UP544 ChatUP: Craig Hepburn on how Agentic AI changes sports consultancy

Craig Hepburn sits at the intersection of enterprise technology and cultural institutions. He spent years as UEFA’s Chief Digital Transformation Officer, overseeing its digital ecosystem, OTT platform build, and Innovation Hub. He moved to Art Basel as CDO in 2023. He is now an independent AI strategist, Perplexity Fellow, and prolific writer on the structural implications of AI for organisations and industries. His Substack has become essential reading on the gap between AI hype and implementation reality.

Hepburn’s central thesis is that most people and organisations are “tourists in someone else’s architecture.” He draws a sharp distinction between using AI (prompting chatbots, generating content) and building with AI (constructing proprietary systems, workflows and tools). He argues the latter is what will separate winners from losers — and that the window for making that shift is narrowing fast.

Crucially, Hepburn’s argument extends beyond sport. His recent writing on “The Builder and the Billion Dollar Lie” contends that entire industries — consulting, systems integration, transformation programmes — were built inside the gap between the person who understood a problem and the person who could build the solution. Agentic AI, he argues, is starting to close that gap. That has profound implications for the agency model in sport.

Unofficial Partner is the leading podcast for the business of sport. A mix of entertaining and thought provoking conversations with a who's who of the global industry.
To join our community of listeners, sign up to the weekly UP Newsletter and follow us on Twitter and TikTok at @UnofficialPartner

We publish two podcasts each week, on Tuesday and Friday. 

These are deep conversations with smart people from inside and outside sport. 

Our entire back catalogue of 500 sports business conversations are available free of charge here. 

Each pod is available by searching for ‘Unofficial Partner’ on Apple, Spotify and every podcast app. 

If you’re interested in collaborating with Unofficial Partner to create one-off podcasts or series and live events, you can reach us via the website.

Transcript

Craig Hepburn 0:00

if you look at the technology industry specifically, if you look at Shopify I think Bloomberg even started to do this. Salesforce, all the big tech companies are opening up their data to AI agents because they know that's the future of their business. Like that's who's coming and accessing it. So to some degree the question is, is the sports world. the sports world thinking like that, or do they have the appetite for that kind of idea?

Speaker 0:21

Hello, Richard Gillis here. Welcome to Unofficial Partner, the Sports Business Podcast. Today we're talking ai and we're talking AI with Craig Hepburn because very few people know more about it than he does. He is a strategist and builder perplexity fellow and former Chief Digital Officer at Art Basil and at ufa, I think his output on LinkedIn and on Substack is pretty much unmissable. If you are interested in this subject, you'll have heard him at our live event at Fuse with 21st group before Christmas

Speaker 2 0:51

I Wanted to talk to Craig about the particular vulnerabilities of the sports marketing agency sector from ai, which is much talked about. Something that caught my eye. The advertising agency world is already feeling the pinch. 60% of US senior marketing leaders reported spending less on agencies in 2025 as a direct result of ai. I just wanted to know what he thought the future was for. Sector and in terms of what the response will be, what the particular vulnerabilities are, and what the opportunities are for growing a new it's interesting conversation, which uh, might surprise you. Certainly surprise me. Here's Craig Hepburn.

Richard Gillis, Unofficial Partner 1:34

I mean before we get going, have you noticed on LinkedIn a sort of uptick in people? cause you've really done that AI thing and the sorts of things that you are talking about are beyond. the chatter and the cliched stuff that you tend to find on ai.

Craig Hepburn 1:48

it's funny'cause I've learned a lot about myself over the last few years, especially coming outta like full-time employment in the last sort of year and a half. Because obviously sitting within a company and organizations, you're kinda restricted a little bit in terms of like,

Richard Gillis, Unofficial Partner 2:00

Yeah.

Craig Hepburn 2:01

because I've been in tech for over 30 years and I'm probably quite a deep system thinker like. From both our strategic commercial point of view all the way down to like, I mean, I work with CEOs and executives, but then I'm literally sitting in a room with a developer building like Linux and code. So I kind of feel like I have, I sort of spread a really broad spectrum and it's always really interesting when people have tried to put me in a box over the years. It's like, My brain sort of like racks up a lot of thinking as you start to kinda map the, I suppose to some degree the infrastructure, but then the scaffolding around it. and also, I think being in tech for a long time I loved it for like 10, 15 years and then I think for the last 10 years I've just felt really frustrated by it.

Richard Gillis, Unofficial Partner 2:41

Yeah.

Craig Hepburn 2:41

think I have a lot of opinions on what's broken about the whole technology space and

Richard Gillis, Unofficial Partner 2:45

Well, I remember you saying, when we talked, I remember you saying you are really enjoying building things again. Which again feels like there's a bit of the whole thing, which, not that I'm a builder of any sort, but I get that bit of it is really exciting when you hear people talking about, okay, well you can now whip these things up and try them very quickly. There is a sense of wild west to it, which I really like.

