Ep 28: Revolutionising Advertising Briefs: AI's Impact on Advertising

Our overall product, omg, has an AI layer that we're building around it, which we're calling slipstream. One of the biggest problems that you get in the advertising industry is the quality of brief that comes in from a client to their agency. So we used to run briefing workshops with clients we still do from time to time to get their people better at writing a brief. But when a lot of the AI technologies came around, it's like, well, we can do something here.

So you've augmented the brief based on the information that exists within the knowledge base already. So you've got a rag database sitting behind the scenes that's absorbing all of the information that's been shared in. They're then putting in their brief. It's being analyzed, it's being checked against specific things. You're reframing it back into a consistent brief format, showing that back with the customer and getting their feedback on it. You're encouraging them to produce a higher quality brief by actually doing some of that work for them.

Hi, I'm Freya. Welcome to the Skynet Chronicles Business in the AI Age, the podcast that dives deep into the world where artificial intelligence meets business. Join us as we explore the latest trends, challenges and opportunities in this ever evolving landscape. From innovative startups to established enterprises, we'll uncover the secrets behind the successful integration of AI technologies. So grab a cup of coffee, sit back, and get ready to unlock the mysteries of business in the AI age.

Oh, good morning, Rob.

Good morning, Tim.

Rob Goddard, Director of Technology Solutions at Oliver Group and car enthusiast. As we've just spent quite a bit of time talking in between, hopefully you've.

Got all of that out of our system. But I will apologise for the number of car analogies that I'll probably drop in today.

We do car analogies and not Terminator analogies. We'll probably be fine.

That will keep me more honest. Yeah, yeah.

Okay. So tell me about your role at Oliver.

So I am the Director of Technology Solutions. I joined Oliver 10 years ago this December, maybe 10 years ago today, actually thinking about it. That's crazy. And when I joined Oliver, build in house solutions for our clients. So going back 10 years, we were about 350 people. Not really anything central. We were all split up across client offices. So it was very difficult for people to piggyback off of other people, people's ideas and technology. So one client would come and say, we need an asset management system. Our team would go away, they'd employ a freelancer and they would build that asset management system, but nobody else in the group would know about it. So another client comes along and says, can I have an asset management system? Someone else goes away and builds something completely different. So when I joined 10 years ago with a couple of people that I'd worked with for a number of years previously, we started to productize everything that we saw across the group. So it was things like asset management systems, automated artworking solutions, and things like that. And then we looked at the. Because it wasn't our first rodeo. We've done this kind of thing before. We knew what we needed in order to run the company as well. So we looked at the workflow solutions, the project management, the production management, you know, automated workflow solutions and things like that. And that's where the Oliver Marketing Gateway, or as we call it, img, came from.

It's a much better name.

It's not a bad name. We're a creative agency. I'd like to think we came up with something better, but, you know, we had to. It literally got to the point where it's like we've got to register a name somewhere just so that we can actually start building something. So that's where the name came from. And it stuck. It's gained a bit of traction. It gets people's attention in a room, whether good or bad. So, yeah, so it came from that. And yeah, 10 years later, here we are. So we've had quite a lot of very aggressive growth within Oliver. So when I joined, we were about 330 people. We're now five and a half thousand. So that growth rate has been very, very aggressive and very hard to keep up with in a small business or what was a small business. So we were also responsible for all of the IT within the group as well. So we couldn't dedicate all of our time to pure software development. We were looking after all the IT solutions and that involved as well. We've just won a new account with 50 people on it. It starts in two months. Get all the kit set up, get the servers set up. And because we're inside other clients environments, it's not always, here's a server, go away and use it. It's where do we put that server? Does that go within our network? Does it go within the client's network? If it's within the client's network, how do we access it? Are we using our laptops? Client laptops? There's a whole load of things to consider in all of that as well. But we got very, very good at the implementation side of it. So we had dedicated teams of people that work on implementation and you kind of just follow that blueprint over and over and over again. So, yeah, 10 years later, we've got a very well established product. It runs our business day to day and I have a very good team of people that I work with very closely on the product and development side. We've got an outsourced development partner. We use that. We've got this 10 year relation at Oliver with. We've had a relationship with them before as well and we're very, very productive. If we look at the output that we get compared to where we've been previously, I think our ratio of productivity is probably higher. Overall delivery is less because we haven't got as many people. But I think we're in that nice little niche where we've got a small enough team that we all know what's going on and you don't necessarily need to have too many checks and balances because we've all got an awareness as to what's happening. I'll caveat that we do have all of the rights, infrastructure and processes around it for ISO 27001 and everything else that we need to make sure that we're doing everything properly, following the right procedures. We do pen tests and all stuff like that as well. So, yeah, a very. Probably quite a standard software development life cycle, but a quite aggressive rate to keep up with where the business is.

And that business model is interesting, isn't it, because you're inside the companies that you're working for and because they're on a mix of paying for people and paying for assets, depending on the particular relationship you're in. Yeah, we started talking some five, six years ago about the project management stack of how you could move to a point where it's task and output driven and managing that potential transition has then been overtaken by the capabilities of AI and how that gets baked into the process.

