How The Telegraph is Redefining the News with AI with Dylan Jacques, CPTO

(Intro Narrator at 00:00:00) Today, we're talking to Dylan Jacques, CPTO at The Telegraph, about how newsrooms are harnessing and navigating the power of AI. Thank you to DigitalOcean for sponsoring this episode. For simple cloud and powerful AI that's built to scale, visit digitalocean.com or just click the link in the show notes. You're listening to Joel Beasley, Modern CTO.

(Joel Beasley at 00:00:26) Yeah, I want to jump right into it, right? But let's start at the beginning. Like, what is it that you actually do at The Telegraph?

(Dylan Jacques at 00:00:33) Yeah, so I'm Chief Product and Technology Officer. So I look after our product development, roadmaps and our plans, and how we invest in our digital products to deliver value to our customers. I look after our plans as to how we utilize AI, both looking at our internal workflows—how we can work better, more effectively—but also as we, like all companies, move into a different era, how we can use AI as a tool for growth. And I've championed a lot of our approach building AI into how we operate and really building the organization of the future.

(Dylan Jacques at 00:01:22) So that's a big part of my role. But also the technology part is we have all the usual systems and engineering and products to support and make sure that they run effectively and that we build them effectively. So really our whole digital ecosystem. We're a subscriptions-led organization, but we also have a lot of things like advertising, lots of complexity to look after. That all comes together that really creates revenue for the organization. So it's quite an interesting, dynamic, fast-paced, and complicated but interesting part of how to run a modern news publisher.

(Joel Beasley at 00:02:12) So how do you think news is—the newspaper model—how is that changing with AI?

(Dylan Jacques at 00:02:17) Well, I think that almost every business, really, is having to reinvent themselves to a certain extent. And the news publishing industry is no stranger to general disruption, and AI isn't a huge amount different. So we've been through print to digital, to then our audience is moving online, to lots of different revenue streams, the online ad model changing, how products are developed, all the way through to that moving to a subscriptions model. And these are all changes that have caused us to really reinvent how we get revenue and how we sustain the journalism itself. And AI is the latest megatrend, and it really is impacting all organizations, as I say. But it's impacting publishing in all sorts of different ways, from the fact that the market has fundamentally shifted. People are used to using tools like ChatGPT now, and I've never seen a technology that has been adopted as quickly with such an exponential growth curve as AI in my whole career.

(Dylan Jacques at 00:03:46) We've seen lots of things. I mean, smartphones is the other disruption, and that change of consumption patterns. But this really—and I was talking to another publisher recently, and they were saying that their impressions of their content on these platforms has been doubling literally every month. And the amount of growth—I was on the train this morning and walked down the carriage, and every other person was on ChatGPT asking it questions, how was my day? Like, that was just a year ago, that was just totally not the case.

(Dylan Jacques at 00:04:19) And that's going to have an enormous impact on the amount of eyeballs that end up on news, how people go after information, and the implications for the sector are huge. But I'm an optimist, I suppose, generally about it. And I think that the news industry is very adept at adapting to these changes, and there is an enormous amount of growth and productivity potential in using AI tools. I think it's going to enable us to produce far more at a far higher quality, ultimately. We need to think thoughtfully about how we use AI, and we need to protect the fact that, really, journalism at its core is a high-quality, human-produced premium product. But I think the tools that are going to be at journalists' disposal to do things like deep research, to interrogate massive databases of information and ask the right questions to get to the heart of the story—that means to do that is going to be massively improved. So I'm quite excited that we're going to hopefully get to a new era of being able to have human-led, but with AI tools, much more breadth and much higher quality and maybe more creativity too. But that has to obviously be done in the right way that's not using AI for AI's sake and diluting the journalism, which is ultimately the core. I think AI tools are very good at making sense of things, but they can never—AI tools are terrible at asking the right questions, and I think journalists are brilliant at asking the right, often challenging questions.