Craig Hepburn 3:09

I suppose, I mean, when you say you're not about, I mean that's the point, Richard, technology really in, considering how long we've kind of been at this game for like the last 50, 60 years, the idea is it should be democratized so that more people can, like, we said about you, the technology you're using now, if your podcast like. The idea is probably we've made technology super complicated for companies and people and that's because the business model is to kinda make it complicated. And and don't get me wrong, like. from a deep technical architecture point of view, there's a lot to unpack. But I'd say, yeah, I think now what. What people can do now is probably, most people haven't, haven't realized that the capabilities now not say that everyone's gonna go and build their own Salesforce platform tomorrow. But the fact is that, it's possible now for you just to build a quick product, a quick app, personal software that you want to use for the day and like. That kinda democratization of being able to kind of unpack that and build things is pretty profound. Like even for me and a lot of my friends

Richard Gillis, Unofficial Partner 4:09

Okay.

Craig Hepburn 4:10

in this for a long time that I'm working with, like, we kinda get excited, but we're also astonished by what's possible. But everyone's kind of got a framing and I think that technology is hard and it is difficult and that's the weird thing. It is. And it can be, and it to some degrees. There's certain industries that it that it should be. But there's also a lot of stuff that we pay for, we buy we think is difficult, that it's actually a lot easier to access. And I think that kind of, that horizon and that kind of idea of, to some degree we're making it more accessible to people and there's a lot more you can do it, but also it also like there's a lot of complexity to unpacking it as well. So it's kind of, it's weird. It's kind of a bit to some degree it's it's quite profound, and I think that's part of the challenge that everyone's getting struggling with a little bit in terms of what models can do and what they can build in the products and things.

Richard Gillis, Unofficial Partner 4:58

so the question that I wanted to ask you about, what this podcast is gonna be about is what this moment will do to agencies, the sports marketing agency, the classic sort of, could be a data agency or a research agency, whatever it is, the cliched view is, what. Agencies are renting people out to do now, can be done in-house at a fraction of the cost with a few smart people and a language model, and. I just wanted to sort of interrogate that a little bit and just go a couple of stages further just to see, well, first of all, whether you agree with that framing, and that's the direction of travel. A lot of people listen to this work for agencies, but it's also a question of, well, yeah, but what? Actually, it's never as simple as that. So just gimme, gimme your sort of sense. Gimme your first sense to it.

Craig Hepburn 5:48

Yeah. I think that is the, think that is the key point, like. Like any business, there's always more to it than just purely the deliverables. So I think what we look at, like we, we've always kind of, from an

Richard Gillis, Unofficial Partner 6:00

Comedy.

Craig Hepburn 6:01

even consulting, building, agency, software, SaaS, like all of it is the same. Like ultimately you're. You're buying, paying and getting access to a group of really smart people who've built and systematically developed something using their own knowledge and intelligence that they've been able to package up and sell. Because you haven't had the capability to do it. And to your point now, if you have someone internally that understands how to use AI well, the models, the tools the capabilities, they can get a lot of that, let's say that general deliverable, pretty much, on, serve to them through the models and the tools themselves. So you know, if you just take something as simple as a website, so even today in a website, you still kinda want to get somebody who's a good web designer, a good developer to build your website. Like even, you probably could have figured it out, but you still wanna outsource it'cause there's a lot more to it than that, managing it, supporting it, et cetera. So. That's something that, you know now with rep, with lovable, with all the different, even with just the raw model now, just coding you up and deploying it, you can build agents that will manage it for you. So there's a lot of capability. The question is just what's the appetite for a business to want to do that? Where I think it gets more complicated is where you've got a lot more value being, being delivered by an agency and a partner. So I think the key point really is what are you charging your time for? Like what is being delivered at the end of the day? Is it just the deliverable of a PowerPoint, a strategy document, a website, an app like we will. Payers to build these things and implement these things, that is still not gonna disappear completely. But I would say the cost and delivery of doing that is either the agency's using the model to do that quicker, faster, cheaper. And are they passing that cost on to the client or the client's gonna do that, as you said internally to some degree, maybe not all of it, but they might start doing 20%, 30%, 50% of it. And so therefore, where does the agency and the partners come in and add more value on top of that? And so really I think what it does is it kind of elevates everyone up from the pure deliverable of the asset, the artifact to what has been delivered. And it just means that everyone needs to sit in a room and have a conversation about what that looks like now going forward. I would say the, probably the biggest opportunity right now for agencies and and anyone in consulting and I think that's interesting. Software companies, consulting businesses and agencies I think are all ultimately gonna kinda converge in the same space.'cause they're all gonna have to deliver something a little bit more than they did before. And so the question is, a consultant firm can now build software. A software firm can now do consulting. So so an agency can kinda do all of it. And so you kinda realize like, what kinda gets unpacked out of that. And I think you're right, it is an interesting question. I would say data is gonna be a big part of that. And then I would say the other thing is obviously just having really good, trusted relationships with people as well is still an important factor. I.