Yeah, yeah, yeah. And I think I had a session yesterday actually sat with our group legal team, so walked in to do a demo and it was sort of 15 lawyers in a room. So I was sort of immediately nervous and I was just sort of taking them through what we're doing and a lot of the questions that came up were obviously legally related around, you know, what's the implication of doing this and that and, you know, let's just SAP all the fun out of all the stuff that you're doing here.

Bun Sponge.

Um, yeah, and one of the points I was trying to make to Them was there's the ability to build things now is so much easier and quicker than it used to be. And I think there's a tolerance to, I don't want to say higher levels of failure, but because you can build things that bit quicker, you can get a POC out really, really quickly. Didn't work. Throw it away. Yeah, you know, that's always kind of been the root in agile development, you know, fail fast and all that kind of stuff.

But the rate actually followed that well and now it's on steroids. It's so much quicker.

Yeah, because I think people still, you know, you wanted to fail fast, but you still had that view of what it could be in six months. And people latched onto that because you've got that mental ownership of that idea that you had. Whereas now you can spin something up so quick, run it through a poc. Is it working? Yes or no. And then, okay, yeah, POC signed off. Let's move that to an MVP. MVP's worked. Let's move that into an enterprise. So we've got quite a detailed, what we call our MPI process, new Product introduction process that takes us all the way through from ideation of something all the way through to final delivery. But we've got yardsticks in it that kind of say, right, what do you need to do in order to reach these different parts?

What's the risk with that sort of process now? Because it's very well structured, it has the right gateways and measures in place, but is it fast enough to be competitive?

I think if you've got the right balance at the lower end of that funnel and you are aware of what the potential pitfalls further up are going to be, it'd be very easy to just come up with an idea, build it and get a POC out to one account that go, this is the best thing we've ever seen. It's fantastic. Let's roll out everywhere. If you've got a legal check up here that you know is going to kibosh that whole thing, get that check in earlier because it's those kind of things. A lot of these things aren't restricted by the technology anymore, it's restricted by the impact it's going to have on your business, any sort of legal guidelines or laws that are in place that might scare certain clients in terms of data protection and stuff like that. So I think it's about adjusting what would have been a typical process five years ago and saying, right, those final checks that we used to do, we actually need to get them further up the line because we need to be at a knock these things out quickly if it's not got any longevity in it.

So improving the level of understanding and what's possible within the legal sphere is critical. Yes, because they've got to be on board for the potential of change. So they can look at it with glass half full rather than look to shut things down just because of the potential upside.

Yeah. And they've got to understand, you've kind of got to get to a point where we're quite a. We're a young business, a lot of young people, not myself anymore, unfortunately. But our attitude is very, very strong towards entrepreneurialism and technology. Every one of our accounts is led by a person who is effectively in charge of growing what is a small business within another client within another vertical. So I think that attitude really, really helps. And it means that when we've got people coming into the business, we're prepping them to say, look, this is the sector that we're in. It's not just advertising, it's advertising, advertising and tech because of what we have to put in place with our clients to support them. So therefore, you've got the right attitudes within those people, ideally from the offset. If not, you've got to build it into them once they've joined the business. So I think having the legal teams and all the other kind of central operational teams aware of what is possible and the things that are out there really, really helps, because you can look at taking them on a journey for one product that you're thinking about. But that's a lot of effort, especially if you've got to do it again three weeks later, because you go to them and say, all right, okay, now we want to do this. If they don't understand, oh, that's that same technology that you were talking about previously. So it's probably okay, but just let me cross the I's and dot the T's. Cross T's and dot the I's, then you can exacerbate that process. And you're not sitting there going through the same sort of legal checklist over and over and over again, or going to the business and saying, oh, you know, we've got a bottleneck in legal, we need more people in there. Train the people up, get them fitting for the role that they've got in the business that they're in. And that helps.

It does. I mean, we've been talking for probably two years now about this squeeze on agencies where the ad Platforms are bringing in more and more capability to directly produce ads. Within those channels, AI tools are popping up left, right and center that democratizing the ability of individuals or brands to produce content. What we didn't originally foresee is the impact of consultants wading into the territory with every AI concept under the sun while reducing the headcount left, right and center. So we know that agencies are in this ever more competitive space. The cost per asset is effectively coming down all of the time. But delivering that sustained value is critical. And so the benefit of the Oliver model of having these people with inside the organization who knows what the brands want and has the ability to leverage the wider capabilities of the group is incredibly powerful, isn't it?

