The Future of Creative Strategy: How AI Is Transforming Performance Marketing in 2026

0:00Going into next year, what is going to be one of the most important and big swing roles that we need to bring into the business. and I think that is the role of an AI videographer.

0:12Got our first Club Neuro event happening on Saturday. How many people are coming? A hundred. Sold out pretty quick as well. How's the events business? It's a fucking nightmare, to be honest. To be fair, I can't say much because Liv's been leading most of the event organisation. but yeah i mean it's it's it's definitely it's one of those things where you you think you're you don't do it for the for the profit it's there for community building purposes um which is is really valuable and it's lining up with a new new product launch that we have as well but i think it's it's there's definitely a lot more hidden costs involved than you think i just always remember arjun from free souls telling us that it was a nightmare yeah they They smash it though, don't they?

0:59Yeah, they do. They've obviously operationalised it and they're so best in class at that stuff. But it just always told us. But yeah, what's the structure? Yeah, so we've got, obviously it's for a neurodivergent brand, so ADHD and autistic people that will be attending. And we have one panel. We have one doctor that's coming in to do a talk. And then also we have like round tables with group we'll split into groups and do some like question prompts and stuff where people can like meet and get to know each other and then i think we're going to end on just announcing the the new product that we're we're launching which is exciting so it'll be a good learning it'll be good to see how it goes but i think um i'm so bullish on community and just irl and i think there's there's so many case studies for that free soul pure sport um especially if it's a brand that you know really did you know really should have a community attached to it which isn't every brand either but i think there's a lot out there that it makes sense for yeah i think if any i'm sure there'll be plenty of neurodivergent people watching i feel like we over index as entrepreneurs and people who operate businesses as neurodivergent so they should check the brand out for sure speaking of that film i swear incredible anyone listening if you've not seen it go you will cry yeah I think it's one of the best films I've seen in years yeah like in terms of like the story the acting the control of emotions the the quality I don't know it was just like flawlessly executed yeah give a bit of context just of what's what it's about so it's about a one of the a guy in like rural Scotland who um develops Tourette's round about probably about 14 but he's obviously the awareness around Tourette's in the 80s isn't it yeah and he gets just marginalized by everybody his family school everybody in life and then he meets this individual who kind of pick up like I don't know empowers him I guess the first person to like accept him for who he is and understand his condition because if you I mean makes you think about something like Tourette's in in times where there's no education around it like imagine just the trouble you'd get into just swearing at people on the street, even down to police and things like that that just wouldn't accept it as a condition.

