft, or be the AWS, whatever, to allow you to take advantage of all this sort of stuff. But I don't know if we have a diagram. We do. I mean, this is like-- I apologize. Right? I think it's like, what is that? But I use this. And I have a massive carboard, like, three meter car of this, and I wander around the building sort of a some time, bringing it alive. And so this is, like, the forge and genetics, sort of a platform we're heading towards. And the very top of it there, what you have is these agents. And I'm sure you've all heard the phrase agents. But basically, it's the difference of moving from generative AI, where you're putting information in, or you're getting information in back, type of thing, versus actually these agents that are doing stuff for you. They're doing, like, end-to-end stuff, you know, copy transactions. It could be, like, complex sort of processes. They're heavy-liff, and they're doing it. So we're building, and we're actually partnering with Microsoft, and DeFrain, and Joe's in the room, shout-out. We're working with them to build this out, and you have this planner agent, also called an orchestration agent. That's like the one that comes in, like, your concierge. And then it figures out, actually, what you're doing. And the model, really, is you then define a specific agent to do specific tasks, because I understand the belief, you can get an agent to be an expert, and something that's just better. So that's what we're sort of building out. But one of the things I'm scared about is, I don't know if I'm going to have a brand in the future. I genuinely don't know. I don't know if the future is, everybody is going to have a digital twin, and they're going to be using, you know, some form of that digital twin, and they're going to be going to-- it could be OpenAI or ChatGPT. That's where they go for everything. So what we're trying to do is an organization, which is why you have this thing on my left, which is the ChatGPT. We're trying to also say, well, actually, if these agents and these platforms want to come and talk to us, let's open the door. Let's make it really easy for them to come into our business and to explore our amazing data and our amazing content and potentially transact and do business with us as well. And there's lots of chat, you know, a few weeks ago, ChatGPT, Fiber All-Launched, and they developed a software development kit, and you can now buy things and Shopify and bookend.com inside that, you know, and I know which way we really go. So we have to build an archery chapter that's ready to play different bets. But when these bots come in and go through this thing called the MCP layer, and that's the question, go away and talk to your CTOs or your CIOs and say, we got an MCP layer, do you know what an MCP layer is? And if they don't fire them. No, and what is MCP? It's to describe what an MCP layer basically is. Like a message of contactable. It's basically like the interface language you use to talk to all these LLMs to make it really easy. You know, if you go back in the olden days when I first started, we used to be really good at search and optimization. So like, how did you make it easy for Google to come into your business, you know, robot.tax file to make your data available? This is the equivalent of that. And I can tell you the way the LLM models come in and interrogate your organization and interrogate what you do. If you're a consumer brand or a rather brand, it's so different from the way that Google come in and do it. And you have to learn how to do it and track it and monitor it. And I have 7% of my traffic today that's coming through AI mode in LLM models that didn't exist, obviously, 12 months ago. Now, it's not converting, but it might. And it might in the future. Model context protocol. Ask your IT director, CIO, CTOs, if they know what you're talking about. If they don't, you've got a problem on your hands. You don't need to-- I mean, that is unfair because like, you know, there's not a lot of people doing understand about that, but trust me, this is a big thing. And this is a thing that, you know, it's, you know, it's a lot of the more advanced organizations that I'm doing and opening up your accent because you want-- You don't need to get to this next week or even in six months, but you definitely have an enterprise application. And you definitely need to be layering your technology stack so AI can be built into it. And you need to be getting to some-- if you're a large-- yeah, a decent-size enterprise. You need to be getting towards NCB pretty damn quick in the next 12 months or so, is that. Yeah, I mean, I have a 25-year-old son, who you know, actually. He's been teaching us this one. Yeah, about 25-year-old son. And I have a 18-year-old daughter, and they don't go anywhere near Google anymore. They're always using, actually, TikTok, Instagram. But more recently, large language models. Now, when I ask them, do you trust them? No. But that will evolve over time, because I remember people say I don't trust a mobile phone to get my credit card to. That will evolve. But 90% of that demographic is using those tools. So you have to be sort of visible within them. Yeah. There is something coming next after NCB. When we're here in 12 months, what will you be talking about? We're going to be talking about ACP, aren't we, a genetic context? No, not agent. Agents context. Agents, agent, agent context protocol. Because that's basically-- an agent basically is then. And it's the same with the thing, basically. It's loving-- it's teaching the agent that comes in to perform tasks. That's