Richard Gillis, Unofficial Partner 8:49

there's the sort of soft end, the people end and you're right in that. Ever since, I mean, I worked for a business in 2001 or whatever that tried to get away a sort of online marketplace. This was in dial up,

Craig Hepburn 9:02

Yeah.

Richard Gillis, Unofficial Partner 9:04

The idea, you sort of got there intellectually, but it just never worked. The tech wasn't there and whatever, and. But the argument against it was, no, it's a people business. It's about relationships. People buy people, all of those things. And so there's a sort of, that's where the fuzziness comes into it, and I think it's interesting. We're an interesting moment in terms of where the value resides in your organization and. One of the sort of things I came away with our event at Fuse with 21st crew was trying to sort of establish that what is the race? How are we defining what the competitive set is? Now what are we trying to do and what can, if we park agencies for a minute, what is it that they're gonna try and build? So it, what, where's the value in. The sports data market, for example, because if everyone can access it via

Craig Hepburn 9:57

Yeah.

Richard Gillis, Unofficial Partner 9:57

large language models, then I don't need someone to house that or warehouse that. And is that protectable, do you think? So if you are selling data to marketing companies, to brand clients, or if you are sending it to rights holders, is it defendable? Do you think is that as the data goes on, as the sort of period goes on, there's a question about what's the moat and what can you defend if you are selling information?

Craig Hepburn 10:27

Yeah. So I think that, I think go one level deeper. So I would say if you think about what, right, okay. What are the things that stay and what are the things that change? So I think if we, again, nobody knows, can predict the future, but let's just. Argue for the sake of it that, we've built all of our products, technology, and systems, not just in sport, but every industry is being designed for humans. So, in other words, all of our websites, apps, all of our engagement is being designed initially for people to engage with. So therefore, we built that kind of, that, that map of all of our technology for people to, engagement views click download your product. And also you build one product like a gaming app, or you build a fantasy app or you build, come get a product and you've built that and you deploy that to the people that, engage with your product. Or you build a website and you put branded marketing on there so that your fans or your, like, engaged with that. but what happens in a world where you know, and we're starting to see this, and again, we're super early, but you know. Where, you have an agent. So there's a big, kind of like movement at the moment as a company called Open Claw. So agents and now you're seeing open AI Anthropic all the companies are building like agents and what are they like, they're not, you get to proper agents, it's, I can then send it off and it can go and look for me for. find me the scores for the Champions League game last night. So I, I literally have one running on my WhatsApp, it's called Neo. I've built'em. I literally just ask him like, go and find this. Go and do this. I even could ask him to go and build me a fun app, on, and it'll go off and build me an application and deploy it in a few minutes. So. Agents now are the thing that every human is gonna have, or humans companies have an agent and they are the ones going off doing that. Now that might sound like, 10, 20 years away, but I would say it's probably a lot closer in the next few, 2, 3, 4, 5 years. as agents start to be the proprie, like the key person you engage with, are you building your product not only for people today to engage with, but you build. You're building a product or stack, or you're building data, or you're building information for agents. So as a simple example is you go to an agency today and you say, build me the Champions League app You Pay them whatever, a hundred grand, couple hundred grand, whatever. They go off and build. And that's a long process. Now internally they could do a couple of things. One, they could just have an ai, they could just build it through ai, so that makes it cheaper, more effective, or even better. Why don't you build a model of all of the data or build some kinda of API of all of your data, all of your content, all of the stuff that you have around your sport, your business. Connect that allows some fans or people to build upon the top of that and they can literally just subscribe to your API, similar to des, subscribing to a model, and then they can, a fan can build their app using, the data of your product to your company. So in other words, if everyone can now build and use an AI or model to build a piece of software, why don't I build Craig and build my own version of my app, using the UEFA for the Champions league or, the PGA's kind of application. So. That's really where companies, like, if you look at the technology industry specifically, if you look at Shopify I think Bloomberg even started to do this. A lot of the big tech companies are starting to deploy, and I noticed box Salesforce, all the big tech companies are opening up their data to AI agents because they know that's the future of their business. Like that's who's coming and accessing it. So, and so to some degree the question is, is the sports world. the sports world thinking like that, or do they have the appetite for that kind of idea? But it means like, my kids are 15 and 17. They spend a ton of time inside, playing with ai, with models, building stuff like they're, they're they love sport but they're engaging with it, as we know, through different channels. And they're also wanting to build things with products and services. So if you think about that generation, maybe, there's a whole generational change that we go through, but I can see that coming pretty fast. I think that's an opportunity. And so if you're an agency or a partner right now, it's like, actually, can you sit and build the agent frameworks, build all the application layers, build the APIs, build all the data systems, and actually capability so that you can still serve people and fans directly, but you're also building for the next layer of kind of the agent AI world that's coming over the next few years.