Yeah. And I think if you look at some of those people, the relationship they have with the clients, it's not a client, it's the team you work with every day. They see themselves very much as you can argue it both ways. I do this a lot actually, so bear with me. The teams that we put in our clients offices are very much, I am this client. I work with these people every day. This is where I work. I understand the business, I know the brand inside out. And they'll often know a lot more than people at the client will know. A lot of that knowledge retention stays within our teams. If you look at the churn rate of the senior people that you have within some of our clients, the brand managers, the creative directors, faster than the churn that we get. So they're going every couple of years. The knowledge stays within our team. And then Oliver, the sort of de facto marketing agency. So I think that helps massively. On the flip side, it's easy to then take what I've just said and think, well, that's great, but you've got a team at one of your clients and that team are then that client from a mentality perspective, how are they Oliver. But somehow there's some very good people at Oliver within the HR teams who've taken care of this. The ethos of what Oliver is, how we work, what we do, who we are as a business is intrinsic in how these people work every day. So when you then meet these people, when we get together for Christmas parties or things like that, it is an Oliver thing. You don't have teams of people that are standing in different corners of a room saying, oh, we're this team, we're that team. We're on that client, we're on that client. You can tell that they are all of the people. So there's Been a lot of effort put in by the HR teams in order to ensure that that stays, because that it's. It's really important that you understand how Oliver works and who we are and what makes us us. There's actually a presentation called what Makes Us Us. And it goes through a number of different rules, mantras and guides about what it is to be an Oliver person. And I think that is absolutely key because that's what gives us the ability to go in and really take ownership in those situations.

But it's also the ability to bring in different views. If you've got an organization of 5,000 people and they're all operating in exactly the same way as the way they've always done, versus gaining the insights from a particular campaign within a customer, feeding back in an anonymized, aggregated way so that you're getting those insights and able to build on that and create new ideas and concepts. That is original ideation, but based on the insights that have appeared within the customer. That's hugely powerful.

It is. And I think it extends further than that when you, you know, we've been doing a lot of work over the last few years on DEI inclusivity, sustainability. We've got central teams who look after this.

Yeah.

And when they, you know, if you go back a few years when a project came up and someone looked at it and thought, you know, we need to think a little bit more diversely on this because of the target audience or the subject matter or the product or whatever it might have been, we've got central teams within our business who take care of helping the other teams for that. So if there is a particular sustainability goal that a client has got, we have an internal sustainability team who can help out on that and come to it, not just from a, you know, go off to ChatGPT and say, oh, how can I produce this content more sustainably? Or don't do a photo shoot that involves flying anywhere, you know, all of that obvious stuff that comes through, that's not really helping. You need people that are actually in this every day and can come to you with some proper ideas and recommendations about reducing it. And from a DEI perspective, it's exactly the same thing. We've got technology that I'm sure we'll come onto that allows us to look at what a client requirement is and say, right, well, how does that actually flesh out from a DEI perspective? How diverse is the client's attitude towards what they are asking for? How inclusive can we be around this? And we can look at that and flag up. There is a need for some really diverse thinking on this because of the target audiences that you're going to. And we can also look at what teams of people we've got working on those clients. And if we don't see that there is diversity on that team, then we're able to look at the resources that we have across the group through our technology and say, well, actually, I need someone who's got expertise in this different sector or this different area. There is merit for us to start looking at. And there's questions around. Got to be careful how I say this. There's questions around sort of personal data. But if people are willing to share the personal data and say, I'm in this demographic, I'm this sexual orientation, I am this age, I am this ethnicity, if they're willing to share that information, because this is obviously not something that would normally be shared across a business, if they go in and say, look, I'm happy for you to know that about me. Because if another project comes up within the group where someone says, we really need someone from this background, from this ethnicity in this age range, then we can say, we've got someone who can help you out there. So instead of just going, oh, you know, I need an artworker in a different region to get something done in the next 24 hours, we've gone way beyond that and we're able to say, well, actually, I need the right person and not just skill set, but the right actual person, the right mental attitude, the right background, the right upbringing, whatever it might be, to work on this, because that's going to give the diversity of thought to what we're going to deliver to our client.

So you've touched on a few things there. You've touched on Slipstream, which we'll definitely dive into. You've touched on Bios, Breaker and the wider group, which we haven't yet mentioned, in terms of the de DEI and responsible AI initiatives that taking place there. So let's start with improving Briefing and how we get a good understanding of what's a great brief. We increase the ability for the client to generate a good initial brief so they get a great outcome. Here goes. Slipstream.

So our overall product, omg, has an AI layer that we're building around it, which we're calling Slipstream, because of the inherent advantages that you get from that kind of thing. If it's built right. Massive caveat. If it's built right and it doesn't turn into something that does a whole load of things Wrong. Very quickly. Slipstream briefing was our initial focus. One of the biggest problems that you get in the advertising industry is the quality of brief that comes in from a client to their agency. The clients, if you look at the statistics, the clients are 75% of clients. I think the number is, think that their briefing is adequate when it comes into an agency? Not even close.

It's a message through WhatsApp.

Yes. A lot of different regions operate differently in the world. There's one region where everything's just done over WhatsApp. You can't be briefing that well over WhatsApp. It's just not. I can't see it. So traditionally we take briefs from our clients that will come in over email. It might be a PowerPoint or a PDF, something like that. It might be a sit down session where the client's actually taken us through a whole massive scope of what they want to do for the next year. And then we'll cherry pick smaller parts of that to actually put into different projects and campaigns. So within the OMG application, we've had an online briefing module for the last 10 years and it's been very formulaic, very form based, so we can customize templates for our clients. Most of them have between three and five different templates that they would use and it depends on the type of work that we're doing. So if I'm doing a simple adaptation, which is take an existing piece of content and change the language because I want to use it in a different country or market, that's a very small brief. What's the original content? What's the message that you want to say? What language do you want it in? Nice and simple. If it's a big creative campaign, I've got a new product that I want to sell. What are the reasons to believe? What are the target objectives? What are the things that we need to know about the product? Who's the target audience and how do we talk to them? That's all the sort of information that we need from the client, which quite.