what an agent context is. A gateways, doorways. Yeah, and I think if you look at the more recent evolutions of what's happening with a particularly open AI, but they're all sort of doing it. This is where they're sort of going. It's that. They want to be the destination. They want people to transact within that platform. And they're obviously walking and partnering. There's a lot of cyclical money going around here that are all like paying each other, which is a bit scary. But anyway, and they're all walking. We figure out they're trying to try and unlock this. OK. Number 3 and 4. So let's do 3. Develop governance that enables. Yeah. Freedom. Freedom, we find a framework we call it. So when we started, I wanted the idea of AI for all. I wanted the whole organization to learn from AI. I used to say, I want to give you a superpower. And so we went through a piece of work to, first of all, define a policy. It sounds horrible, a policy. You want a policy. But it was chaos. It was wild west. People were using their Gmail address and putting spreadsheets into chat GPT. So I needed to stop the anarchy. So I needed to give people the framework. And actually, what we did is we actually launched a freedom within a framework using AI and actually created a digital avatar of one of our employees. We called them AI Josh. And then he communicated the framework out of that today. So people were allowed to play. But within the framework, we gave different levels. Because we recognized that we didn't want to close it down. It didn't want to be sort of a corporate sort of saying, no, now, shall not do. So we developed different freedoms for different people. And together, we developed training and education. And we went through all this training and education to teach people at the very basic level, it was, what is AI? What is an LLM? What is a prompt, et cetera? Through to clearly more advanced prompting and more detail. And we built all of that training and modular inside. And we made levels one, level two, as compulsory for everybody. So as you graduated, you got more freedom. We then developed guardians. So we developed these, like, ninjas where we gave them all, who dees AI guardians. And everybody wanted to be an AI guardian, because it sounded really cool. And we gave them different land yards. And those were the people who were picking, were saying that they really get it. And the thing that really surprised me is it probably wasn't the people who I would have initially thought would have been the adopters. Quite often, it was the people who got, like, terrible jobs that are laid monotonous and really wanted to get out of that rat race. They wanted help. So we developed the AI guardian sort of a program. Then we did the education, the guardians altogether. And then we just encouraged, like, mass experimentation, just going to play. And I think one of the things that really struck me is when we said to people, if you go to say coal pilot, this is our enterprise level version, it's all protected. Go nuts, do whatever you want. Stick your spreadsheets in, do whatever you want. Because you can't break it, and you can't do anything wrong. Don't use some of these Chinese ones over here. But if you want to use-- if you want content copywriting, and you want really good copywriting, well, general cloud is really good for that. So we actually built, like, a little AI tool that was like an agent who could go in and say, what's your task? And then the agent would basically then work out. Here's the different tools you could use. And we then curated 55 plus tools. I think it was that we said, any of these you can go and you can knock your socks off. But then the guardians, because they've got superpowers, they were allowed to go and spend money on some of the more complicated tools as well. And we gave them a different level of freedom, yeah. I mean, when we talked about it, you basically said, you've got to go and find the people in my businesses. There's only 12 of us, so it wasn't very difficult. But in your businesses, there will be people in your businesses using some of this AI in there. And they'll be advanced. And as Graham says, there'll be the people you least expect. You've got to go and find them. You've got to task your executive team with finding those individuals who are already experimenting with AI. And you've got to bring them together. Form a little lab. We're calling it stand-up an AI lab. And that was really your guardianship, wasn't it? You found it. We found these people, and people know that I was really passionate about it. So people would like to please deceive you. So there was a lot of that sort of going on. It was quite close, quite nice. But then I would stand about it at every single old hand meeting, like, what are we doing in AI? We use a tool called Netscope. And anybody who-- we vary between 1,670 and 1,200 employees. Anybody who has an account registered in a business, we track it all through our system. We use a single Netscope. Netscope told us who's using AI, what they're using, not what they're doing, but the depth and the frequency as well. So we had like a scorecard going back to the 40x model. We then said, well, let's have an AI day, because that sounds pretty cool. Then we invited multiverse and faculty AI and real thought leaders in. And we got people really excited. You know the team that got the most questions on the AI day, the legal team? Because there was still a bit of a fear about what you can and can't sort of adieu. We today-- I mean, just moving forward a little bit-- but today we have the doctor of