Richard Gillis, Unofficial Partner 14:36

Let's just pursue that for a second. because you get to control and releasing control, like you're saying, give the API to fans, give it to third parties and let them develop and work on your data. Sport, as you know well, having worked for UEFA is about control. It's about the right to do something. I've sold you the rights to do this, and I'm gonna police those rights from everyone else because that's what I've sold you. I've sold you monopoly access. So is it a technical question or do you think it is a cultural question for sport

Craig Hepburn 15:11

It is both

Richard Gillis, Unofficial Partner 15:11

both

Craig Hepburn 15:12

think it's both. I think it's, you end up where, strangely for me, I would say you. enabling, by opening up a data platform that people can build upon, agencies, fans, everyone can have access. You actually take back more control because you actually build the infrastructure, you build the rules, you build all of the, you can set all of the kind of the rules and regulations around how your data and APIs can be. By anyone. So you can kind of define that. You can monitor it, you can manage it, you can monetize it, you can build new products and services on top of it. So you actually get a lot more control from a technological point of view. And then purely from a philosophical point of view, it's like, we kind of, there's a kind of IP legal control framework thinking as you said, like contractually what you give people contracts to do. But as we know in the world of technology especially, information, data leakage, all of that, people are, the models themselves are training on things. So there's kind of, there's a lot going on anyway, so I'd say actually by building your own kind of like infrastructure and technology. You are actually making a step in and playing and actually strangely making a sort of play towards, we're gonna play the technology companies a little bit of their own game by actually building more of the infrastructure ourselves that we can actually. To take that a little bit more management and control of I mean, if we argue like the world in the next five, 10 years is not going to go backwards, technically, the technology and digital is not going to go in reverse. I, it's just going to compound and move forward. And AI is not going to just stop and kind of end. You could argue every company, no matter if you're a small, independent cafe all the way up to being a large global conglomerate, you have to, you're in a technology business because every business is being accessed to some degree through a technological platform. Like, that's not changing. That's probably accelerating. So it might make sense to actually hire and build some of that capability in house. It doesn't mean you do it all. You still have partners and agencies, but thinking more as a technology company. And as said, this is my argument for a long time has been the models like, take Facebook and Google and Amazon and all the big technology companies, look at the amount of value that they've extracted from the world. Now look, now go and look at how much every other company or business or vertical is extracted out of the world. Look at the difference in the gap. then maybe the question is, should we not be maybe thinking a little bit more like a technology company if we want to get some of that value back? And also for me personally, I also think it's just about that democratization. If we don't want four or five big tech companies, pretty much extracting all the value. I said this when we were together at the podcast, we online, then there's probably, you would argue there's probably a bigger opportunity for everyone to build more of the technology stack themselves, or at least invest in some of that capability, leverage the models, the ai, the intelligence, build into infrastructure. We don't know how it plays out, but at least it gives you a lot more control because at some point when you realize like the your world is changing like really quickly, it's, it is quite hard to catch up. But if you do it early, at least you kind of, you're kind of like. You're moving towards where the puck is kind of heading towards. So you're kinda like, you're giving yourself a little bit more, so my argument on the control thing is the biggest opportunity to take back control of your business and your future is actually to build more of the technology and systems thinking. You're not all of it necessarily, but at least to start thinking about it.

Richard Gillis, Unofficial Partner 18:42

so. the big foundational models, you the very top end, sort of Google philanthropic back group, open AI chat, GPT, all that. They are thus far scraping the public internet or have scraped the public internet. And they are now looking at. The rest. So the data that resides in companies and organizations, and I'm trying to, what sort of proportion of that, is that the, how much is there still for them to gather? How are they going to get at it? And in sport, one of the conversations we've had in the past is sponsorship. The route to getting at. So when, a league does a deal with Google today, what are they giving away? What's at stake? So just

Craig Hepburn 19:31

Yeah.

Richard Gillis, Unofficial Partner 19:32

I'm just trying to sort of, there's a general thing. Let's talk about the general question about what's still available. What do they not use? So when I go on chat or when I go on Anthropic or on Claude. That's accessing the public database or the, on the internet.