(Joel Beasley at 00:06:26) Until we build a model that's specifically designed to ask the right questions. No. So all right, I want to talk about this, but you brought up a bunch of interesting points. The one that I want to zero in on is the adaptation, right? So you've—the industries have adapted, right, when you went from print to digital and all of that. And it seems as if you could think about it like the newspaper was the original UI, right? And so you controlled the UI, you controlled the content in the UI, people went and paid for it and picked it up. Then the UI kind of went digital. Like, instead of the paper, now I have a screen in front of me. And that's a pretty easy transition. Take all the content that was on the paper and put it on the screen. You can have people come to your screen. But now what's happening for me personally—and like you said, how you watch people on the train or how I watch myself use technology—I noticed that I stopped even going to X. I just go to Grok now and say, hey, because there's too much junk on X, too much negative stuff on X. But with Grok, I can say, hey, give me the top five stories. Do a slightly positive tone to it. Like, I don't want horrible things in my feed. And it'll just bump out exactly what I want for the news. And I'm like, oh, okay, cool. And so now that sort of human part of me that's curious about what's going on outside of my little bubble of my family is now satisfied. And so the reason why I'm bringing this up is because the newspaper-to-screen change, that seems pretty simple. The purchase of paper to subscription change, that seems pretty simple. But the fact that they're not even—like, I'm not even going to this. I'm going to this other entirely new thing. Are you guys working licensing deals with the Groks and the ChatGPTs to include your content? Is that a revenue stream for it to stay relevant? How does that all work?

(Dylan Jacques at 00:08:18) Some publishers are. And you are right. I mean, everything's changing at once in the example that you've given—both how things are consuming, both how we monetize news is changing. Personally, I see a world where news—if it—I think for a news provider such as ourselves or anybody is a sort of content stream into an LLM, be that Grok, OpenAI, or whatever it is, I think for a customer, that's not a great result. I think that you don't want a world where you're consuming information, but it is quite generalized by the platforms, and you're consuming it in a similar way. And I think it's quite important for news publishers to have a direct relationship with our customers for the simple reason of trust. So the minute that everybody starts reading our content through a lens of a third party, there's another company in the mix that is presenting the content on our behalf. And for me, that is—there are trust issues to that. And even, that said, you have to recognize that what you've described is how you consume and your expectations as to how you consume information has changed, and you're not necessarily interested in the same thing that all of our other readers are, and you're expecting a personalized feed, not necessarily even a feed, but you're expecting a personalized experience that is either answering the specific need that you've got, that particular question, that particular time, or it's at least dialed into the types of things that you like. It's anticipating what you would like, and it's bringing you quite relevant things.

(Dylan Jacques at 00:10:23) I think if we're solving to that problem, I think publishers can still solve to that using tools. And whether that is the Telegraph app has an agent which knows you extremely well and gets you direct access to our news—I think publishers can absolutely still deliver that in the same way that we've adapted all of our products to shifting consumer trends for some time. You know, this is a major one, but we can still adapt to that. I think a world where absolutely everybody gets their news from one place, one interface, one app, and there are just a number of sources to feed into that—yeah, I don't know about that.

(Joel Beasley at 00:11:11) No. I don't think that'll happen. No. I am optimistic too, and I did like the point you made earlier about premium content. So the fact that you can know, like, and trust some very specific source and you're willing to pay for it, that model has been proven. It's been proven through the citizen journalism. It's been proven through your own subscription model. People build this trust with this organization or person, and then they're willing to pay for it. I think as the AI rises and—I believe that there will—you could tell me what your thoughts are on this, if you think I'm crazy or not. But I think we're going to have another major shift when the deepfakes become real-time good. So, for example, when it's so good that you can't tell if it's me right now or if it's my AI twin that I sent. Like, I think that'll happen in the next three years. And I think when that does happen, it's going to create a massive cultural shift where we'll start doing business more in person, like at conferences and things like that. Because once we reach that threshold of not being able to tell, then it's going to create an enormous amount of trust issues with everything.