Often already exists, doesn't it?

It already exists in a lot of places. So part of the issue was then bringing all of that data together, but also getting the client to think about it in the right way. Because the client just thinks, oh, we've got this product and we know it's going to sell. So there you go, agency, off you go, get on with it. And you get to the point of thinking, right, how can we make this process better? So we were looking at the standard briefing forms that would come in from our clients. They fill it in online. It doesn't stop the client from picking up the phone. This is one of the precursors when I first sat down with Simon Martin, CEO, when I joined the business, and said, yeah, build this great tech platform that's going to run our business. Brilliant. Don't you ever, ever, ever stop a client from giving us a piece of work. So if they pick up the phone, if they scribble something on the back of a napkin, back of a fag packet and chuck it through a window, that is still a brief that we need to get into our system. So don't ever make the system have to. Don't ever make the client have to do it through the system. So we built this and clients have been using that for the last 10 years. So we've got some lovely structured data and we can look at how often a brief was rejected before we accepted it. We don't use the word rejected, obviously. We say, your brief has not been accepted yet and it goes back to the client and they can edit it and resubmit it. That comes back in. And we can look at how many times it's gone through that process. And we know what a good brief looks like. So we know the first one was really bad, we know the second one was a little bit better, but still not good enough. And eventually you get to the point either where the brief was acceptable or you go, this is never going to get any better. Let's just take that brief and then work with the clients to expand it, which we do a lot as well. So we used to run briefing workshops with clients we still do from time to time to get their people better at writing a brief. But when a lot of the AI technologies came around, it's like, well, we can do something here. So clients, some of them, like the structured forms, most of them don't. They just want to come in and go, right, here's your brief, get on with it. So we gave them a very simple text area where they could fill in what they wanted. They can dictate to it if they want to. So they can just click a button, talk load of rubbish. Doesn't matter what it is, doesn't matter what the structure is, doesn't matter if you go off on a completely different tangent and then come back in again. Because what we're able to do is, again, still use a kind of template methodology to it so we can set up for clients what are the checklist of items that we need them to know. We need. So as they're filling in that brief, they've got a little checklist that says, right, item number one is the objective, item number two is the target audience, item number three is your deadline, budget and whatever else. So they're cognizant of what they need to give us, but they don't need to give it in a restructured format, so they can just dump it all in there. They submit that AI, analyzes that, and then breaks it out into the brief that we need to see. So kind of similar to the checklist, you might then say, right, I'm going to identify all the different objectives from this particular brief the client's given me. I'm going to present it back to the client, but I'm also going to do a bit of analysis around it. So I'm going to rate these objectives against the smart objective framework. I'm going to look at giving it a score out of 10 so that when the client gets that back, they've got four objectives in there. That's a seven out of 10. That's a nine out of 10. Oh, that one's a two. I better make that one better, because even I, as the client know that's really not that helpful. So they can then go in, edit that particular objective, edit any other information that they might like, and then submit that through. And then as time's gone by, we've been able to add more and more to this, so we can. Each one of our clients has got a knowledge base that sits behind it. The knowledge base will contain things like, what's your company mission statement? You know, what are your targets, what are your key acronyms you might use across a business? So similar kind of things. Again, when I was talking to someone the other day, he was saying, if you imagine you've got a new starter joining an account in our business, you're going to sit down with them on day one and go, right, here's the really key stuff you need to know about this account. When they say, I need this in the big six, they're talking about the big six markets, not the big six products or whatever it might be. If you get an email from this person, they're really important, answer it immediately or escalate it. These are different acronyms. They might use all of that cheat sheet stuff. We just dump in the knowledge base so the client doesn't have to think, oh, they might not understand what I mean by this anymore. That's all taken care of by the engine that sits underneath so we can Then take the brief from a client and we can look at the mission statement of the business and go, this brief doesn't quite align with the mission statement. Maybe you should think about doing this or doing that. So the client gets presented with a summary that says, right, you've given us this, this little brief. We've broken it out into this for you. Does that make sense? Does that sound good? Yes, submit it. That then comes through to us as the agency, we can look at that, accept it or not accept it, and it goes back to the client. But by this point, a lot more briefs are coming through because the quality has been increased at the. At the source.

So you've augmented the brief based on the information that exists within the knowledge base already. So you've got a rag database sitting behind the scenes that's absorbing all of the information that's been shared in. They're then putting in their brief. It's being analyzed, it's being checked against specific things. You're reframing it back into a consistent brief format, showing that back with the customer and getting their feedback on it. When they want to improve it, they're then putting through something that's automatically better because it's been enhanced through this process. So you're getting rid of the possibility of crap in at the top of the funnel. You're encouraging them to produce a higher quality brief by actually doing some of that work for them. But that's happening automatically.