Craig Hepburn 19:48

Yeah. Okay. So to unpack that, yes. So all the models have been trained on the open internet. so the corpus of basically all texts written that's accessible online plus. Books and videos and images. So, so all the models, you're right, is kind of being the models are being pre-trained on all of that. There's also something called pre-training and kind of reinforced learning, uh, reinforcement learning. So there's kind of two technical aspects. So what the model is trained on, as you said from that, and then what it can get afterwards to add additional context. And there's a whole range of technical terms rag.

Richard Gillis, Unofficial Partner 20:21

Yeah.

Craig Hepburn 20:22

these other things, but I think purely the model's. Yes. So they've been trained on that open content. then have the products themselves, so GPT, uh, Claude Google, Gemini, they are like they have both two, they have two things actually, which is interesting. They have the raw model themselves, and they have the product that you work within. So when you use GPT. that's a the chat GPT, that's the product itself. If you use a free version of that. There is, and there's, I think there's some configurations. You can turn them off for training, but ultimately, there a lot of that data leakage does not into the model itself directly When you chat to GPT, the model itself is not training directly at that moment in time on that data, what you're doing is storing that data within the product stack itself, and there's leakage that happens into the product through that, especially if you're using the free models. When you have an API, so when you, A-P-I-I-E build a product off the back of maybe cloud or GPT or Gemini, what's called an API and you pay for access to that API, that the terms of service, if you read them, they do not train on your information. They do not store your data through the a p. So, if we believe them or not, but that's in the terms of service. Where, when you have an enterprise version of Claude or GPT. So in other words, if you're a company right now let's just take, was working, I'm working with a business at the moment. installed, uh, Claudes or Andros, kind of Claude model as an enterprise, and they paid the money to have that enterprise version. of the training of your company information is switched off and it, all of the terms of service again, means that we're not training on your data. What they do is they build up databases around that to organize and manage context. So the best way to leverage the models themselves is generally through the APIs. And you can keep and the way that it's architecturally set up as well, you can set them up in such a way that you can. Access to intelligent models, but keep all of your data completely separate. And so that's why a lot of companies are building proprietary pieces of technology around the model for companies. Again, from an agency point of view, big opportunity there to really understand data models, training, help companies understand exactly what that looks like. There's a lot of myths as well in businesses. There's a lot of misinformation and the lack of technical knowledge. But I don't wanna go too deep, but what I would say is there's a huge opportunity for agencies and companies and every business to learn a little bit more about that. to your point then, how does all that information within those business kinda get unpacked? You're right. I mean, at last count it was like a very small percentage. I don't know what the number is now, but it was maybe like less than five, 10% of like internal. Every company that owns internal intranet and all of their internal databases and knowledge, none of the models have really get full access into that. And that's really where the kinda standoff is at the moment between allowing AI into businesses. I know a lot of sports organizations and a lot of, any organization has put a lot of restrictions around AI within the business, and rightly so, because they wanna make sure they're protecting that data. is though, the tension becomes. You are gonna have to go over that quickly. You're gonna have to either A, understand that, understand what the technically it looks like and work with people to help you unpack that. Because just staying still means that someone else, another business I think it was it Redding FC just hired, I dunno if you had Redding FC

Richard Gillis, Unofficial Partner 23:49

Yeah.

Craig Hepburn 23:50

head of ai.

Richard Gillis, Unofficial Partner 23:51

Yeah.

Craig Hepburn 23:51

Yeah,

Richard Gillis, Unofficial Partner 23:52

I think that's right. Yeah.

Craig Hepburn 23:53

So, but I, again, probably, again, having someone inside the business really helping unpack like where we can take leverage some of the internal data to build new products and services. So I would imagine he's looking at all their internal data and saying, how can we build or train our own? I don't maybe even fine tune their own models, use APIs of the of the major models, the frontier models to build products and services. He would, I would imagine, would have quite a good knowledge of how that works and therefore they're gonna start

Richard Gillis, Unofficial Partner 24:21

So the.

Craig Hepburn 24:21

really interesting moves and build some products and services that probably, move them forward. I think the internal. The, yeah, I mean the big play for all of Anthropic and OpenAI, even more so now is if you look at the investment they've put into all of their models and their products, they're all playing for the enterprise. They're all playing for the big companies because in order to kinda get, even recalibrate some of the numbers and that investment, they're gonna have to like, play the, those bigger, revenue numbers. And that means that how much tokens are you using, how many tokens are you using per month? I've got a couple of companies that are spending, hundreds if not thousands of pounds a month on tokens. So that is money that's going directly to kind of, into the. models themselves. And so there's a big kinda play right now to how do we become valuable for you as a business? Harnessing our model with your data to build new tools and systems or commercial opportunities for you. And really the money is made by the API calls and the token usage that you use for building that product. Again, that goes back to my point around understanding the technical architecture, building the right systems so that you can also, potentially make some revenue on those as well, and building those new products. So I would say you're right, the big revenue play will be. and I know GGPT, well, I know for example, I think anthropic lose money on the 20 pound a month subscription for users. So they really want the enterprise play, and I know GPT are now starting to move more towards building their intelligence models for enterprise. So what I would imagine, Richard, is over the next six to 12 months, you're gonna see more and more products come out from the companies. That are specifically designed to help enterprise businesses take advantage of the models inside the companies. So the question is do you sign up and kinda like license those? Those products that they're building for you. But the interesting part is you don't necessarily need their products to do the things that they're offering. You can do a lot of that just with some smart developers and some APIs. So I think that, that's gonna be where the tension lies at the end of the day. Do you wanna subscribe to their product to, you just wanna use their API to build products on top of their stack?

