UP528 Chat_UP Live - The Race to Build SportsbizGPT Pt1

0:00Hi there, Richard Gillis here. Welcome to Unofficial Partner. So this episode and the one on Friday is derived and based on a live event that we did in December, just before Christmas, in partnership with Fuse in the UK and 21st Group. Like you, I've been wondering what to do, what our position at Unofficial Partner is in relation to AI, to give it its banner headline. But the things like large language models, co-pilots, foundational models, all those questions that we're all sort of bombarded with. And at the top end, quite often the conversation goes to the big foundational models, ChatGPT, Anthropic, Gemini, which one to go with, is it a bubble, how much of our money is going to go with it.

0:41All of those questions feel out of reach, but I'm very aware that there is something, there's a race going on within the sports business bubble, and there's going to be winners and losers. And I thought it'd be interesting just to try and do something different. One of the aims was to just push a bit further than the usual panel chat format and to solve one of the problems of the conference format genre more generally, which is that we forget about 90 % of what was discussed. And we have little way of interrogating the views and opinions of those on stage beyond the Q &A, which is a sort of flawed device quite often.

1:18So we, and I talk about we in a moment and who we are, we created a second brain, which I sort of co-pilot for the evening that captured not only the onstage conversation, but it took in audience questions in real time. And then I think this is the really smart bit, offered up a range of evidence based answers and then extrapolates on them to push the conversation forward in in pleasingly sort of non-obvious ways. You'll hear us mention the visualisation behind the stage, which lit up sort of common threads and there was a bit of theatre involved. And again, I'll talk about how and why we did that as we go through.

1:59And I think this sort of thing will, I think, quite quickly become the expected product in the conference world. It wasn't perfect. It was an experiment. And it was all the more interesting and fun because of that. And I've not seen it done in the same way. And it goes without saying that to achieve that level of sophistication requires deep tech and sports sector expertise. So we went to the smart people at TFG Labs, which is the offshoot of 21st Group, who essentially created the thing that we're going to be talking about. It was a really fun experience and I hope that comes across in the podcast.

2:35And I want to thank some individuals before we go any further. Helen Burford, who has been really supportive of the idea and instrumental really in making the whole evening work as well as it did. The venue was superb and we're hugely grateful to the team at Fuse, including Louise Johnson and Monica Conway and their team, in particular Zainab Zaman and Annabelle Wilson. Thank you very much for your time and effort and hard work. And you're here on the podcast, Lucy Basden-Smith, Managing Director of Fuse International, who took part on stage with her usual intelligence and smart insight. Sean Betts, who is Omnicom's lead when it comes to AI, features heavily in Friday's episode, which is more of a deep dive into AI and its impact on the sport and marketing sectors.

3:21And we'll talk about that separately. our other partners were 21st Group so again a big thank you to Blake Worcester and his team including Omar Chowdhury who you'll hear later on one of the panels Andy Shorer who built the model for the evening again you'll hear through the podcast and also Dan Zelizinsky hope I said that right Dan Chris Woodcock and Conal Milligan amongst others so big thank you first up you hear myself and Andy Shurer just set up the first panel but also if this topic is interesting I really recommend Friday's episode because again it builds on this session but also takes it much deeper into more of a conversation about AI and its deep implications and it gets back to that question about who's going to win the race what are the characteristics of the winners in this race to build what we might call sport business GPT or Bloomberg for sport who knows but keep listening it's good so I came with a really sort of ill-informed basic idea about okay well what could we do and then I spoke to Andy and every time I had a meeting with him he said well we could do this we could do that and I said well hang on a minute you can't do that that's ridiculous So it just became, it just got a bit of momentum.

4:43And I'm enormously grateful to the work you've done. Let's just talk about, first of all, TFG Labs, first of all, and then we're going to talk about what this is and what they're going to do. Thank you, Richard. So I'm Andy Shorer. I'm head of TFG Labs at 21st Group. We are essentially the innovation arm of the company. We experiment with emerging technologies and introduce them into the sports industry. Okay, right. So let's talk about what this is. What are we trying to do here? There are two phases. There's a first panel, which we'll explain in a minute. And then there's a second panel, which is more about AI itself.