3:29It's the best use of, the best control of tone I've ever seen in a piece of cinema, so everyone should definitely go watch it. Yeah, you didn't warn me and I was crying my eyes out. It's tough, it's a tough watch, but in the best way it's so heartwarming at the same time, but yeah. Exactly. Go watch it. Cool. Anyway, welcome back everybody. we're on a roll of solos again might recognize these outfits from the from the last one we did we do wash and change um but yeah obviously last solo episode we we focused a lot on incrementality tactical things that are working since then lucas has released his master class on creative similarity and how we're thinking about process and adapting that now everyone if you've not watch that pause pause now go on youtube subscribe first then watch that because i think it'll give you some really good insights into how we're thinking about operationally and output wise but i think it's worth starting there still as a bit of a rehash um at the start of this episode from a top level perspective yeah so maybe let's start with you just giving a quick rundown again of what creative diversity means now um and a breakdown into like that piece around entity ideas creative ids and what those differences are yeah so i it was kind of felt like this all this narrative started a few few weeks ago with the episode we did with sam meta released some better guidance on creative similarity in the pursuit of better incremental reach and more spend and more growth and why brands might be struggling with that we got some data from a client that sort of side-by-sided some ugc that was two different creators two different messages but were being grouped as the same asset and i guess that's the the perfect example of content idea entity id right so the content ids are different but they're being grouped under an entity id because the diversity in that content is not being recognized as significant enough by andromeda or by meta therefore that's limiting your audience pool through those through that group of assets um and it's limiting your reach as a function of that so it's like right with that as a starting point how do we think about adapting what we've been doing previously to get the most out of the algorithm and the ecosystem we find ourselves in as as brands and operators yeah exactly and i think yeah essentially if a creative the goal is to be assigned a new entity id right because it will start fresh then in terms of audience and learnings and unlock incrementality so true creative diversity is is important compared to minor tweaks to ensure that we're unlocking net new audiences so i think right now us and i think a lot of others will be looking at iterations um as the thing that is going to be impacted most by this change because i think you know we do a lot of iterations on a week we essentially right now set up iterations with weekly sprints so we are launching net new uh campaigns and uh into the ad account and net new creative cycles every 30 to 35 days and then on top of that we are doing weekly iterate iterative sprints we were doing some creative with those iterations that were more visual changes but a lot of them naturally came down to messaging tweaks whereas now the way that we need to be approaching iterations is very different because we can't be we can't be doing those small messaging tweaks in an iterative way because they will be grouped under the same entity id and therefore essentially piggybacking off learnings that are already within the ad account yeah and i don't want this to think i think the easy thing to think is like right we shouldn't be making iterations of any concept so we shouldn't be taking a podcast ad and doing like six to eight hooks or six to eight variations of that of that footage which i actually think is completely false because i almost feel like you need to separate iterations of existing performers and iterations of new concepts and those are two like distinctions that need to be made prior to like really diving deep into the the process on a on a weekly basis because whenever you're testing that net new that net new piece i think higher volume of different hooks and messages gives that new visual messaging vehicle a higher chance of hitting and succeeding exactly but it's then once we've found something that's working what's the iterative feedback loop and process look like off the back of an asset that's winning 100 same messaging tests are still working obviously for net new concept variations because there's it's it's launching into the ad account and there's nothing already there with that so using the same visuals for net news isn't a bad thing because the goal there is to identify which messaging resonates most with the target persona and how to communicate that message most effectively yeah and i think if you stop doing that your cost of creative is going to be astronomical so we need to we need to retain that approach to big swings lots of iterations across that that big swing concept in the pursuit of finding something that we can then run with but on the iteration of something that's really driving performance already whether that's static it's a ugc it's how do we inject more visual diversity across the process exactly and i think the so what the the impact that this is going to have let's let's park like net news for a moment and just look at it through iterations because that's what's going to be impacted most in terms of how we need to change our approach it's going to add more operational load um and you know and and resource to the process because by doing more visual changes within those iterations, that's going to put more strain on post-production, more strain on design.