Richard Gillis, Unofficial Partner 26:26

And I guess the agency question that we started with is that there is the gap and there is the opportunity between the sort of universal and the specific, isn't it? So you are saying, right, okay. Here is a space that sport has or requires a particular set of, I sound like Liam Neeson, but a particular set of skills and insight and intelligence, but also the knowledge of on the sports side. So that's where. That's what an agency is really when you boil down to it. But the people will be different. Presumably, there'll be a, there'll be different types of skills that you require inside the agency, but the agency as an idea will still exist because it's a different shaped thing, but it's doing slightly different jobs or very different jobs. But there is still a market there for advice, but also building.

Craig Hepburn 27:17

Yeah, so exactly that. So the mo, so if you say, going back to original point about agency, so before, we provide strategy. Let's just say we create assets like websites, applications. We build data systems like all of the things that they build. Now, imagine. What they deliver now is, agencies themselves have people who understand how to harness the models, understand the architecture under, a good example is, I've got my son who's a, who's not got a huge amount of money to spend in APIs, but he does a lot of building. With stuff and a lot of his friends, so they don't have a huge amount of money. So they've become really experts in open models. So open weights, models, uh, open source models where. It's a product, there's a platform called Open Router. And so they can go there and they can find the best models at the cheapest cost to deliver the best value and the best outcomes. So what code's the best, what can build the right tools, what products and services they can get. So they've become really expert in understanding the models. So I actually, I've actually, he's been advising me on some projects because he understands the capabilities, the models. He also knows how to use things like cursor. How to work with rep, how to, he's able to as a, as an individual, build full end-to-end stack products so he can do the strategy, the research, build the product, build the PRE so he doesn't deliver just pieces of the pie. He can do the whole thing

Richard Gillis, Unofficial Partner 28:35

You start renting him out, Craig.

Craig Hepburn 28:37

well, I'm keep now his mine. Uh, yeah, like at the moment, uh, his mom's like, he's at college. He's at, uh, he needs to get to do studies, but. But yeah, you're right. I mean, to be honest, it's like these young builders are learning what's valuable and they're seeing that actually being able to build things develop products, build systems, understand how to use APIs, is kind of a skillset. As well as we've went through kind of an interesting sort of like, time this year. We're trying to identify the right skills and what we've found is, the people who are very kinda broad knowledge, very good problem solvers are very inquisitive. They are open-minded, they are very technology like they, they enjoy working with technology. They really get into playing with the models and the AI tools and systems. They've started to been able to realize, to our point earlier, they've been able to build things that they didn't think they were able to do. They've kind of realized that their skillset, which was kinda generalized. Now instead of relying on developers, strategists, researchers, they can actually do a lot of the work themselves. I wouldn't say they can do all of it, and there's still a lot of edge gaps that you still need expertise in, of course. But what they're saying is they, we can now do a lot more of the work And be a lot more autonomous. And so if an agency model, you would think they have creative designers, researchers, developers, builders. All of those skills to some degree still matter. It's just they matter at the edge. They matter at a smaller percentage or where the value lies. And so now it becomes like one or two people can do the work of five or six, but you might still need a couple of people to validate very important aspects of that deliverable. I would say the other thing is testing validating outcomes. the most amount of work we see now right now is not building and developing. It's doing a lot of the evals, researching the outcome, the validating that you know, what it does is right and correct and accurate. Making sure the information it's using is good. Making sure that the platforms you build deliver the right outcomes. So I think a lot of the work now is shifting upstream a little bit or downstream or however you wanna put it, but where it's, the work itself was where you spent all of the energy. Now the work is. You can do that a lot quicker. Now the value moves to how can we use that to be a lot more valuable? How can we evaluate

Richard Gillis, Unofficial Partner 30:54

it, checking that it's right, that it's not, yeah.