5:17But let's talk about what we're doing here. What does this do? So we've designed a system just for this evening's event where we wanted to give you a kind of idea of how we think about AI. And that is not really a chatbot that's going to spew out wallpaper responses. it's not going to be a useless summariser of an evening that you're going to get after the event although I don't know if you're going to. I'm a useless summariser by the way don't come for that gig. So you can ask questions or send in concerns or ideas about the topic we're exploring tonight and this isn't just going to send back a response from an LLM that's trying to please you we have a number of principles that we follow where we break down problems we want to point AI at the right data to solve a task and not just respond with pleasing language.

6:09So the conceit, okay, we needed it to be about something. So it needed to be about something in related to the sports business. So what we've landed on is we're going to buy a football club. So we're going to look at the market for football and the football clubs. And the first panel is just going to sort of pick apart the sorts of questions that would be useful, would be valuable, and interesting. And the ultimate question is, well, can AI help this? And is it going to actually improve or just make things faster? How can the audience get involved? So you should be all within close proximity of unsurprisingly, a QR code, which you can scan.

6:46If you send an email, click on the link, you can submit your question. I'm sure probably more emails are going to get queued up than I've been testing with late at night. But within a few minutes, you're going to see your response on the screen. And you're also going to get an email back, which contains a lot more information than what you're going to see here this evening. It actually contains strategic advice that's from our private company archives, which hasn't strictly been signed off by Dan, our CCO, but I've decided to give it away. Since like all of the sensitivities removed, of course, and you're going to get back some of our proprietary data.

7:26So some of our models, which might have club valuations, club ratings for this evening's event only. Okay, and what is this? Right here you're seeing questions being classified into six different areas. Someone's just asked a question about fan engagement, for example. So we might see the yellow cloud burst in activity. We're going to hopefully see towards the end of the evening the brightness of this visualization increase, which will represent the interactions that we've had here in the room tonight. Right. And recency affects the speed of the clusters. So we're really hoping that the topics that the panel are talking about will be soon represented behind them in terms of activity.

8:08OK, right. So find that thing there, point your phone at it and just ask whatever questions you like. We will again, it's an experiment, this. And I want you to sort of come with that frame of mind. They're taking a risk here. It's OK for me and Sean, unofficial partner. failure is sort of embedded in the DNA of unofficial partner disappoint vague disappointment is what we stand for but other people you know are serious about this stuff so take that where it could fail there's a nice element of jeopardy to this and I want to just thank Andy again while he wanders off to whatever that whatever's going on over there dictionary corner I think dictionary corner perfect right round of applause for him and the football group can we get them on so joining me We have Maggie Murphy, fresh from Journey from Aston Villa, who's the managing director of Aston Villa Women, as everyone in the room knows.

9:03And we have Lucy Basden-Smith, who's managing director of Fuse International, has a shorter route to the stage. And we have Omar Chowdhury. Omar, I want you to set us up here because the job of valuing football clubs, can you just give us a frame this for us? Because everyone in the room will be saying, OK, well, it depends on who the investors are. It depends on the sort of types of the framing of what type of club it is. But just give us an idea and then we'll throw around some question ideas. Yeah, I mean, that's exactly right. The valuation comes from the investor in the first place around what their overall thesis is.

9:40And you can start with a thesis of, for example, wanting to find an undervalued asset and seeing that there's potential to take advantage of natural growth in that. That might be a Sheffield Wednesday, for example, today. You might have an investor who thinks about investing in kind of efficient, well-run assets and can therefore ride the wave of a good management team. You might have an investor thinking about a completely different business model. That might be women's football. That might be a player trading model where I'm going to buy a club in France and be able to connect to clubs in Belgium and England and sell players.