surgery going on today back home. So basically, we took the guardians. They're all the white coats. Each one of them is called like a doctor. So we have doctor Otto, who does all the automation. You get an automation problem. You go to the doctor. And we create surgeries. So it's just this constant sort of like embedding AI into the organization. We have a war that I give every month in the organization about the people who are using AI most frequently, but also the most innovative ways. And then we share the stories about how it's sort of helped us or really deep, deep, deep into the organization of culture. So you've got to create the environment where it's safe. You've got to put some guardrails around it. You've got to put some governance around it. Find the talent internally. You don't need to go external yet. Find them internally. Give them a freedom of framework. Make the best of them. The ones that are really innovating your champions. They're the people that are going to solve. And Richard's going to come up here. He's our champion. He's going to show you some of the stuff that we're doing. This is not-- this sounds quite difficult. It's not that difficult. You just make it a priority on you to do list. Speak to your executive team. Go and find the talent once you've got the talent. Bring them together. Governance, and then shaping what we're going to work on. Exactly, we've tundered that. The one thing we'll say, sorry, it's not a contradiction, but just to build on, it's like, those of people internally really talented using AI. And I think we've got to slide with some of my numbers to share. But also, I then recognized if I'm thinking of horizon free, I'm maybe even horizon two, I worked with my boards and my investors to say, can I just go work and hire the best people I could possibly afford, and quite frankly, even people I couldn't afford, who could help us understand and unlock this. So we then went on a journey to figure out the capability and massively invested in product people who understood how to do a conversational design, data scientists, world leading data scientists to come and work on this stuff. And what I found is-- and yes, we paid them quite a lot more money. But what I also found is smart people who are doing this stuff, the one who work on really clever things. So some people who came to work for us now ask me a year ago, I would have thought we never wanted to come to a little cottage business in Chester. If you get into this, you can attract the talent. People also spend money on apprentice. Yes, it is a good point. So this year, we will spend 750,000 pounds on apprentice AI apprentice trading. Now, I'm sitting because I'm not really spending it. I'm using a levy, working with multidirest. I don't know if people know multidirest, UM blarers. And so we have a bunch of people that are going through their apprentice of a program. And we have that whole program, and we make the apprentices stand up for the organization and showcase what they've done, what they've learned, et cetera. One of the other phrases that you use is it's democratizing talent or democratizing smarts as the phrase I sort of attend to you, yes. So democratizing smarts, because there's a lot of talk around. I mean, yesterday, again-- this is Shoshia Gika, an open AI, or touch of tea, 5.1, just launched with personality. Don't quite know what that means. But it just launched with personality. And actually, the war and the future people talk about is there's war and emotional intelligence, because actually AI is democratizing intelligence. Now, I actually believe it's like combination of the human and the technology sort of working together. We always talk about AI can make things up. It can hallucinate, don't trust everything, you're sort of a seeing. That is how you blend those two things together. But I think you can see where they're going. We're trying to build an emotional intelligence into that sort of a capability. And if anybody's ever tried sesame AI, they have an incredible agent called Maya. Download sesame AI and talk to Maya. It is scary. Sounds like Scarlett Johansson, which is quite a cool one. OK, number five, build a user case pipeline. Yes. So we've got our data. We've got our application stack. We've got some governance with safe. We've got some talent internally. Now, what can we do with it? Yeah, so what we discovered is in order to get the highest level of adoption, there's tons of things we could sort of do. And there's millions of ideas coming through for some of people. But I wanted to find things that were like monotonous, that were like, you know, very process-driven. But actually, we're maybe not enjoyable for some people in terms of doing their job. How can I use AI to fix those parts? Shitty parts of the deal. Yeah, friction that exists within the organization. And, you know, we're a 50-year-old organization. You know, we've built a lot of stuff over those years and created a lot of debt. You know, so how can AI help us fix some of that? The other thing I wanted to do was to find things that could benefit the widest group of people in the organization, because that was my ticket for a high adoption, because they could start to see, you know, what sort of happening there. And then the other thing I looked at was, where are things that could really help me create more value in terms of revenue or EBITDA, and, you know, I'd add to hire lots of people to do or invest a huge amount of money in to create. So I can give you a very simple example. And there is, you know, we have 30,000 properties that we look after in