Yeah. And there's some little gamification things. You know, the scoring that we're giving to this, we, you know, we've made it a lot more visual. So as that brief comes back, you got a little star system across all the different sections that say, you know, Your objectives are 3 out of 5. Nobody wants to see that. Right. I'm going to. I'm going to do better at that.

You're failing your brief exam.

Yeah. So when that brief then comes through. So we've then got a secondary set of questions that we can ask that are tailored towards our internal. Our internal teams.

Yeah.

So where we've been saying to the client, you know, the prompt might say, you know, how well does this brief meet the mission statement of this business? Which is blah, blah, blah, blah, and it gives something back to the client. When that brief then comes through to us, we can rerun the analysis and it will pick up another sort of 10 or 11 different things. There are things that are very specific to us as Oliver, in terms of how we work and how we rate the quality of what we feel a brief is and what we think the quality of work can be that we can output and we can make suggestions around how we can accelerate that and get further up the quality. But there's also things around the sustainability, again, the dei, but questions that you might not put to a client just yet. As a client, you don't want to submit a brief. You don't want to submit a brief and get told, oh, that's a bit racist, isn't it? You know, you've got to be careful what you're putting across to someone when you're trying to elicit that information out of them. But we still want to catch it in that process.

You want to maximize the input and then have some.

And then have someone be able to do some validation on it. So we can then do some of those other validations. Again, still using AI to say, right, how can we increase the diversity in this? How can we increase the inclusivity? What can we do from a sustainability perspective? But bearing in mind not the client's sustainability guidelines, but Oliver's sustainability guidelines as well, and all the extra knowledge that we've got. So we're then able to flag those things up internally that our teams are able to look at and go, okay, I can see what the client meant when they said that. But my knowledge of the account, the business and the brand, I can see how we can extrapolate this and make it into something different and better. But a lot of that information is being brought together by the technology. So where a person used to have to go and do that, go and get the insights, go and get this, go and get the tone of voice, information and all that kind of stuff or things that were just in their head, it sped that process up by going, right, okay, I've got all the things that used to take you two or three days to go and do, dumped it all in this brief in a structured format. So now the client has seen. Has submitted a paragraph, they've seen a full page of structured information and gone, yeah, that looks brilliant. I've written a really good brief. The account team have then picked that up and gone, right, okay, I've already done a lot of my initial analysis of this based on the information that we've got within our rag stores to then be able to start crafting the response to the brief or the creative brief that's going to go into the rest of the business.

It's magic, really, isn't it?

Yeah, we'll call it that. Yeah. But I mean, when you look at it. Yeah.

In terms of the culture shift, in terms of the quality of the input originally given, it just empowers people to have a far better outcome.

But I think that's the thing. It's empowering people to have a better outcome. It's not taking the people out of the process. No, it's not yet saying we don't need the client to brief us. There are things that we're looking at to help us be more proactive and say to a client, we're seeing this happen in your sector, your competitor has just done this. Perhaps you should consider doing something to combat that. They're things that we can do, but you're still going to need a client thinking about, all right, what is the best way to come and combat that? What can we actually do about that? And the technology has given us the ability to get a lot of that information in front of people to be able to make that decision. So the term human in the loop is, I think, key and something that I like to think will always be there. I think you're wrong, but I know you do. That's why I said I like to think, and I sort of looked up. I think you're absolutely right. It won't always be there. And I think people, again, this is either a conversation for a little bit later or in the publisher. But the roles that we've got are going to drastically change. We all know that. But for the time being, we're able to give people more information in a better structured way and higher quality information in a better structured way for them to be able to make the decision about what they need to do. And I think that probably one of the responsibilities that we all collectively have is to try to take people on that journey. I don't want to be responsible for drastically changing the way that people work overnight. I don't think that's a constructive thing for people.

I don't think people can change that quickly. So it has to be iterative. It has to be a journey. You have to demonstrate, here's the benefits now. And there'll be still some people that struggle even with that slower level of change. So you can't go, you know, this entire process that you used to do, we don't need people in anymore because culturally, businessly, it just doesn't work. You can't make that shift, but you can augment the process. So in the way that you've been describing, you've gone from having a brief process which was pretty inconsistent and had varying levels of success. Based on who was briefing in, essentially. And there was an awful lot of work internally to compensate for poor quality briefs where you had project managers and creatives trying to understand and dive deep and feed back to the customer. And all that back and forward is time, which ultimately is cost. So improving the quality of the brief in means that you're accelerating that process. You can deliver more great quality work. And all you've done is look at, okay, how do we apply in this space? And so if we look at the lens of where can we make a small change that can make a big difference, but works in conjunction with people within the business, then it's relatively easy to implement and to get traction with. If you said overnight, and I know Oliver will never go down this path, but we're not going to use people for this part of production. We're not going to use OMG anymore. We've got a newfangled thing that just does all the thinking and produces all the output for us. There's no way it'll get traction. You've got to take people on a journey. And I think the way we apply this technology is it is in combination with people, it is to improve the greater capabilities. But if you're not doing it, you're being left behind 100%.