10:15and whatever that thesis is you then assign a value to that based on where you think you're going to be able to appreciate that asset value over the course of 5, 10, 15 years work out how much money you're going to lose on the way because football clubs lose money and they pretty much all lose money men's and women's and if you can get that equation to add up then you know you're onto a winner. Okay so we're looking for sleeping giants Sheffield Wednesday or we're looking for for not necessarily lower league clubs, because we could start to say, right, okay, is the Premier League a good place to invest?

10:49It could well be. So, you know, if you take, I don't know, a club like Arsenal today, I think if you, let's say the Cronkies were interested in seeking minority investment to support them in a different part of the business, we believe in sporting performance kind of underpinning valuations of particularly men's clubs. One of the things you might want to know with Arsenal is their success recently how much of that is Mikel Arteta they found an amazing coach and therefore when he leaves the whole thing is going to collapse or they've built something more institutionalized within that and so do we know the answer to that question well that's one of the things that I guess we try and find out and maybe some of the the kind of AI can help us understand and understand how does a team's performance vary based on different circumstances different managers different ownership etc because that again if you're an investor coming in and you're looking to invest in something that's already kind of working then having a kind of feel for that is really really important okay maggie when i know you know this but you were ceo of lewis and that when you went through a process of someone trying to buy you you had mercury 13 coming in and looking at the club it didn't happen but you through that process what did you learn about this question about what the money wants from a football club well first of I think the money is different.

12:08There you've had roughly three motivations. Financial return. I feel like there's something in there around ego. You think you can do it better than someone else. You've got a better idea. You know what to do. And then the third one might be about reputation. Something like that. You're actually interested in investing into a big shiny club. Don't really care about the return. And it could be a sports washing on that end of the scale. Or it could be just you really love to be aligned with a nice London club. So I think there's three motivations. And then that doesn't really matter if you don't have the cash to fulfill those ambitions.

12:40You can ramp up the cash, you can ramp it down or you might not be able to ramp up or down. And then the control element as well. There's no point you wanting to go and control, have a majority control if you don't have the cash or if you don't have the ego. If you don't, you know, that's it's like cash. I think you can ramp up and down control. You ramp up and down. and then maybe level of certainty you ramp up and down because you might want to go into a very, very stable club or you might want to go straight into a top tier if you want to be more sure or you want to go into a closed league if you want to be more sure.

13:12But if you're happy with the level of risk and uncertainty and jeopardy and excitement, then you might be more likely to be in a free-flowing position or be lower down and think that you've got scope to grow. So I think just on the very first basis, the money is different and I don't think this consortium here has come to an agreement. I'm not sure in the drinks whether you did as to what kind of investor we are. OK. If I am a sort of hard nosed American private equity, this is sort of method acting here. But what what will I then want? I imagine you want something that is probably for the future as opposed to the past.

13:51So maybe am I right in thinking that you'd probably want something where they have an amazing digital infrastructure that then bleeds into a really kind of active fan group you'd probably want a good base of sponsors and maybe a famous I imagine you'd want a famous football club that's going to give you a big brand with security through merchandise and the commercial and a maybe a pretty good commercial based to it in the first instance but then would you also want like how embedded into the community would you want it would you want with it like lots of land are you just looking at a team with a big stadium and a big fan.

14:28So you're getting to quality of the operations. Atletico Madrid springs to mind here because obviously that's a recent case and one of the stories running is that it's very much a property buyer with a football club attached that might be doing Atletico down a bit but that's one of the readings of it. Yeah and I think the really interesting thing with private equity is you know you think about the life cycles of funds you know five to seven years doesn't tend to work in a football context because there's so much uncertainty that exists in any given year you know if you were to you know try and understand what's the probability of Spurs making the Champions League each year it's probably never going to be above 30-40 percent and so some years you'll you'll be really high and some years you'll be really low and um and so with Atletico Madrid and you look at private equity investing in Chelsea there's an element there of trying to pay a premium for a level of certainty in that sporting product because you know, Athletics have probably qualified a majority of the last 10, 15 years.