9:31Those teams that are going to be making those changes, it's also going to put more strain on your creative strategist because they're going to have to be more intentional with how they're briefing those iterations in. I also think it puts more strain on your brand department because you have to find a way to do creative diversity whilst remaining on brand if that's something that's really true to your business in a performance in that in a performance lens so it's like compounds all the way through the org in some ways it's so true i think it's a really good point actually i think brands that are very obsessed on it almost completely contradicts the idea of brand consistency doesn't it so brands that are very obsessed with that consistency meta's update does not align to that and you need to almost break out of or re-look at how you are approaching brand versus performance 100 i think it's like trying to find ways to remain on form a common identity through assets but allow for more flexibility and message and presentation yeah that could probably be a whole podcast in itself so let's park that but looking at directly at performance creative process you mentioned creative strategists you mentioned post-prod production it's just more cost right it's just like it's going to cost more to output the same amount with higher visual diversity exactly high diversity and the thing the way we've been thinking through this is how can we drive as much if not more impact but at the same you know same amount of resource input um which again i think in some ways isn't possible but i do think we're it is possible in in others and sorry what are you going to say no no i was going to say something's got to give somewhere exactly something's got to give somewhere and i think the the best way to ensure that iterations are still driving as much impact as they were previously is to so what we're doing is obviously we're moving from a model where we are going to we were producing we were doing weekly iteration sprints um every single week to a model where we are doing bi-weekly iterations but and less of them but more intentional big visual changes so let's say you've got a hunch based on reverse engineering from your your volume your spend targets into the amount of volume you need to produce you we did a whole episode breaking down how we get to this point but you'll have the amount of iterations that you want to do across a month what we are now starting to do is reduce that by 30 40 50 percent doing less but more intentional and then weighting the other 30 40 50 percent towards more variations of net news because by doing so we are still driving the volume through iterations but they are bigger visual swings and then we are giving ourselves more chance of success and being able to do more messaging tests of net new concepts when when we're getting to that campaign batch makes sense and i think that that word intention i think is super important because i think like the iteration cycle if you're not careful can just fall into that category of like busy work like just throwing things at a wall just changing things for the sake of it so it's like really being it's more creative i guess and it's and it's at the heart of it right to drive that exactly but i think because the way we need to be thinking through this lens is like you could do triple the amount of iterations but you know 70 percent of them are getting grouped under the same entity id so it's absolutely worthless activity to be doing so we need to be thinking at it you know we need to be thinking through it and it's in a different way to ensure that we're still driving that impact and you know helping drive incremental reach which is obviously so important yeah that's we always use strategic volume internally internally and i think it's that balance it's like people will always be like quantity versus quality and i don't think it's it's more like how do we continuously drive more volume without losing quality and impact off the back of the output and this kind of refocuses minds a little bit more on that i think um again it's it's common for us to see brands just making a shed load of creative and not being able to grow spend and see an incremental reach decline number one signal that that's not quality that measure of quality and quantity isn't at the forefront of your operation needs to be exactly and i think this is why essentially where what this means in in many ways in the way that we can use ai as a tool to increase efficiency whilst also aligning with creative diversity is leaning into things like sora to vo3 and nano banana for designers and statics to be able to still drive that volume but through bigger visual changes without adding huge amounts of time to the process so this kind of took me on a tangent of thinking okay going into next year what is going to be one of the most important and big swing roles that we need to bring into the business at least from a delivery perspective and i think that is the role of a an ai videographer or visual prompt engineer is probably the better way to put it yeah um and the way the way i think about this is like we need to you need to be treating like ai production the same way that you would treat production because a creative strategist or an editor they're not they would never go out and shoot the content themselves you'd have a you'd hand that off to a production team with a brief to shoot that content that would then come back for ticketing to then go into post prod and i think we need to be especially now with the volume that we're going to need to be producing through ai video and image gen and how how much that's even advanced over the last three months since we first did that vo3 episode um it means that this role is going to be incredibly important and be viewed almost as its own vertical and its own production department but through the lens of visual prompt engineering 100 since that vo3 episodes we've had i think nano banana was before that right no i think it was after yeah and the volume of assets through nanobrann the amount we're using that in our process is like crazy we've then had sora and then we've had vo 3.1 and it's just like again it's just moving so fast um but yeah i really liked the way you kind of thought about that or introduced that sort of mental model for how do we bring in ai and like it's almost like compartmentalized but like and without having to really drastically change our current operation it's like bringing them in figuring out our way of like what goes into that production team what goes into internal production team and then then again hypothesis would be like you'd expect the volume of assets produced by the ai department to increase as a total percentage of asset volume over time to probably i think we'll get to a point of like 80-20 AI and most the 20 % will be the very human dialogue dialogue long form founder customer things that we've talked about on the previous episode authentic authentic I guess is the common theme through all of those whereas AI is able to handle a lot of the products probably a lot of the UGC to some degree a lot of those a lot of things that that maybe don't require that same level of authenticity 100 % and I think the beauty of it is like it really it's quite simple to slot this in because you've already built the process for production or UGC briefing and it's very similar but you just need to treat it as a different type of production in many ways so just like we brief UGC creators to shoot content or production teams to shoot high production content your strategists will start to brief AI videographers or visual prompt engineers to create AI video alongside real content production.

17:23So initially that will be hook variations as an example, product shots, and also B-roll elements. Because now, you know, we didn't have this in the UK. It took a while to come to us. But the ingredient to video feature on VO3 is incredible now. So you can upload an image of your product and easily put it into a scene. we had one of our brands nothing fishy we'll put the example on screen which is the product just floating in a tank and the fish are all going up to it and it looks super realistic it looks it looks really strong um so you know we're thinking of it through the lens not only of how this can drive more creative diversity but also what elements can we use that's going to reduce efficiency and strain on say the production unit yeah and it's that combat to like you see the So you kind of see this creative diversity requirement as a cost pressure, and then you see the rise of AI tools as the opposite.