Craig Hepburn 30:56

Yeah. But more than that, not just checking it's right. Is this the right thing that's delivering the right value? So let's just take an example of building an app would take three months or six months. The bell quite expensive. You deploy it and realize. Ah, we made a cut. The wrong decisions. Our hypothesis wasn't right. This doesn't work. Nobody's, it doesn't really work the way we thought. Now, imagine you could spin up four or five or 10 versions of that app in a few hours or a few days. Deploy that test and learn, figure out like what's working, do the evaluation, run lots and lots of hundreds of kind of, research trials or research on that and self-improve the product. can do a lot more work on that. And then what you're doing is helping the partners or you're working together to get a better outcome. So whereas before, once you've done all the energy, put all the effort in, deployed it, it's quite a lot of work now to kinda change that whole thing and do it all over again. So I would say that there's a lot of value there and moving upstream a little bit to be a little bit more valuable there. I'd say that's a huge opportunity just now.

Richard Gillis, Unofficial Partner 31:58

On the sports side. So you look at one of the questions in the previous era, we now call it, the sort of social media era and how they relate to what they do with their data, all of those things. And one of the answers was always, one of the sort of questions people would put forward was pooling. So football leagues, for example, share a whole load of similarities and pooling their data when they're getting into bed with sort of various third parties. Whether that makes sense, because actually individually you'll only get so far because it's your small and you haven't got the capability, you haven't got the money, frankly, so one of the questions is can they pull together that data back to the, that question about the sort of open versus closed data that the, uh, the models want to get hold of. You can sort of start to see, well actually there's a sort of sport ink answer. Because of the point you put about the last generation of tech platforms has just been so extractive,

Craig Hepburn 32:55

Yeah.

Richard Gillis, Unofficial Partner 32:55

In terms of just taking all of the value from the marketplace. I wonder if there's a way, there's a sort of, don't get fool again type solution. They're coming around again with a different product, with the same expectations, presumably. But I'm wondering if you would advise Sport Inc. To do something different this time around.

Craig Hepburn 33:15

I think that the thesis makes sense. I mean, there's a lot of, probably a lot to unpack in terms of different sports work. I mean, everybody's

Richard Gillis, Unofficial Partner 33:22

I re I realized, Craig, that it's almost impossible to get sports bodies around a table and agree on anything. So this is a massive hypothetical, but I think it's worth asking that question because actually. You can feel it happening again, can't you? That everything will just be given away.

Craig Hepburn 33:38

yeah I think that's be because it looks like. Inde independently. If you're kind of individually having a conversation with, like we've had,

Richard Gillis, Unofficial Partner 33:47

I.

Craig Hepburn 33:48

with social media, same thing. Let's have independent conversations as, independent sports or businesses. And then you kinda, like, you look to Facebook, YouTube, Google, and all those companies is a way to drive you traffic engagement and value. And, and we all went through that. I worked through that in all of our companies that I was involved in over the years and but I think the biggest difference now is. We needed a lot of that technology company because they were building these, building these systems. And as I said, I think that the, to some level that the playing field is now more level. So you now have access to be able to develop and build more of the things yourself. The biggest challenge is the network effect because I think what those big products and platforms provide you is a massive network effect. Now your point, I think makes sense. Now, imagine Sport Inc. Got together and says, well, actually we, yeah. We could start to form. We have a lot of data, we have a lot of context and you, we hear this a lot in ai who has the context and proprietary knowledge and data that makes everything valuable in the first place. Sport has that in abundance, right? So every single event, every game, every experience provides a vast amount of incredible data even getting into what every athlete has in terms of. Their data and what you can do with that to build products and services. So, you can imagine the unbundling of. Data into some form of product, into some form of new value opportunity for every fan that can engage or buy or rent or subscribe to the sport. So imagine I now subscribe to, the Premier League I subscribed to the PGA, right? I pay them a subscription for their digital products and services. But the difference was they're now able to build, manage and to some degree kinda get together and build the right context. So. That's why, to be fair, Google meta, all the big kinda social networks, I mean, put YouTube to one type of a video point of view. But if you think about it, like value is becoming the, or the real value is the intelligence models and products connected into your data, that becomes really valuable. And so the question is like, what does that network effect look like? What does the commercial opportunity look like? And and how does that kind of like. Push some of those social networks to one side. And again, I think we're early. It's really hard to kinda see how this plays out, but I would imagine, there's going to be a whole new there'll be a whole new operating system that will be built on top of what we have today. And the question is have, can sports organizations, as you said, get together and start to think about that more broadly? And, and as they get a lot more mature with technology and digital, which they all have done, right, they've all invested a lot of money and a lot of technology. Built a lot of, hired a lot of, great talent into their businesses. Now the question is, do they just partner with third parties to build everything or do they start to think about building some of that themselves? And as you are getting together, the opportunity's great. And to be honest, it's not binary either. You can play both. That's the beauty of it right now. You can play, start to build things at a cost. An ease of build now that's never been available before. And you can still play into the existing business models. And so there's a lot more leverage there and a lot more opportunity. But yeah, I think you're right, Richard. I think there's, that, that's the thing. The overhang is really interesting. I. was chatting somebody the other day. What people believe is possible in terms of the capabilities. If we stopped every, stopped building anything today for the next five, 10 years, you would still have a huge amount of capability that people could build versus what everyone knows is available. And so I think the difference between what. Individuals, sports organizations, executives understand the model and the capability versus what they could be doing. There's a big gap and the big opportunity right now is to try and fill that gap with knowledge and information to make better informed decisions. And I suppose that's kind of what, to some degree, you are doing with this podcast and we're trying to kind of unpack with people and give them that information.