15:26So that's one of the things that private equity is looking for is certainty in a pretty uncertain sport. So a bigger club in a less competitive league is a better bet for private equity because you can guarantee, they can guarantee Champions League football more than Spurs can because as you say, we're mercurial that's the yeah I mean even a club in Denmark like an FC Copenhagen or a Belgian club like Anderlec at Bruges have got uncertainty around their revenues and around their performance from year to year like Champions League for them might be worth 30 million one year and then Europa League 10 million the next year and that's a big variance Maggie there is a button there in grey community what does that mean and where is the value do you think if you under that enormous heading of community what does it mean in football terms and in terms of valuation yeah it's interesting coming in and seeing and i don't know what's going on behind my head with this big visualization behind me but i thought it was interesting because that's the first time i've seen it just so that everyone is aware and those kind of six categories are slightly different from the categories i would have chosen when thinking about building a football club but maybe just because i need to add a couple of categories we're not going to put them on the ball that's fine we don't know we don't have to change everything just for me um but like say take um you were talking about the community i think there's something in there about people full stop and obviously you're looking at the quality of your players your quality of your um academy like the potential that you've got in there but you've also got the quality of your staff and i think the the leadership there and their relationships and connections and connectivity with the local community for example there might be access to land that you can purchase but only if you've got a good relationship with your local council that's quite a difficult different thing for you to be able to monitor and just check from the outside doing a scan, you don't understand people relationships.

17:12When it comes to people, you can also then include your fan base. Are they active? Are they not active? Do you have a lot of fans? Do you have few fans? And I think that this is a really important thing, again, coming back to the motivations of your investor, because in some cases, an investor might not actually want a big fan base. And I know that sounds counterintuitive, but this comes back to your motivations again. So let me give you a real life example. And this is just me throwing out propositions to the floor. During the negotiations, the conversations we were having with Mercury 13 around potentially investing into Lewis, we have a very active and very vocal and punchy fan base.

17:51They have opinions. They care deeply about the club. The first investment that went into a club in the championship went to London City Lionesses through Michelle Kang. They had a very limited fan base, close to just a few hundred. People didn't care that much. There was no big fan base that was going to push back. But also you didn't have a fan base that was going to buy shirts and was going to show up on day one. So the investor there is taking a completely different track. And they're thinking we don't actually want a major fan base. We'd actually like to go in, clean sheet, build it from scratch.

18:25We're willing. We've got the cash. We want the control. And we know that we can invest into this and the fan base will come. And we're going to focus on the product first and then we'll figure out the rest. So I think that's like a, when you think about community, your fan base, a big fan base, a boisterous fan base can be brilliant, but not necessarily in the eyes of the investor. Okay, and what about, so Lucy, what about this? Let's just pursue this then. You've got fan base, you've got audience. One of the stories of the last few years has been Ryan Reynolds and Wrexham. How does that play in, do you think, into the sort of total value scheme?

19:01Is the Ryan Reynolds experiment, is that useful or do you think it's an outlier? I think he's shown a different way of running a club, of introducing different audiences. I think in Europe, everyone has done football in a particular way for a long time. And he's shown that actually a global audience is very useful. I like the fact that he's taken and championed a team with a really strong community that were definitely not at the top of their game. and then he's created an entirely new revenue stream through streaming, through documentaries and what that's opened up with regards to just even celebrity, just the power of celebrity, that celebrity endorsement and then how he's lent into the community on it.

19:51I don't know, like Maggie, how important or how detrimental would it be then if you're investing in a club where the fans have equity? does that make a big difference like not just a voice but I guess again it can't it kind of comes down to the kind of investor that you are and the kind of level of control you want I mean there's plenty of clubs take Brentford they have a seat at the table which is a fan they protect representation and equity on the board for the fans fans are represented by I think somebody that's on the foundation but is you know coming at it from a fan perspective plus the owner there is a fan so So there are instances where you have highly successful clubs that are highly influenced by a fan base.