I think even if you go back any number of years, the whole automated content production processes that we've had available to us in the advertising space for a number of years now, just for bulk producing print material and things like that, there's still a lot of clients who don't necessarily use that either because their content isn't suitable enough to template or because they haven't got the right processes in place in order to do it. So bringing people on that journey has always been difficult, will always be difficult. And with the rate of change that AI has brought us, it's going to be even harder. And I think there are some people, people need to be really careful not to get left behind. I think, you know, you've said it before. Many people have said it before. Probably you're not going to lose your job to AI, but you'll lose your job to someone who's using it. I used to say that right now.

Do you know what I mean?

Appropriate. You're going to make it even worse now, aren't you?

Yeah, you're going to lose jobs and it'll be to people using AI. I think there will be people who are displaced.

I absolutely agree there will be. I. I have some issues with the ethics and morality of how we get to where we're going to get to. There's a responsibility that we all have to each other.

Yeah.

And I think the drastic rate of change that we are doing because we can do it, not because we should do it, is something that doesn't always sit particularly well with me.

And we're aligned on that completely, but I don't think we can.

It's not going to stop because there.

Was an attempt to stop developments greater than GPT4 for six months.

Yeah.

Where did that go?

Didn't go anywhere because why would it stop? Legislation has always been slow to keep up with any technology. Keeping up with AI, it just doesn't stand a chance.

No. I mean, it's a greater rate of change than has been experienced in mankind's existence, which is mental.

Also, what is AI at that point? If someone brings out legislation, someone's going to find a way around it anyway. So it's a journey we've all got to go on whether we like it or not. And I do like it. I've worked in technology my whole life. I'm the one who sends the emails around at work going, look, the robots.

Are going to take over.

Look, there's more proof someone's just gone and done this.

Is that really taking people along with you? Rob?

One of the things, I drive my wife mad a lot. I haven't done it yet, but I will possibly do it. I will always argue both sides of a debate. So my wife and I will be having a conversation about whatever I've done wrong or whatever it is, and I'll be defending one side of the argument and then I will flip because I, like, I don't. I'm not going to say that I overthink. I just like to think everything through, every possible scenario. Not to the point of procrastination or anything like that, but I just like to make sure that I've debated the whole thing. So we'll be talking. We don't argue, actually, we just talk. And I'll be talking from this way, she's talking from that way. And then all of a sudden I'm over here and I'm going, but yeah, but what if we did do it that way? But then that's not going to work and this isn't going to work. And then she's like, what are we doing?

So this is how you get around buying another classic car? Is was.

Until you said that. Now she knows that I'm in all sorts of trouble. I'm not going to look at One today, Timothy?

Definitely not. I know.

No, not at all. So, yeah, I think I like to look at things from all the angles and sometimes when you're in conversations and talking about technology, especially you've been to our tech summit. So once a quarter we have a three day tech summit and we bring all the tech people from the business in. We fly a few people in from different locations as well. Other people join remotely and we have three days where we talk about what the goals of the company, how can we meet them, what technologies are out there that we can leverage, how well have we done since the last meeting that we had in meeting the actions that we've got and are we on the right trajectory? And it sounds like a lot, and it sounds like meeting once a quarter is a lot, but you kind of got to do that if you want to stay up.

Now the change you guys have been through over the past two years is phenomenal. And those tech summits are great and they're heavily AI focused now, aren't they?

Yes, massively so. We started introducing it as a small sort of two to three hour slot a couple of years ago, whatever it was, and now we've got a couple of hours to cover the other stuff that's not AI related. But in those sessions I will bring up that the LLMs are all lying to us. And if you ask a really, really, really specific question that needs a really specific answer, it's likely to be 80% wrong, but with a confidence rating of about 100%. 100.

Well, if you believe in false maths, beyond, beyond 100%.

So I used to say to my kids, you've got all the information. You know, what's the point doing homework, what's the point doing exams? All the information out there on the Internet. Like we used to argue about, you know, your math teacher saying you never have a calculator in your pocket. Guess what? I've got a calculator in my pocket. I don't even need my pocket. So when you're talking to the kids about, you know, doing their homework and why they need to do it, it's that I used to say to them, you know, you've got all the information out there in the world and you've got access to that on your phone at any point, your job now isn't learning that information. Your job now is discerning what information is true and what information is not. And that's before AI. Now all bets are off. You haven't got a clue because there's so much data that's being generated and we're seeing massive exponential increases in the amount of content that is being produced. There's the whole. Social media's a whole big lie anyway because everybody's just posting how great my life is. And it's not actually that great because you just posted that one little snippet from that holiday that you went on where it rained. The rest of the time, people are being very careful.

Don't be careful.

Okay, fine. People are being very fake around what their life is and projecting that onto other people. And then we, we all feel miserable because everybody else's life is better than me.

If you consume.