the UK. I have half a million images that are sitting on my database. I cannot afford to go out and create a video for every single property that people like videos. I also can't afford to go out and reshoot every single property. But guess what I can use to help me do that, and to make it, you know, incredibly slick and fast, and turn static images into video images without hallucinating. And that was one of our biggest problems it started to hallucinate with properties where it was putting swimming pills and it didn't have a swimming pills. Like, that's not going to work. So if I showed you a prompt that we first started doing this, our prompt was, like, free lines was started, our prompt now to make that work to stop hallucination is like this big. But that's a great example. That would have cost hundreds of thousands of pounds. And it helps me because, believe it or not, people buy all the digital images. So actually, the point is here, you start with the mundane, boring tasks that actually affect quite a large proportion of the employee base. There's a lot of people that go, oh my god, you can take that away from me. That's amazing. You free them up to then do the higher value work. This isn't about taking jobs away straight away. It's not. Yeah, and it's not. And that is a great point because people, you know, I stand up here. And I had to be really clear with people saying, this is not about taking your job. And then people are like, yeah, really? It's not about taking your job. Because we're a growth business, it's like, I won't need to hire as many people moving forward. And I want to make your job easier. And I want to give you a bit more free time so you can use your brain to deliver better service or to learn or to walk on those things. And I was very conscious about making sure this didn't become a story of taking people out of the business. Yeah. Yeah, this is about freeing people up to be more productive, to produce higher quality work in the Monday tasks, and to set themselves up, and the organization up, to create value in a way that they haven't been able to do before. And it starts turning through opportunities that are nutty, difficult problems that you think can kind of have to spend a bit of money on that. And actually, with this sort of software and this sort of technology, you can start solving some of those problems. And it changes language, so someone comes and tries to hire someone that we're a bit like, what can AI do that? And actually, in my org chart, it's a bit weird. It just shows you how mad I am. In my org chart, I've got AI agents that sit in my org chart alongside other people. So I was trying to see if I had the legal team, a growing legal team-- people like to get more litigious, whatever. So instead of hiring power legals into the organization, we have an AI agent that just does the legal activities for us. And we went, and we found a four-party company, who specializes in this as a niche. But that agent is $25,000 a year agent that sits on the payroll, technically, in the org chart. Let's do measure value creation. Yes. We're not just throwing it out there and saying, OK, go play. Have a wonderful time. No. And you're going to come back and be more productive and creative. And we're all going to be fabulous. You are measuring every step of the way. Yes. So we're measuring-- as the next slide, are we going to talk about the-- Yeah, so I'm sorry. It probably looks horrible from the back of the room. So this is a weekly little power BI dashboard that basically shows what department is using AI. So the total number of interactions, technologies, winning, six and 1/2,000 interactions last week, and then we have marketing, data, analytics, product, et cetera. Then there's a division, which is the number of interactions by the number of employees in that department. Then you've got what roles are using AI most, and then you've got who's using AI most. So you've got a combination of a depth here going on, as well as overall sort of a usage. So we're measuring that. And then we're trying to work out how many hours-- because this is where I landed, because of so many different ways I said, how many hours do we think we're saving the organization by either not hiring people or as a result of AI sort of a productivity. And we have a measurement framework. And it's not perfect. But it's something I can get people to talk through as a story. And what I've discovered is, I think we're massively under indexing. I think there's so much more going on in the organization. I came up with the idea of 150,000 hours a year of productivity gains through using sort of AI. But what that doesn't capture is all the things that we've done through our agents or through our agentic chat bots or through our conversational AI up front. It doesn't really capture that stuff as well. So that's like a framework we use for measuring and employ utilization. But then also we then try to turn that into like a value as well that we can report. I mean, what are the interesting-- The dashes. I mean, people and cultures really-- Huge. People and culture team absolutely love it. So our employee engagement service that go out-- they get all the information back. They chuck it into AI. They purchase these amazing sort of a dashboards. They then turn it into using a product called Hagen. And they've got a digital avatar of Victoria, who works on a team, and we use 11 labs voice of cloned voice. And then she delivers like the employee engagement back to the organization. I mean, they are, they are all in-- Finance. I know everybody in their room