(Dylan Jacques at 00:12:18) I totally agree. And you're right. I think that we are not far away from a world where it's quite hard to see the wood through the trees, and it's quite hard to discern between what is a trusted fact and what is just—an article—I asked ChatGPT something. They get better at sources, but it is unclear where it has learned that information from that it's providing. And I think that there will be a counterculture to it, and I think that there will be, ultimately, a premium on well-researched, human-collated—that's, you know, we—myself and other publishers work to high journalistic standards that are just not going to exist on these platforms.

(Joel Beasley at 00:13:10) Some organization in a tweet, right? Yeah, yeah.

(Dylan Jacques at 00:13:11) Some organization needs to ensure the enforcement of journalistic standards, and that's what a news brand is. Like, we're talking a lot about how things are consumed and how information is consumed. But if you're getting your information from just one bucket that is pulling that from scraping it from a news source, pulling it from Reddit, combining it all together and predicting it, like, then you're a million miles away from what we think of as a reliable news source, a million miles away. That's not news. That's information that's been smashed together from a thousand different places. And whilst that might be a great solution for, right, I'm having a conversation with my boss this afternoon, how should I phrase something? You know, that can be handled. That's just language in the right way. But when it is, you know—when we're talking about—we did a story recently on our deputy prime minister who then resigned. Like, when we talk about holding governments to account, like, there has to be a much higher standard to that, and the LLMs are not going to do that.

(Joel Beasley at 00:14:30) How far—so I think most people would say that having a conversation with a nuclear scientist, like a really well-trained LLM on that data, you can have a pretty high—you can have a pretty detailed conversation and get some value. You can bounce ideas if you're a nuclear scientist and you're trying to bounce some ideas. And so that works. How far do you think we are away from the LLM being as good as a human reporter?

(Dylan Jacques at 00:14:54) Well, I think that from the sort of standards of information gathering, the way that people talk about this is, you know, we're at postgrad level or undergraduate level right now, and we're getting to a point in 2030 where they're going to be professor level. And the line that you cross is that they are going to be more capable than the most intelligent human at that particular thing because it's been trained on a whole corpus of research. And if you've got a professor of, you know, whatever and there's a whole detailed syllabus and there is a whole background reading, you know, I can see where you get to the point where you can ask questions in an effective way and somebody's going to be, in inverted commas, expert in that thing. But you've got to think of journalism as a slightly different concept to that. We have reporters that stood on the ground on the front line of the Ukraine conflict and are interviewing people and are getting a sense of what's the impact of conflict in these countries. And it goes back to my point as well on how to break new ground, not just be an expert in previous events. How do you ask the right questions? How do you—and there is a lot—there are a lot of personal relationships to journalism, being plugged into the government. And I would view having an academic understanding of a subject is slightly different from the act of going and creating journalism.

(Joel Beasley at 00:16:43) Yeah. And you will definitely have that natural, corporal delay because, you know, these AI LLMs can't go walk physically to, you know—at least without a human counterpart recording or something of that nature. But there will be some delay. That's why I'm generally optimistic, and nothing really is happening overnight. It usually takes a little bit because human behavior takes some time to change. I do want to talk with you. I saw this interview you did last year about twelve AI products in twelve months. So I'm going to ask you for an update on that. But first, I wanted to chat with you about QuantumMetric because that's how we got introduced. So how do you know QuantumMetric?

(Dylan Jacques at 00:17:24) Yeah. So we have been working with QuantumMetric for a few years, and I can't remember how we were introduced precisely. It wasn't through myself, in fact. But I was interested in the technology and being interested in how we evolve our products. It's the first time I've seen as powerful tools—sort of, really, typically, what we have and a lot of organizations have are a lot of after-the-fact reporting. And the problem at the time was, you know, often what's happening on our products right now and trying to get an understanding of, okay, if we change this, what would the impact to user behavior be straight away?