If you consume it. But it's hard, especially when you look at the younger generations. Why wouldn't that be true? Why would someone be posting, you know, something that's not true? Why would they be telling me a lie? So it's gone from being understand what's right, understand what's true and what's not to. I can't really help you anymore. You know, perplexity is. Perplexity is my new favorite. I can go and I can ask a question and it's got all the citations of where all the stuff came from. So I can actually look into more around what the source of that information was now. It's great, but there's limits to how much time you've got when you're just asking one question to go away and then look at all the sources. But you've at least got something that's saying, this is where I found this information from. And you can start to make some decisions. But the rate of consumption of content in kids these days, and they are the next generation that's just going to start coming through is so dramatically greater than it was when I don't. We didn't even have this when we were young. But the ability to denote this is true and this isn't is becoming a very, very gray area. And that's a very difficult thing to manage going into the future. And that's why I kind of like to bring these things up with people to say, look, someone has to take some responsibility for this. And if the people that are developing the technology aren't, then the people who are using the technology are going to have to. And I don't know what the answer is before you, before you ask, but there's an awareness that has to go with this stuff that, you know, we're building things because we can, not because we should. And a lot of it is great. And as a Technologist. I think it's fantastic. And on one hand, I can't wait to have a robot walking around my house doing, you know, go cut the grass, rebuilding the engine on my car for the, for the second time. But all of those things that they're going to be able to help with are fantastic. And then you've got the flip side. You know, that Hollywood has been warning us about this for, you know, years and years about what the robots are going to do, but, you know, that's not how it's going to work. But I think having an acknowledgment where we are is absolutely key because when I then look at the rest of my family, my brother, my sisters, they've got no idea what's coming, no idea at all. And they're an age where they don't necessarily need to worry about it to an extent, but they've got to worry about it for their kids. We've had the conversations about, my son wants to go to university. What does he want to do? Computer science. Don't go to uni, get an apprenticeship. And that's absolutely the route we want to go. But if he doesn't, you sort of think, well, okay, go to uni, get the life experience anyway. There's some validity in that to an extent, but it's not really what you want to be going to uni for. You want to be learning. And that methodology of learning isn't necessarily suited to that market anymore. So the shift is absolutely huge. And I just think that collectively we've got to do something and be aware of what we can do to help people progress from what is today into what is going to be tomorrow and the day after and the day after and the year after, because the rate of change is just going to be catastrophic.

At one point, we've already got huge volumes of content being automatically produced with lifelike looking people on social media. And this is just with level one. So this is just conversational and early generative AI. This is without the intelligence, the ability to reason. This is without agents, the ability to operate autonomously and complete actions. So the rate of change we've seen so far, I think, is slow compared to what we're going to see in 2025. And that's quite a daunting prospect because it already feels like it's coming out of a fire hose and you're trying to interpret information and change when it's coming at you full bore.

Yeah.

So I think it's very exciting to be time to be in Tech. Never been a better time to be in tech 100%. But how we manage the understanding of change and take people on a journey so that as many people as possible are able to benefit and not just be left behind is challenging at a family level, let alone an organizational level or a country level.

Yeah. Was it a person is smart, people are dumb. It's very hard to lead a big group of people when there is so much influence from elsewhere and a lot of that influence is unqualified and just pulling them completely the wrong direction. It's, you know, a lie will proliferate a lot quicker than the truth will. You know, if you go and you try and calm a conversation down around what's happening in AI. If we were to go down to the pub now, start talking about AI and you talk about the benefits that it's going to bring to the average person across the uk, and then I go, yeah, but it lies and it's really, really confident in the answers it gives. But it lies and it's tried to replicate itself and it's going to take over the world. What ones are they going to remember? They're going to remember what I said.

I mean, this is the way the news and press has always worked, that people focus in on the negative 100%.

We don't watch the news at home because of this and it's only going to get worse.

Well, I think there are ways to limit and control that. So if you think of the consensus model within anthropic models, if you think about bias breaker, where we're estimating and construing any bias that could exist and then compensating for it. There are ways of putting in checks and balances to ensure that content is true, content is unbiased. So I think the whole large language models being incredibly confident bullshit machines isn't going to be a sustained problem. It is just with the current vintage of technology.

Yeah. Which I think is very important because at the moment they're very much a mirror to humanity. They've learned from what we have. We've been a bad parent to AI. It's learned off all of our bad habits. That confidence versus accuracy issue has not come probably from something programmatic. It's come from looking at the data that's all there.

We don't know what's true really. We're only going on judgment. So how you then program that? I think the XAI approach of trying to build the ultimate truth seeking AI is a fascinating angle of looking at it because anthropic are going consensus. Evaluate the input, make sure it aligns with effectively our ethics model. Check that the output aligns with the ethics model. OpenAI arguably have taken a Californian approach to it and have something that's relatively woke in its reinforcement learning. So there's different approaches to solving this coming from different directions. And I think that that means we might end up with some very, very different AGIs.

Yeah, I don't think that's avoidable. Whether it's the training data or the people that are building it, it's going to come with inherent bias from those people. Regardless of what you've tried to take out, there will always be some foundational level in there.

Yeah.

The ethics and morality one is a very interesting one. You do sort of question what their ethics morals are and how many of them really understand what ethics are and what morals are in terms of how.

The AI or in terms of the companies.

The companies and what their attitudes are. A lot of these companies had nice big ethic departments when they started. They haven't got many more.

What was Google's mantra? Wasn't it do no evil and then they bought Boston Dynamics.

Yes, yes. But that's kind of. My point is that these companies are.

Doing.

What they can, not just what they should.