Richard Gillis, Unofficial Partner 37:36

Yeah, it's also again. Second guessing where the value of your organization is not just today, but further down the line. And I remember talking at our conference where the question was Google and does Google have a strategic advantage?'cause it's got YouTube. And in terms of if the future of robotics is modeled on Premier League. European footballers.

Craig Hepburn 38:02

Yeah.

Richard Gillis, Unofficial Partner 38:03

okay. I could understand that. In which case, what are you selling to Google? If you're getting, if you are doing deals on YouTube, if you're doing deals with Google today, what is it that they're using? How are they viewing? on both sides of the. table, what is it that they're looking at? Because when I've heard people say, right, okay, Warner Brothers discovery, that's all about ai. So again, a question there, I've no idea whether that's true or not, but it immediately changes the lens on why companies are buying other companies, what it is that they're seeing and training data. So it's, fundamentally, is that just the game that is being played? A level of, billions of quid.

Craig Hepburn 38:42

Yeah. Yes, I'd see. The training. Yeah. I mean, the Google thing. Yeah. So training data. Yes. The Google thing's really interesting because I think they have they don't just have YouTube and all the products they have, they actually own the whole stack, which is really unique. So they own the TPU. So Nvidia sell GPUs to all the other companies. Google have what's called TPUs, their own, essentially their own chips their own right into the hardware, right into the metal. have the metal in their server infrastructure all the way up to the model and all the way through all the training data. So all of Google, all of YouTube. All of their products, all of their search information, all the way up to obviously Gemini Genie what they're doing with DeepMind, like they have an incredible stack top to bottom. So I would say Google are in an incredibly strong position, unlike OpenAI. Have to rent all their, the cloud, all their infrastructure. They have to find training data. They have to rent or buy all the GPUs. So I'd say there's a kind of thinking about it from a stack point of view who owns all the layers of the stack. And to your point, how does training data become valuable? Where, where you have access to the technology, the models, the tokens, the compute, because it comes down to really some simple dynamics. Who has the compute, the chips, the tokens the AI models that are capable enough to ingest knowledge and information to then generate products and systems that people want are to generate some kind of outcome. So, so the more so to some degree, anyone that holds huge amounts of really valuable proprietary knowledge. It a lot of leverage now because if you can figure out the right products in the right monetary or the right commercial model to build on that data, then you can either, partner with the right people to kind of like access that. But again, you can access the APIs in the models to build some of that with compute power. The one you said about in Hollywood, I think that a lot of it was. I think the vision is, I think Elon said it the other day, in a few years you'll be able to just ask your TV to generate you a a 30 sec, a 30 minute show on, this subject with these character, and, a lot of people are starting to sell. I think a lot of the, not a lot, but some Hollywood actors are starting to kind of. I think Matthew McConaughy and a few others are starting to license their data, their voice, their image likeness, two models who can train on it, and then they get paid for that information. imagine in sport can happen there with all of that information about. Being able to replay an old match, create an experience, build things like, there's so much potential there. But ag again, the question is can they get together with the right data, put it in the right structure, have the right teams that understand the opportunity and the value of that opportunity, and then negotiate that in terms of what they can do with it. But yeah, I think training data is. It's a big thing. Yeah. That, to your point earlier about enterprise, who can who can monetize that training data and that information and create new products and services. I think that's definitely a big area, and again, agencies that start thinking about that and helping people figure that out is, it's definitely an exciting space to be in.

Richard Gillis, Unofficial Partner 41:48

Brilliant. Okay. Right. We'll talk again at some point in the near future.

Craig Hepburn 41:53

We

Richard Gillis, Unofficial Partner 41:53

Thanks a lot.

Craig Hepburn 41:54

a joy. And I do like, I do like you're tapping into all the right things. It's kind of all the questions that really I think people are trying to figure out right now. So,

Richard Gillis, Unofficial Partner 42:01

So, uh, just about grasping the questions, let alone knowing any of the answers, but, we'll get to the answers in due course, thanks a lot, Craig. Really enjoyed that.