(Dylan Jacques at 00:18:16) Publishers typically did that by looking at a load of reports, changing something, waiting a few months, looking at a load of reports again. And I think it was the first time I'd seen genuinely real-time being able to actually see and analyze our customers' experience within our products. And that was a huge leap forward at the time. And we were also able to trace through where we've got points of friction in our products, and if we were able to change that, fix that, what would that mean in terms of knock-on engagement, and then what would that mean in terms of revenue? So it was the first time we were able to actually use tools to think about that rather than many months of after-the-fact analysis.

(Joel Beasley at 00:19:03) Oh, nice. So from a business perspective, other CTOs and VPs of engineering, those types of people are listening. What is the one reason why you would recommend they even take a look at QuantumMetric? Like, what's the problem they'd be experiencing, or what's the business value they would communicate to their C-level peers?

(Dylan Jacques at 00:19:21) I think that understanding the business value of a change to your UX—and that sounds quite a fuzzy thing to say, but, you know, if you're an airline and you've got thousands and thousands of e-commerce journeys and the fact that you have the facility to be able to see at volume, like, this delay or this UX not being quite right—we can see a pattern in real time that it is causing a load of customers to falter. And some of them eventually work it out and move on, some of them give up. If you were to fix that, what would that mean for your business?

(Joel Beasley at 00:20:12) That's a pretty good value prop.

(Dylan Jacques at 00:20:14) Right. That being able to do that gets you out of—I think a lot of organizations get into the trap of you try and manage a product development plan, and there's a lot of opinion-based stuff. Like, I think it's really important to do this. I think it's really important that we do this. I think, you know, in this world, there is a lot of data, a lot of customers.

(Dylan Jacques at 00:20:40) You really need a data-driven, objective view for—you know, if I was to change these five things, you can go into tools now and say, right, put these errors that are on our site, put them in order of business value. These friction points, put them in order of business value, and your teams can start working through those in the right order that hits the biggest returns first.

(Dylan Jacques at 00:21:09) I think that's a problem that actually everybody has. And you need the right tools to be able to deal with that.

(Joel Beasley at 00:21:18) That's true. Because everybody—there's no shortage of logs and things that are wrong. Being able to organize them from business value sounds like the thing you would want to do.

(Dylan Jacques at 00:21:28) Indeed.

(Joel Beasley at 00:21:29) Real quick, before we continue, I wanted to give a special shout-out to DigitalOcean for sponsoring this episode. DigitalOcean is cloud infrastructure that's simple to spin up, complete with integrated AI dev tools, plus 99.99% uptime SLAs and industry-leading pricing on bandwidth. We all know DigitalOcean. It's super reliable. It's been around forever.

(Joel Beasley at 00:21:51) And now with DigitalOcean's Gradient platform, developers can train, fine-tune, deploy, and scale AI workloads all in one place. DigitalOcean is reliable and affordable at any budget. Companies can save up to 30% on their cloud bill when they come to DigitalOcean. Use code DO25 and get $200 in free credits to get started. Simple cloud, powerful AI, built to scale.

(Joel Beasley at 00:22:16) That's DigitalOcean.

(Intro Narrator at 00:22:18) Now back to the episode.

(Joel Beasley at 00:22:20) All right. So 12 AI products in 12 months. I feel like I might be holding your feet to the fire. Is that something that you said in the interview last year?

(Dylan Jacques at 00:22:28) Yeah. No, it is. It is. Yes.

(Dylan Jacques at 00:22:30) Yeah. So it sounds like one of those promises that you make in an interview, and then you keep holding you to account for it for the next five years. A lot of organizations, we spent a bit of time trying to make sense of what—when GenAI hit three years ago, everyone sort of took a bit of a beat to say, what does this mean for us? How should we handle it? What should our policy be? And no one really knew how fast it was going to accelerate. And at the time, there was quite a lot of talk about policy, quite a lot of talk about, you know, how we should separate what we do from AI and how we should think about it. And so really, what I wanted to do is—the concept of 12 in 12 months was it's really about how you deliver at speed. So one of the consequences of AI, again, to everybody, is that the current or old way of developing software and building new experiences now has to change because we are not going to be able to do it fast enough to operate competitively in a world with organizations like OpenAI and others. People can spin up an app and website without any knowledge of coding very easily now. And so how we build products had to change. 12 in 12 months was an effort to say, right, okay. Whatever we do, we don't know all the answers to how AI needs to transform our company. We don't know exactly that if we say, right, we're going to build 12 things over the next year that 10 of them will still be live in two years' time.