is going, what can we do with finance in AI? I mean, that's like really heavy process, like you know, complicated sort of a tasking and process going in. I mean, the finance team, although ones were had to spend the most time with, and to sort of help them, and they haven't blessed them. And because they were a bit like, if I use AI, I'm going to be-- I'm not going to be with the clever and your ticker job. And they were quite-- It was quite a high element of risk with finance, isn't it? I mean, you know, with people in culture, markets. Well, my board report was produced using AI. And it's a dream, I have to say. So that's used to sort of AI. But we do these things called surges. We stole it from Elon Musk, where he would go into any of these businesses, and they do these really intense surges. So we got the product team, the technology team, AI guardians, and we went in, and we want to all the processes that we could use AI to help an automate. And there's tons of, as you could imagine, like, information going out, letters going out, payment collections, automation, chasing up debt collection. That team saved 1.7 million last year just through improving the way they collect debt that was old to us. A big proportion of that was through AI. You've got to quote down here. I know, that's the we say 1,000. So I've got another quote. Every agent has to earn its place by proving business impacts. Yeah, I mean, because we could have-- Hunter's there, look, every agent. Yeah, yeah, it is. The timer's in the way. We could have hundreds and hundreds of agents going out doing various sort of different things, but actually it gets a bit chaotic. And we're learning all the time, you know? So one of the first agents we built was an agent called Kate. And what Kate would do is when we close our contact centers at, say, 8 o'clock in the evening, Kate would then come and actually give two choices to someone running us, you know, a potential new owner. So like, high value customer, you know, I don't really run 24/7. So Kate would answer the phone and she'd be like, do you want to like, you know, we'll close, you know, want something to call you back tomorrow? Or do you want to talk to me, I'm Kate. I'm an AI agent. I don't think we use the word AI agent. We use a slightly more subtle, but we're very clear that you're not talking to a human. And then Kate basically has a job to, like, have a conversation, to learn more about your property, to give you some tips and advice. Her ultimate job effectively is to book a diary appointment in, to value it, whether we want the property, and then to assess it, and then to put it in the right position in the queue. So the humans come in in the morning to access it, you know. And that was a valuable role. But then, we built other agents where, I mean, one of the interesting thing about Kate, sorry, slight tangent, is like, people are having 15 minute conversations with this agent. It was quite scary when you were coming in the morning. So we actually had to tone Kate down. She was too good at her job. We had to actually, like, stop her, you know, like having random conversations with lonely people at 10 o'clock at night. And so she's now, you know, I'm much sort of a smarter. And interestingly, one of the things that we've done is we've tried to humanize all of our agents. So we have Hermione. That's the HR agent. We have Kate, who's, like, that voice agent for property recruitment. We have, then, Aidan. Aidan is a training agent that we use for M training, the, you know, training at LLN, effectively. We have Maya, who's, like, a digital personal, sort of, assistant, et cetera, so. But they all have to earn their stripes and so to speak. And we're always evolving these agents trying to make them better. We've talked about a bit of, I'm training already. Yeah, I mean, I think the start, probably, to lead people-- well, on this particular slide, is if you go to the one where we've got the 25% of my organization, it's every employee in my organization that we track. So it's about 1,200 people. Because I haven't found a role yet from a housekeeping team to use AI. If anybody has any ideas, please tell me that my housekeeping team, possibly, yeah, yeah. So we've gone from 28% of the employees using AI to 68%. But actually, more interestingly, the depth of how they're using AI is really deep as well. So you've got people using it, but also the depth of people using it. You have to have that uptick and adoption, and you get the efficiency gains. Yeah, I believe so, yeah. And we need to keep evolving the training as well. Because it's changing all the time. It's like, you know, there's new, smarter, clever ways of doing it. Could you get that if you weren't into this? No, I don't think so. How important is the CEO? I genuinely believe it's very important to democratize it, to lead from the top, to be saying it, and to surround yourself with champions who think live and breathe. And I'm very lucky, you know, my chief product officer, I sent her last week to the FT AI summit, and she came back like, you know, just like she said, at the best conference she's ever been to in her life, and it was life-changing. So that gives you an idea of how into AI she is as an individual. So you need a, you need a sort of a number of people. And when my proudest moments in last year as my CFO came to me, and she, at the beginning, she was a bit skeptical, and she basically said, "Grey, I've built a GBT CFO that, instead of the team asking me those are questions of disappointment to that, and they can like answer questions or responses, et