20:35For me, it kind of comes a little bit back to the governance and maybe the purpose of what your club is. Do you have a really clear purpose? And maybe that's another, does the club have a clear purpose and does the investor have a clear purpose and do they align? Because they were very clear with the Rexham case, they did want to be part of a regeneration project. And, you know, they wanted it to be part of something bigger than the club and prove that football could have a massive impact. I don't know whether they would have had the story without the community. So I don't think they could have done the...

21:06Well, my question was wrong, wasn't it, in that sense? Because actually I was saying that question was more about fan bases in Western US rather than the local community. Like Birmingham City. Yes, exactly. So they've come in and then all of their bigger partners are now global partners. but then actually their story I imagine will then become a community story because if they're then building out if the ambitions that you're reading are true around them creating a sports quarter in Birmingham I imagine it'll go from a global piece in the first instance to then a community piece in the second how that translates to what we see or consume from a media point of view or how it's streamed or how we find out about it I don't I don't know yet.

21:53Omar there's some very interesting questions coming up that are channeling over what if you could just flick through there was one from Ed R I don't know who Ed R is we might get him to say something he might he might actually exist in real life here we are I've got it are you able to quantify the level of irrationality in a given football club valuation relating to an owner's ego status fandom versus what would typically contribute to a rational valuation which is commercials so can you quantify the irrational the ego yes good question um i think you probably can to a degree i think you can you know you can look at ultimately like football club it's is it i always forget the terminology is it the winner's curse like whoever kind of is bidding for a club you end up ultimately overpaying and i think that's partly a kind of function of i looked at that and thought that it's getting to a sort of product and brand question because i think there might be a correlation between measuring the the intangible of a brand and the ego irrationality of todd boley yeah and whether or not that is a similar process someone will jump in with econometrics or whatever but there might be a way that you could get to a part of the question one of the bits that we're going to get to in the second part of this is actually can ai help with these questions so we're storing up these questions and then actually we're saying right well is it just a fast idiot or are they actually going to help us do the really difficult stuff I think there's also some rational irrationality so for example you might put in all the things that you really want from a club and then it spits out that you should buy Plymouth Argyle and then you're like I don't want to go down to Plymouth yeah Plymouth is lovely but it's you know you might just be like actually I'm flying in flying out I can't fly in fly out to Plymouth so there is like some level of rational irrationality yeah and there's also I mean everyone talks about the losses that that championship clubs make if you kind of do the maths on it all the probability is that someone will win someone will get promoted and they will make money because the valuation of their club goes from you know 60 million to 250 million or 300 million yeah but a small subset is winning the rest are all losing so ultimately most people who get into it probability you is that you'll lose but the expected value is that you'll win and so there's a as you say rational rationality but we just did a thing with Brentford and they are always a case study because it's a you know we're in the English market and there's there's variations of this across Europe I'm sure but are Brighton and Brentford have they sort of sold a false dream because you look at someone like Ipswich and we had the people from Ipswich on you know a couple years ago and their plan was let's get to the Premier League and they did and then they're back down again what does that do is that just okay once you get into that yo-yo category yeah does that mean that you're better off than Ipswich five years ago well firstly I mean there's a lot of luck involved so Brighton Brentford in their first season first couple of seasons in the Premier League they weren't a million miles off being relegated for both those clubs they would have had a 20-30 chance of being relegated and given point in those seasons so there is another narrative where there's another universe an ai universe or whatever sliding doors yeah exactly where those teams go down but they're able to stay up and then reinvest etc so there's a kind of flywheel effect that a lot of clubs you know are seeking to benefit from what those clubs have demonstrated though is that you can massively punch above your weight if you've got some level of operational organizational intelligence that that enables you to do that and again with some of the metrics i'm sure appearing behind me like speak to how do you find those clubs that are doing something doing more with very little i've got a question before and we're going to sort of spin the groups in a minute when buying a women's club do we know what that is now if you're looking at it from a 10-year perspective sevens is that one bet might be that sevens becomes the thing that sets the thing alight in a commercial sense and that goes off and becomes a sort of 2020 of women's football.