(Dylan Jacques at 00:24:35) But it is not just about what you develop. It's about building a muscle. It's about building capability, and it's just about getting going and how to do things fast. So I think we're in a world where regardless of how you get it done, I think you're going to need to be shipping a major feature every month at least and doing that on a disciplined regular basis. And, you know, hopefully, they're the right ideas. Some of them might be the wrong ideas, but there is really no—oh, that didn't work. It is all muscle. It is all capability. And even if you launch something and you think, well, customers didn't engage with that, you know, there might be a slightly different use case using similar capability that you launch in two months' time, and at least you've moved the ball forward, and you've got teams used to doing this. And, really, it was a bit of an effort to just get going as fast as really you can. And at the end of 12 months, you'll have delivered 12 things. You know what has worked, what hadn't worked, and you're better placed to tackle the following year.

(Joel Beasley at 00:25:45) How did the team respond to that? Were they pretty pumped up for it?

(Dylan Jacques at 00:25:48) Yeah. So all the managers thought you're massively overcommitting. So all my colleagues were like, you're setting yourself up for a massive fall here.

(Joel Beasley at 00:25:55) They started calling you Elon Musk. Like, okay, Elon.

(Dylan Jacques at 00:25:57) Yeah. Take me to Mars. And the team thought it was a little bit ambitious, but, actually, it was okay. And the key thing that it really does is right-sizes what you are biting—you're biting things off in the right chunks. And these are all AI tools, features, experiences, some of them sort of products.

(Dylan Jacques at 00:26:30) But, really, it was we have to get used to the fact that we're going to have to deliver things on a cycle like that. And, you know, first couple challenging. But actually, after a while, people got used to it. It was very clear that we were expecting a launch at the end of the month. There was no drift. There was no, oh, we hadn't thought about the launch plan. You know, we have to think about a launch plan every month at the same time. So you end up with a sort of formula that, oh, it's week three of my month. We haven't actually agreed the launch plan for this thing. We need to get on with that.

(Dylan Jacques at 00:27:11) And it effectively establishes a cadence that then everything has to work to. And, you know, you start something you think, actually, this is really, really complicated, and this is going to take a while. You quite quickly have to work out, we have a broad backlog of problems that we're all trying to solve. And if we think this one is really too big, you need to have something else to swap it with quite quickly. And getting the team to just, at all costs, preserve that 12-month cycle with whatever it is, and get going was the key, really.

(Joel Beasley at 00:27:53) Do you think it was a net positive?

(Dylan Jacques at 00:27:55) Yeah. Absolutely. I think it was a net positive. I think that with AI, you want to hedge your bets in as many directions as possible. And, you know, you could very easily fall into the trap of delivering one or two major releases over a year. And the issue there is that if that one thing doesn't work, then that's all you got.

(Dylan Jacques at 00:28:31) And, you know, if you're delivering 12 things, you know, some of those are going to work. And if they don't, you're figuring out where was the value, why was that—you're learning 12 lessons. You're learning you've got 12 different parts of your technology stack that can now work in different ways. You know, you've got a lot to build on. The downside to it all was we did end up with 12 things that we then had to manage. So you also have to be quite disciplined about what you turn off at the end of that as well.

(Joel Beasley at 00:29:10) So I've got some questions about generative AI and how that's impacting publishers, specifically for you and me as CTOs. Right? So in both of our countries—you're in the UK. Correct?

(Dylan Jacques at 00:29:23) I am.