18:26Exactly. It's a relief to that from a cost per asset perspective, and that should, over time, continue to take effect. Yeah, I think that's going to be a huge unlock, and then it's just a process problem, I guess. Like, how do we optimize the inputs to that team even make it as efficient as possible yeah exactly did you watch i well i know you watch because i sent you it the andre carpathy i can't say his name the guy who used to be um he used to work for tesla self-driving ai i also believe he had a founding team researcher role at open ai where he made quite a bearish case for the current state of the ai and almost pointed to a bit of a bubble in a way that you would look at the dot-com bubble of sort of early 2000s yeah really interesting i think i mean what what were your thoughts because i it was the first it was the first it's a unique perspective but one that i've not seen especially for people that deep in in the ai industry my first thought was that's the first podcast that's impossible to watch on 2x like it's they talk so fast it's a really i can't it's dar i think it's darmesh the guy the podcast he's on i think um we'll put the link below yeah it was a it was a really it was it was a really like it's always good to have a counter argument to what is just such a bullish environment um i think i also saw a really interesting stat the other day that um gdp growth in 2025 has been 0.1 % when you strip out investment in data centers.

20:07So if you remove the investment in data centers, which when you look at the flow of money there is quite self-like, it's like a flow chart of just companies investing in each other in different ways, like Microsoft, NVIDIA, OpenAI, etc. Which I think paints a bit of a bearish picture maybe for the 2026 economy when you strip that out and maybe how the consumer versus tech is looking. Yeah, especially when you see China now starting to actually build these centers underwater in the ocean. Yeah, China's, that's another point. I saw Brian Chesky from Airbnb talking about how they're basically only using Alibaba.

20:57Alibaba have got loads of models, which I actually didn't, I wasn't aware of. but they're faster, cheaper and the quality is just apparently very similar but obviously it's just a Chinese company and he's using most of their internal AI processes run off Alibaba models which I need to look into but segue but yeah, it was interesting to get like it was quite, felt a sense of almost relief that maybe we're not all doomed within like three years with AGI coming that fast but yeah, fascinating to hear his like bearish case on the and I'm not going to go into detail on what he said because I'll butcher it, everybody should just go watch the video like a bearish case on the sort of ability of LLMs and our understanding of building true intelligence and how a lot of that requires reinforced learning and the way that we're currently building these LLMs isn't creating that environment that creates a system of reinforced learning over time based on incentives and outcomes from from tasks that are completed interesting hoping i haven't butchered that no i think you said that you said that better than i ever could that's for sure um but yeah it was just interesting to get a bearish case on on on that whole environment um from from someone with such technical understanding yeah it's not it's the first time i've seen the the other you know the other case um demonstrated especially after coming off the back of like watching that podcast with the the ai godfather or whatever he's called um which was like very uh very depressing watch um have you seen by the way have you yeah i had a played with a um chat gpt's new browser coming off google yeah i think that's a bit of an overrated drop do you think like i think it's interesting like it's obviously built built on chromium which is chrome's base like i think it's it's open it's what it built chrome on top of so it's it's like i don't it's not chrome wrapper that's doing it in injustice but it's an interest it's an interesting release but like just feels like they're just churning out so much stuff yeah and it's product roadmap is wild and it's like i also think the agent builder that they brought out in that big product release like i don't think the agent builder is great either yeah and that doesn't mean that it's not going to become amazing but yeah it was just i didn't i haven't found it that like it's not i don't use it i've downloaded it and played with it but i haven't found it that like appealing to interact with on a daily basis yeah if there's one person you should you should back that sam yeah he's uh but like because i you look at the first sora as an example pretty poor product no one ever spoke about it after it was released sora 2 on the other hand has created waves in the industry and i think you you see that a lot with their v1s and you know getting it into the ecosystem to to start gaining context and learnings and i think i think it personally it's going to change the whole way that we operate within browsers like the fact that you can you have you're just so easily accessible to agent features there and then like going to a website and saying i need to book a flight for this find me the cheapest cost and it will just take over the screen and do it.