25:59Is that a realistic, is that a conversation that you think is being had? Because these sevens competitions are coming up, a lot of money attached, so they're getting the attention of the clubs. I think it's a separate podcast, probably. I think there's still lots of uncertainty, right? And so, again, coming back to that phrase of like dialing up or dialing down, you need a lot of cash. you have to be happy with not very much certainty but with that you do get a lot of control over where you're investing so you've had about 100 million i think dollars um put into the world sevens uh format which has only had one competition just at in the summer just gone it's got its next one in the us in a couple of weeks time i assume there'll be another one in six months time in Europe and I do think that there is a potential that if that builds momentum interest the first one was popular more digitally than in person but there was a level of intrigue and popularity that yeah I do think that there is a potential world where in five years time you go to Arsenal's website and you have men's and then you have women's first team 11s and then women's first Team 7s and it might be that they're completely different types of players are being produced for a competition that is lucrative, that gathers its wheels like the flywheel effect and then takes off but there's so many things that have to come into play for that to even vaguely happen.

27:24And the investment thesis of that is going to be a 20 year bet isn't it? Yeah. Okay right we're going to talk AI. Thank you very much football panel but we're going to yeah round of applause. Right. Okay we're just going to do a quick debrief and how was it for you? What was going on over there? So we had way more interaction than we expected, but very pleased to say nothing fell over. We had over 40 questions from the audience. Excellent, that's good. As well as quite a few input from the panel discussions and I can assure you there are quite a diverse range of questions. Okay, so can we access those post-event?

28:00You can access them now and yes, we're going to let this live on with a quick scan over the content. I'm sure everything will probably survive. Right. Just very finally, just so I know for sure. So people point their phones at that thing. They'll get access to that and they'll be able to interrogate the questions, the conversation we've had this evening. But also it takes them several stages further than that, presumably. Yes. So on another page, which we haven't surfaced here, we've actually got AI to suggest some AI solutions that could be built to address problems. It's a bit like Inception.