(Joel Beasley at 00:29:24) So both of our countries, we have a concept of free speech, but there is limitations in both the countries. One of the largest ones is incitement of violence. You know, that's a common one. And my question here is with these AI models, if one of them goes off the rails, are they jailing the CTOs? Do they have laws? The UK tends to be pretty good about making laws ahead of time, at least moving faster than a lot of other countries. Are they legislating anything around this? Like, what happens if you're a publisher, you have an AI, and it starts hallucinating things that are jailable speech? Like, who do they come after for that, or who is responsible? How does that work? Have they outlined it yet?

(Dylan Jacques at 00:30:16) I mean, look. I don't think it is totally clear. However, I'm operating on the firm assumption that if we publish something under the banner of our business, you know, there's a level of responsibility around that. And, you know, we try and have human in the loop. And more than anything, we should not be—in our output, have an automated unchecked feed from an LLM. That's not what we do. If you want that, just go to the LLM. There is no point in us having a redirected feed from an LLM through us back to you. We're not adding value really to that. But the way that I look at it is, you know, it's a tool. Whether you use Google to research something. It's not Google's fault if you look something up on Google and quote it to somebody and, oh, don't sue me. Sue Google because that's where I read it. You know, I don't know whether it has to be a lot more complicated than that, really. And I think that we can use AI-grade tooling and applications to be better at what we do, but I'm really not sure we should be in the business of mass-produced AI output as something our customers consume.

(Joel Beasley at 00:31:59) Oh, yeah. I don't think so either. I was just wondering, like, mistakes. Right? Like, if you have an AI agent and it kind of goes—you've seen the agents go off the rails just like talking to them normally. Right? Or if you haven't, let me know.

(Dylan Jacques at 00:32:12) No, no, no. Totally.

(Dylan Jacques at 00:32:14) And actually so we did—we experimented a little bit with summaries. And so one of the problems, particularly for our younger audience, is, you know, our content could often be quite long-form, quite dense, complicated geopolitical issues. And if you're on the tube or you're, you know, you're on your commute and you want a bit of context, a bit of lead-in, quite a high-quality rich summary can be quite good. And then you can dip into the bits within that summary that you're interested in. And I think you're going to see this more and more, you know, whether it's audio summaries, video summaries, just making the news more digestible, I think, is the sort of problem. And we can use an AI to create a summary, but it's quite important that you ground it in the content that we create. And the challenge becomes when it's having to fill in a gap, what it will do is it will ultimately—it needs to produce the output, so it will pencil in, pencil into the gaps. And that's where you get that off-the-rails risk. So the way that we've looked at that is really a technical challenge. And if we're producing a summary of something, you know, or any of the use case examples that I've given and it needs to fill in the gaps, we have a grounding process that goes away, and it takes the summary, and it then redigests that to say, is there anything said in here that doesn't exist in all of these articles that's genuinely, we think, additive? And if the answer is yes, then it goes up a risk level, and it's perhaps not included.

(Dylan Jacques at 00:33:57) So there are ways to try and engineer out the problem, but also test to use LLMs effectively.

(Joel Beasley at 00:34:06) I fully agree. It's just an interesting thought experiment problem. So, like, without AIs or LLMs, you can always point back to the person. Right? It was either a rogue engineer that did this or someone had credentials and they made this change and said this thing that's a jailable speech thing. So there would always be a trace back to a person who could be held responsible. But the moment you have an AI, who becomes—who's responsible then? Like, who is it? Is it the CTO? Is it the CEO? Is it the engineer that made the mistake? Like, who—where do you put responsibility? And I think that's something that will just, you know, have to be played out in time for when it does happen. I think that's just a court thing that will go on.

(Dylan Jacques at 00:34:55) I think that there is, at the moment, a lack of accountability to those big technology companies in two different ways. I think that, ordinarily, if a company produced a technology and that technology, you know, led to harm in some way that led you to trace that back, I mean, ultimately, it's that company that is then responsible. And we should hold those companies to account if harm is caused. The other thing is that these models have been built on information that has been scraped from people's sites without their permission. And I think that's—we're still trying to get our heads around where the accountability for that lays.