24:21The easy access to the LLM chatbot on the side to then speak directly with what you're seeing on screen. I think there are so many, I can just see how the way it's going to change the way we're using browsers. I think it'll get better and better over time, 100%. I think that'll be not a normal, that'll be the way we interact with the internet. They also acquired Sky, which is a company that allows them to, it's like a does the same as they've managed to accomplish there for the browser but with the whole mac os system so they can take control of your whole computer um so they're quite a business that allows them to do that with a team viewer yeah kind of but like ai integrated um i i'm still very bullish on google as an ai i use gemini more than anything i think as a tool like i use their suite of ai tools what i think more than than i do open ai i think it's got a better con it's got a really good context window especially the more recent updates uh obviously the video aspects and some of those elements nano banana vo3 also like the we've we've we're working through a lot of context documentation at the moment how we can build that into like better prompt engineering and we i'm working on like a really strong prompt for that uploading videos into gemini to then give you very clear actionable direction on like how you can add creative diversity in because it can read the video it can understand it if you give it the right context it can hypothesize on what's working and not based on benchmarks and soft metrics and that's something obviously gpt can't do yeah i just think the most recent gemini as well and there's i think there's one release in very soon, maybe in October.

26:06He's got a really long extended context window, which I think is really, really valuable. That's the most valuable part. 100%, that's the biggest limitation. I also think Claude are putting out a lot of interesting stuff right now. Claude as well, that's the problem, they're all very good for different things. Like Claude for script writing without a doubt. Anything written. They brought out skills obviously where you can teach it like specific i don't even i've not actually used it and i don't want to again butcher the explanation but they brought out this almost the ability to like train it on specific skills and then leverage that as you would train probably an individual in a company yeah exactly i guess it's a nice there's a bit of a segue but it's a nice like bridge to getting a bit deeper into some of the ai work we've been doing internally or you've been doing specifically with the help of sam Tom and Harry who's a consultant we're utilizing um so yeah it'd be good to dive through that I think one thing that we're working on just top level is like data warehousing which I think we've made some really large strides over the last sort of three months specifically obviously we've got like add data context warehousing combined shared learnings but we've also got we've started to really think about ingesting core logs like really just warehousing all of the data that we're creating as a business that's then promptable and something we can interact with but from a creative perspective like how have how have you been sort of driving this forward last few last few months yeah i think the the thing that's become more just more and more apparent is just the need for incredibly strong context to ensure that we're putting the right input in to get the best possible output.

27:59I think the way you need to think about that is like an LLM is never going to have the understanding or the knowledge that you're going to have around a specific area. So for example, that might be in a really micro sense, top of funnel script writing. we've got that knowledge and we've got that context to be able to give it more directional inputs so you can get the best possible output because if otherwise it is it is working off existing knowledge that it might have through what it can find online which isn't going to be anywhere near the standard of knowledge that we will have that is specific to our business so one thing that we've built is a really extensive context library for every part of the creative strategy process whether that's hooks um so how to create incredible hooks with like examples this is pairing so there's two documents for each so one is how to create incredible hooks with all of the knowledge we have of what makes great hooks all of the best performing hook frameworks hook layers so from sound visuals text on screen etc etc um and then the second document is best performing hooks from the last 90 days.

29:14And by pairing those two documents together and uploading that with your prompt, it is going to give a very directional output that uses the context that you've uploaded to make it much stronger, essentially. We're doing that for every part of the process. So whether that is the research phase, whether that is iterating. So we've got a master iterator prompt. So you're training it to become an incredible, it's incredible iterating top of funnel scripting and then having context documents that break down the frameworks of all of your top spenders for statics UGC high prod from the last 90 days because by giving it that as I say that directional context it can do an incredible job at just really improving the output so if I was anyone that's listening to this right now I would start thinking through that lens.