It's a race, isn't it? It is a race to see who can produce AGI the most quickly. Because once you've achieved that point, your business or your entity or your country is arguably ahead for a long time.

Yeah.

And so this race is then subsequently drives huge investment, requires a capitalist model. You've got to build these monster data centers, you've got to have compute available, you've got to have as much data as possible. And so it's almost a race for so many different things at once. And the morality of that isn't in check. It feels like a developed superintelligence or bust race.

Yeah.

And that creates great capabilities. How it gets balanced with the needs of society is a challenge.

Yeah. I have a very close friend who's an ethics professor and we talk about this a lot over beer. And he sort of likens AI to a knife. It's a massively useful implement. You can buy a knife and give. This is not his analogy that I'm going into. But if you give one person 50 quid a knife and another person 50 quid a knife, one of them will go away and buy 50 quid worth of food and prepare a lovely meal with that knife, and the other person will take their knife, stab the other person and take the 50 quid. They're very, very extremes just to paint a picture. So you've got these massively diverse mindsets around what to do with the tools that you're given. For whatever reason you have to do that, that person stabbing the other person might not have eaten for God knows how long. There's all sorts of reasons as to why people do what they do. But all of those attitudes and the data behind all of that is what these systems have been learning from. And I think that's part of the issue that we have in terms of, as I was saying, the confidence levels, quality of data and everything that becomes this gray area of what's accurate, what's not. And I think I'm not sure there's enough. I say I'm not sure. I'm literally not sure. I'm not quoting any numbers on this one. But I don't know to what extent people care about that. In terms of the ones that are building the tech, do they care whether they're building a knife for stabbing or a knife for cultivating?

I don't think any of the AI labs are going into it with building a knife for stabbing, but I think it is a race to build the most intelligent thing they can that can potentially take action. Yeah. And that could accidentally produce.

And that's the point. It's the building it because you can, and not necessarily thinking through all of the reasons. And it's not just, you know, I'm not sitting here thinking, you know, robot's going to take over the world. We've had this conversation many times. It's the small nuances that come along the way. You know, if they'd been thinking about that from day one, would there be a difference in the confidence levels that you would imply from an answer that you get back from AI if it wasn't quite sure on what the answer was? When we were developing Slipstream out, we were very careful about the prompts that we used. So that if it made an assumption, it was very clear that it made an assumption. So, for example, a brief comes in, we're analyzing it for the market that the client wants to operate in or the markets. And there were examples coming in where a market hadn't been specified and we literally had no idea. And in some cases a budget was then specified and the budget was in pounds. So we wanted to make sure that when that comes back, we don't go, it's for the uk. We say it's for the uk. We're assuming because the budget was specified in GBP or someone mentioned this particular language. So it will be either in either of these two countries that have that base language. Whatever the reason might be, I think we've got to be clear about that so that we can make a better decision on what we do with that information to go forwards.

But that's effectively taking a vast capability and applying it in a very narrow way to solve a specific problem. And that's the way that we're having success across industries with AI at the moment. We're not relying on a large language model that can produce great content that isn't always 100% accurate to run a business. We're looking at how do we break down the steps in a process and where can we apply the intelligence within these capabilities to solve a specific problem. And so if you're daisy chaining those narrow events together, then you can build a reliable solution?

Yes.

If you're saying, now make me a new business model, it might be amazing, but it's unconstrained. So it's making sure that we're applying these capabilities in bite sized steps now while we build the confidence and that we can check that those outputs that are being produced are accurate and on brand is critical going forward, will we put more faith into wider applications that are delivered by agents acting in sequences and together in ways that we may not understand or even be able to monitor? Perhaps. But at the moment, focusing in those super narrow applications where we can take these amazing capabilities and make things more efficient is key, I think.

Yeah, you can take that right back and liken that to human nature. If you're a jack of all trades, great, you can get a load of work done for someone, but it won't be 100% to the quality that someone else could have done it. You can get builder in who can do a bit of woodwork. They're not going to be as productive or as good as a dedicated carpenter is going to be. And I think that's one of the advantages of looking at the agentic flows that we're going to be able to put in place now because you've got little experts. And as long as you train those little experts correctly on the right data, which means you're not training them on all of this data, you're just training them on this bit. I'm going to teach you how to be a carpenter. This is what you need to do. You've got, you know, a couple of tools and some wood. That's all you need to worry about. Someone might ask you to build a thing and you're going to be able to say, I can't build it because that's not my job. You need to go to someone over here for that. You need the plumber, or I can do that for you, and I can do that. And I need to speak to the plumber because he needs to pass a pipe through one of those floorboards. So I'm going to need to put a hole somewhere. So I'll go and ask that agent what he needs to do. So if you're building a house, you go to a builder, he's going to bring in all those little experts. If you go to a jack of all trades and he's like, yeah, I can build your house, but he doesn't bring those experts in, you're going to be in a whole world of pain.

So I'm incredibly tempted to go into a more detailed version, but I think we're going to save it for another episode because there is definitely a conversation to be had about how those suite of experts operate as an agency.

Yeah.

So let's save that for another day. It's been an absolute pleasure, Rob.

No, it's been brilliant. Thank you very much.

Thank you.