(Dylan Jacks at 00:35:47) I think it comes down to we're moving really, really fast, and we don't really have all the answers yet on what is appropriate, what is a good model to do this well. The technology is moving faster than our ability to have good constraints around it. So I think that is just generally interesting.

(Joel Beasley at 00:36:13) I'm just trying to keep CTOs out of jail. That's, okay, yeah, we're in a new world. This is going to be really fascinating to see how all of this stuff plays out.

(Joel Beasley at 00:36:25) I do want to touch on a couple leadership questions as we start to wrap up. Are you okay with that?

(Dylan Jacks at 00:36:29) Yeah, yeah.

(Joel Beasley at 00:36:30) Yeah. So one of the things I always like to ask guests is for a piece of leadership advice, but I'm going to give you a couple constraints for it. Okay? It has to be something that you were told that you then implemented and it stayed with you for a long time. You found it very truthy. You're like, "This is great," and it's become a core part of who you are as a leader. What advice was that?

(Dylan Jacks at 00:36:54) I think that you've got to lead from the front, and that you've got to show that you are setting the example and you're not afraid to roll up your sleeves and lead from the front. I think that is, ultimately—I've had the privilege to work with some really, really great leaders who have just totally led by example. And rather than being briefed on what to do and what not to do, I can just see what good looks like. I can see how to operate. I can see that there is credibility there that you have to build, and you can't ask for that. You just have to—that has to be how you operate. People have to just look at you, your day-to-day, how you conduct yourself, you know, asking the right questions, setting the right example, and the people that work within your organization, that's how they will start to operate. They'll start to operate looking at you, seeing how you are and how you handle things. So I think if you want your team to be the best, you have to understand what being the best is to you, and you have to do that all the time and really lead from the front.

(Joel Beasley at 00:38:09) Nice. And then how did you go about, in your career, how did you become the CPTO for Telegraph? What is it about you that allowed you to climb and grow in your career?

(Dylan Jacks at 00:38:22) So, "always put your hand up" was actually another good thing, and I think that was something that has helped me. Just, there are always challenges. There's always new things that might be in your role. There might be outside of your role. But being good to work with and recognizing there are all sorts of dynamic challenges—and yeah, we've all got busy day jobs, but as challenges crop up, just putting your hand up for things. When AI first started—I mean, AI broadly has been around for ages, but the sort of inflection change of a few years ago with generative AI—I saw this as an interesting move, an interesting shift, and I just wanted to get all over it. And I just sort of set my stall out to become, as far as I could be, as deep an expert and get on with stuff there. So within my little area, I formed a little mini lab. We were debating, "Should we assign development resources to AI? Should it be a big part of our investment program?" And it wasn't at the time, but I just carved out part of my team and said, "Right, we're going to just become the AI lab through doing stuff and starting to deliver things," and really just getting on with it.

(Joel Beasley at 00:40:01) And did the CPTO role exist before you, or did that become a creation out of the necessity of exploring these new technologies?

(Dylan Jacks at 00:40:09) It existed in different respects. I mean, we sort of ended up joining two things together, but I think we sort of recognized that the two now are going hand in hand. And we now have to be engaged in—you know, it's not about iterating our existing way of working and our existing products. We have to think about, what is the new way of work? You talked very, very well about how you now consume news in a completely different way. And will an article format even exist in three years? I'm not sure. So I think we have to start thinking about how you bring everything you need together to make journalism super relevant and our business super relevant in ten years' time. We've got to start thinking about what those problems of the future are and just start solving for those problems, not just thinking that we're still in the old era and we just need to gradually improve.

(Joel Beasley at 00:41:18) Yeah. Instead of trying to solve six months down the road, solve three years down the road.

(Dylan Jacks at 00:41:22) Exactly.

(Joel Beasley at 00:41:23) Yeah. I love this. Well, thank you so much, Dylan, for doing this. I mean, we made a podcast. How do you feel?

(Dylan Jacks at 00:41:29) Very good. Very good. Reinvigorated. Ready to go and solve the big challenges.

(Joel Beasley at 00:41:34) Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn or send me an email, [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.