30:07How can we create context documents that almost creates a framework for the LLM to work in to give you the best possible output? So each stage of your business that is. And as well as that, lean into the data. So you know what your top spenders are, you know what's working in your account, download those, create frameworks off the back of them, put them into documents. Because just by uploading all of that context into GPT or whatever LLM you're using, it's going to give a much better output and it has much better context of what's currently working right now and it can do a great job at consolidating all of that.

30:46Yeah, you mentioned different stages of the business. I think you can do, we obviously speak a lot about this in a creative context, but you can do it in any area, sales, marketing. And I do think it's just like, it's like, it's quite boring is the wrong word but it is like documenting things isn't the most exciting thing to do in the world but it's like it really is the key it's like doing the boring work spending the time the getting into like a flow of just building out those context documents building out what best practice looks like taking that out of your head and putting on paper which i think people probably don't spend enough time on exactly because the way you the way you got to think of it as well it's like imagine imagine closing your eyes and just shooting a gun to try and find the target you're not you're going to miss nine times out of ten whereas if the the mask comes off and you've got a very clear target you're going to shoot and you're going to hit that target that's how i look at context documents and uploading them with your prompts because you might have a trained gpt bot that you think you've uploaded 10 20 30 documents about the brand all of the reviews it's it's too there's too much there to be able to give directional context to the thing that you're asking it you've only got a certain context window it's like a script you know if you have a one-page script you can only write so many words on that once it gets past that it's irrelevant it's not going to be able to read that information which is why we need to become like more directional with the context another example just like outside of creative and um you know d2c is this podcast so i was talking i was talking to you and matt the other day and i've got a context document about what a great communicator is and how to communicate well on a podcast and i just pulled that from actually watching i need to find his name but he's a communication expert he went on diary of a ceo i literally pulled the transcript it's like a two and a half hour episodes so many actionable takeaways from that which is all about how to present well, how to communicate, how to avoid filler words, all of these different things.

32:53I downloaded the transcript, which again, if you just go to any YouTube video, there's a little button that says show transcript, copy and paste it, put it into chat GPT and tell it to create a context document out of that. You can maybe pull context from a few different videos and then it will create a great context document for you about how to become an incredible communicator. i then use that context doc i upload the transcript from one of our episodes and ask it to critique me and give me feedback based on that and because it's got the context to become what to know what a great communicator is it can do that god i'm not doing that i don't do it for every episode by the way but it's something i've it's more because i've noticed i use filler words occasionally so i'm trying to get it out of the system i also just i have a habit of saying the same phrases over and over again so it gives me it gives me direction on how to say this in a slightly different way next time it's a good idea um i do i think like um so that's an example of like using ai is like a personal development help which i think it really is very it's really good at that um in a world where you maybe as a founder of a business you don't get as much feedback as other people i think it's like how do you look and build a process to give you feedback we're building right now a similar thing for our sales team so as every sales call it gets we we we've uploaded context on our business what we're selling icp got loads of call recordings every sales call whips into into chat gpt or an ai tour can't remember which one and then goes back in a slack channel to our sales guy with a real concise breakdown of what he could have done better and it's just like shortening that feedback loop we're training that on loads of like high ticket sales leaders and experts and that's an example of building a feedback loop yeah i think it's really valuable to think through that managers as well like if you've got new managers in your business and they're holding or conducting reviews for the team you know being able to to do the same thing you just mentioned but for a manager and having context that and being able to to score based on how could they have improved in giving this feedback what did they score on empowerment or empathy directness, candor all of these different areas that means that you can get that instant feedback loop and become better at what you do essentially without having the strain of having someone else to do that that context example is third party I also think we've mentioned this on a previous podcast but I can't recommend that whisper flow enough.

35:33Oh, that's unreal. Alex Cooper put us on that. Yeah, he did. I think it's my most used thing on my computer. I use it every day, multiple times a day. It's just voice dictation, which allows you to just upload personal context at length without having to type it out. Yeah, or even like a Slack message. Yeah, yeah. Anything. It's so good, so good. switching topics a little bit towards video gen because i know we've got some bits to chat through here we've seen as we've touched on since we did load the vo3 deep dive just a massive shift in this market again dread to think what q1 is going to look like but um yeah be good to just touch on like how we're utilizing video gen now versus maybe three three months ago yeah so obviously Sora 2 has come out which is better than Vos 3 I think right now in certain ways especially for like low prod style video that feels yeah feels very very authentic and real.