28:34I've had to feed in actually a lot of strategy on how we think about AI to try and guide these systems and we've also laid out some TFG strategy how we would approach problems or concerns that you might have we obviously won't leave this interactive system live forever because we still need to do this kind of work and make money but for this evening we'll let it live for a little bit longer before we freeze it for everyone to take home and read okay so it's a sort of reward for turning up for these people in the exactly in the room i came to you and 21st group and blake and i said i've got a sketchy idea the sort of thing i'd like to do i don't want to talk about ai and i want to just you know have something some theater but also just see if we can test something on the night and then that obviously fell to you what was your initial feeling and how did it evolve so initially when i when i heard you ask i wanted to do something different from what people would expect at an ai event and i think i think i'm aware of how everyone's using ai at work and the various frustrations that attach that and um everyone's fully expecting a chatbot to be going nuts and outputting a load of wallpaper responses or maybe on the other end of the spectrum providing useless summaries that skip details and not really adding anything to a live event so this brief being a live event and supporting a panel in a live discussion is quite unique in the sense that the client would never come to us with that kind of problem and if we're answering a complex question we wouldn't be limited by say what we could show on a big screen to be digested in in a few seconds yeah okay so there's a i'm trying to think if you had like a you were narrowed by the bot as an idea because we'll talk about that in a minute in terms of exactly as you say people's sort of usage currently is via chatbots and language models and we're bumping into all of the sort of questions that you know everyone's got about hallucinations and summaries and all of that stuff so it's quite a difficult in some ways a challenge that isn't so much technical as more of a sort of what does it look like performing ai on stage yeah so there's two elements we wanted audience participation so we needed to provide something compelling from a visual sense um obviously we don't want people's eyes glued to a screen but you know why wouldn't we want to create a piece of art that could surround the panel and almost subscribe to the feeling in the room and to not guide the conversation in the sense taking over from your role on the night Richard but to provide a feeling of topics people were interested in and the parts of the discussion that people were engaged with in the room so we wanted to provide something interactive and something useful when also something beautiful you know you attend an event like this in the evening and you expect to both learn something about a by ai but also be inspired so it was a lot of boxes to tick for us yeah yeah yeah and by the way i was sort of bowled over by it so again as we said many times on the you know on the on the way into doing the event you took my sort of expectations and blew them up so it was quite interesting and again you get to the difference between someone like me and my expectations and someone like you and your expectations in terms of what's possible and people who were there saw on the screen it was sort of visualized really nicely and you had these questions so and then you get to okay it's about football and the football acquisition and the value of football so we had a part of the conversation with Omar Lucy and Maggie where we were just talking about football valuations and then trying to sort of categorize the questions in some way to give the model something to play with and so how did that work for you know on the on the night in terms of because we gave the audience the route in and they then what happened so yeah audience were allowed to email in questions and that was relatively easy to handle from an automation perspective but in a dream world and perhaps something we weren't able to accomplish was to have a live transcription of the conversation on the stage being the inputs of the system and using those as stimuli to go off and do some work and come back with something useful now it was a technical challenge that dealt with a large magnitude of data and ultimately it was the uh the technical stuff on the night which we we couldn't stand up so we decided to manually splice out some of the conversations from the stage provide them as inputs label them and you know within i'd say 30 to 60 seconds we had the output of a long automation pipeline that was relevant to the point of the discussion and could be shown on the screen and also could activate part of our brain visualization to indicate that particular spoke of the topic was was now active so how complicated a thing was it because i you showed me the workflow and the the sort of diagram or you know the what was happening diagrammatically how complex was it really because it feels like the idea was really sound the process actually we created something that was really interesting and we'll talk about what actually came in in a minute but how difficult was it for you it was difficult from the magnitude of work that had to be done was difficult I lean heavily on AI tools but even then that kind of increases my productivity to a certain level but can also introduce a lot of problems so I mean if you imagine a big decision tree would take an input from a person on the panel and it might ask questions like is this a request for data for analysis or it might say is this a stimuli which needs some more context do i need to go and look up what this person's referring to should i find a football club that's been in this situation and find out what happened in order to advise the conversation going forward so many parts of the decision tree can be followed in reality this decision tree had many many parallel paths and let's be clear automation and programming nothing new but right now we're seeing automation platforms explode because of the ability to plug in large language models into them and that can help in a couple of big ways the first is helping alignment with the user so that is the marriage of it's the calibration of the system against what the user intends to do and sometimes if the user input is very small, if the prompt is quite lazy, that's a very difficult challenge.

35:41And ultimately how the system behaves without any back and forth for clarification in an event like this can be unexpected. And so large language models also help us decide which tools to call. So if we don't want an LLM to kind of provide us with financial data, we have a source for that in a database which we trust and then lm can read all of the tools we have access to it can read the task at hand and it can decide to execute some code it could even write code and execute it but that's kind of dangerous so automation combined with large language models is simple to an engineer. The architecture composing all of those kind of decision tree nodes and all of those pathways and testing the system with a huge, huge number of potential inputs and validating outputs is the difficult part.

36:42So when I was thinking in the run-up to, and part of it was to just sort of test what's possible, again just purely from someone who does a lot of events and sits on lots of panels and you know does these things on a regular basis and i was wondering how language models are going to then help improve and where they will improve and where they'll help and where they won't and that's part of the fun of you know part of the experiment and was trying to work out well is this what a conference panel is going to be like now and when you said we have got data by by that you mean 21st group have got data and the reason we used football as a sort of route in is that you've got a very rich heritage of collecting data in football so it's not something that someone's going to be able to do from scratch without that well of information is it it's kind of i'm trying to get to the point and say right okay is this going to be something that people will quickly become the normal expected product at a conference and what you're saying is that's going to be quite hard that's absolutely right i mean without proprietary data anyone providing a system that can react to a live event is essentially a glorified microsoft clippy and that's that's what not to aim for okay so providing no value anyone has the same tools at hand what companies are doing right now like like tfg we are sitting on loads of proprietary data which perhaps we didn't even know had any value a few years ago.