cetera, you know, so really getting into it and understanding it as well. Yeah, I don't want you to be bamboozled too much audience. I want you to think, because we're, Graham is where we want to be in two and a half years time. No, you want to be there six months. What you can do between now and then, though, is a massive amount of automation using AI. You know, you can save yourself time and money and new hires, and you can drive a capability into your business to then bring you towards step seven, which is progressed towards agents. Now, we've been talking, you've been talking a lot about agents. Well, this is our big bet, right? Agents at AI and our business. So, and I'm really happy, because our investors who have retrieved, you know, like, all laid on this, they're like, you know, this is the future. So, what we're retrieving have sort of said, so it's like, go big on building an agentic platform. So, we're building this year, as they're working with Microsoft, working with different as a partner, and other people who are building this idea of conversational AI, because we think that way that people will be by holidays and the future is more conversational. Remember when you used to do a travel agent, and you're a really good travel agent, you said, "Don't have a conversation," et cetera, you know. And we've been guilty over the years of designing and developing products, where we sort of, you know, here's fatty thousand properties, you know, off you go, use some filters and so forth and read some reviews. But I think using AI and using this agentic conversational experience, it allows me to move my business from beyond a business that's very good, you know, and very strong at being a booking agent, to being a business that's phenomenal at planning with an AI buddy for a conversational agent, really good at the booking part. But also it allows us because of the capability to then bring it all the services to allow you to enjoy an experience, or because they can bring in, through my MCP layer, all the different data sources, all the different information, an AI, the agent, can make sense of and turn it into like a real transactional, conversational sort of experience to hand hold you through that. And that's what we're sort of building as a big bit, effectively, in how we think people will be interacting with us. So we haven't given you mind-blowing seven steps. They're not sort of changed the world's seven steps. But we believe you have done it, and it'll all pep talks is maybe halfway through those steps. It makes a massive amount of difference. So what's our parting shot to the audience in terms of what they should be doing next? They're going to have a conversation, 20 minute break out, and they're going to tell us how skeptical they are or not. But what is your parting piece of advice to this audience when it comes to adopting it? I mean, I think, I mean, just a curiosity, just take your hands if you're using AI, your organization today, or your organization. So I mean, you're already there. So it's not like we're saying 80% of it. Yeah, I know exactly. And every time I talk or have these conversations, people are already doing it, it's like just maybe trying to systemize it, trying to think of a framework, trying to put it into your strategic conversations, challenge yourself and say, actually, we're really investing enough into this to be understand, are we thinking about our data? And maybe role play, where do you think you're going to be in 12 months time? 24 months time? What does that future sort of look like? How can it really help you in your organization? Are you going to be disrupted or be a disruptor in terms of what you're ultimately planning to do? Make it possible. And learn as well. Like every day I'm learning, I think that's certainly an image of the books. Yeah. And all we're going to recommend is some recommendations. We're going to recommend some books and podcasts. Like every day I'm learning, and you just can't-- so it's just like keeping up with everything. And the hardwares evolving as well. So we've just bought-- and this is really geeky and old-- but we've just bought an AI super computer from the video. It cost me three and 1/2,000 pounds. Basically, it's a petabyte of computing capability, which is like a 1,000 trillion notions of AI capability a second, big numbers, big zeros. What it basically means is, for 3,500 pounds now, some of my guys have got an AI super computer, which would never be able to afford a craze computer in the future. And they're chucking elasticity models in and image more and do all that sort of stuff as well. So I'd encourage you to think about the hardware, like the AI glasses and where else. That's all going as well. They're not quite at robot level yet. But you can pre-order a robot. If you're interested, I can send you a link. Got a referral code. So I think about the hardware. But really, I'd just say, just dive in, learn, experiment, and have a bit of a play, really. Make it part of your job. If you find the talent, internally, the job will come a lot easier. Because you'll be riffing with them. We're riffing. And they then go away and do. And that's-- so it's not quite as difficult for the CEO, as you might think it is. And people want to do it. Yeah. They are doing it. They're doing it at home already, and playing with thing and having conversations. If you want to do it, as a 18-year-old daughter, it's constantly playing around. We're sort of AI building a thing, and she wants to be a non-driven owner, and she says, one of you have a job. Yeah. OK, that is a job. [APPLAUSE]