38:17A completed project had a process we followed which yielded success. It might have been a very niche project. We obviously have more defined processes for problems we commonly solve. And in the context of advising owners and investors of football clubs on how to make decisions, we have a whole archive of proprietary data and a lot of that can be sanitized. They're processes we codify and follow internally. And so they can be used in the context of problem solving at an event. A lot of questions that come up are not novel problems and they have solutions. I'm a romantic and I love the organic discussion and the chaos in the live event.

39:00And although I'm a technologist, I'm quite skeptical of how much of these things need to be driven by ai and so i want ai to be piloted by humans and i want it to enhance the value of attending live events yeah yeah yeah yeah no i agree and i think i probably you're trying to fix bad events in some way and actually a good panel or a good conference or whatever it is doesn't need this addition so actually you'll probably you know and this is me i'm starting in a different place i'm thinking okay well how would you solve a problem like conferences but be better at conferences would be one starting point and not throw you know ai at it and so i was trying to be very clever about you know wanting a conversation about ai itself and then having this other bit of on on football and i my my reflection is that the bit about ai was was interesting the bit about football was interesting but these these two things weren't connected particularly i didn't make that link between them so there was a sort of the bridge between those two two panels didn't work and that was a human failing i.e mine rather than uh you know a sort of technical one but i think that's a really quite an interesting learning and we came in here with everything was about test and learn and let's see what happens and learn in public so i'm i'm pleased with that and i think i've learned quite a lot in terms of just managing i think it's about managing my own expectations and i think that's probably true of my relationship with large language models more generally actually well i think you're you're quite hard on yourself there richard i mean it was one big experiment and we did try and cram a lot of ambition into the event and and when you say you know we're aiming to provide a conference you know an event people will talk about and remember when you compose panels of guests they each come with different you know very diverse set of abilities some are there to educate some are there to be the authority some are some are perhaps there to entertain and provide a bit of chaos and the ideal composition is probably a bit of a mix and I think when you have a system like you know AI is responding to one of the inputs from the various people those nuances those differences and personalities and the people's intentions aren't picked up upon and unless it is calibrated that way it's very difficult to tune into somebody and provide the optimal kind of piece of material on the screen to support them because some people don't need any whereas the educational perspective perhaps would warrant flashing up some supporting data which they could use to flow into the next conversation yeah it's something i think about quite a lot at the moment which is it's not the use of it as a live tool which is what we trialed it's actually afterwards and i think about this about podcasts that as everything becomes sort of input for someone else's language model, then there's some questions for us as an organization, you know, in terms of unofficial partner to say, right, okay, but what are we doing?

42:25Where is the value? And actually, we've tried a few times to say, right, okay, what about taking a podcast or a series that we do and running it into and just collecting the transcripts on a very basic level and making that available to listeners and subscribers for them to just go and interrogate afterwards using, you know, notebook LM or some other or Claude or whatever their chosen sort of foundation model is. So I think that at the moment, I think that's where we are. And sometimes the conference is probably the same in that you sit there, it's interesting or it isn't. Okay. But it's a different medium, but actually afterwards I think, okay, well that was quite interesting, but I've remembered about 5 % of what's gone on.

43:14it would be quite useful to have the transcript and then to interrogate afterwards and then extrapolate using the model's sort of capabilities. So I think that's probably where I think the initial value is going to be and where we will sort of try and push. Do you think, do you see what I'm getting at? I do. It's a really interesting point to discuss. And I don't necessarily agree with you that you should open your transcripts and allow people to interrogate it. It kind of reminds me of the difference between reading a book and being able to reread passages and providing those, you know, you can pause on demand and think about something.