36:39If anyone watching has a way for us to access these models ahead of the UK release that doesn't involve building some lengthy API tool I'd love it like please let us know in the comments message me we'll pay you like because we still don't have access to sora 2 in the uk and we're just having to use like apis which yeah it's just doesn't work the same no but um yeah anyway but but yeah so obviously we we can't we haven't been able to really stretch our legs there we've obviously been doing we've been able to do a few things like those secret filming um style uh videos where it's like a faking a cctv camera or ring doorbell and stuff like that but But I think just wanted to just before we go into maybe actually let's start here and then I want to go into what next year probably looks like.

37:29Yeah. And some some predictions there. But I think in terms of through the lens of VO3, as I mentioned earlier, we are now using that more and more for like product B-roll because of the ingredients. The ingredient. What's it called? I forgot the name of it now. Ingredient to video feature. we are uh able to now very easily produce pretty strong b-roll that would have taken us you know an hour two hours to have set up and shot in a studio which is you know reducing the operational load on the production team the problem is right now obviously as i said we're looking to to bring in ai videographers and visual prompt engineers it puts more strain on post-production so that's obviously something that we're working through the second thing that i am just starting to see work better and better now is vo hook variations and this very much aligns with what obviously we're doing with andromeda and i want to talk maybe through the the best vo hook variation frameworks that we're seeing right now working number one just has to be the visual metaphor so take taking an abstract idea and transforming it into a literal and exaggerated image almost so example gut you know if we're talking about gut health as a problem um that could be a visual of showing uh a light a light in the gut going from red to green or showing a stomach blowing up like a balloon anxiety so you know the hook might be uh does your anxiety ever feel like your head's blowing up like a balloon and that someone's head just slowly slowly um expanding skin uh we had a great example that's a top spender for a brand the original creative didn't work and the second one was talking about um was the same hook but the visual changed and the hook was after 40 are your collagen depletion um your collagen depletion almost triples or something like that and previously it was just the shot of the founder vsl style and then we just overlaid it with a vo shot of a woman's skin from going really young a close-up of her face going from really young skin to the wrinkles suddenly appearing and now that i think got like 40 tssr and working really well so those visual metaphors are working great i guess that's yeah it's like it makes sense because it's very like um it probably captures the tension it's super it's a little bit more obscure than maybe something you're able to produce naturally with footage i've also seen a few examples of where brands are like able to like double down on the feeling you're trying to convey in more ways in an asset so maybe you'd be trying to talk about brain fog or energy this is me off the top of my head and it would have like someone just looking sad looking like under the weather but then they've overlaid loads of clouds like the environment they're in looks like a bit grim like low low energy like they've multiplied it in in ways beyond just the person exactly that was that was the next one so great work exaggerated cause and effect i actually didn't read that on the notes so yeah exactly that so taking almost the the problem and exaggerating it through through the visual pushing it to the extreme and that isn't just necessarily problem related it could be based on desired outcome as well so for example if it's a product or beverage that is supposed to give you like all day energy taking a sip and suddenly you've you know you're at the top of a mountain things like that um but actually out of all of this the stuff that is just working best for vo right now is the stuff that isn't as crazy uh obvious and what's funny about this is like it's actually my opinion on this has kind of changed where previously it was like let's lean into the obviously ai yeah to actually i'm i'm seeing the stuff that isn't as obviously ai performing better in in in some cases and i think an example of that is we did a creative where it was talking about does your stomach ever feel like a balloon about about to explode but it was actually we made through vo3 an animation of and it was like a very basic looking animation that didn't look like it was created through ai of someone's stomach piercing someone's stomach and a load of sand falling out of the stomach um and i think again that's something to to also be considering here alongside these clearly ai driven creatives how can we create through vo3 in a way that doesn't necessarily make it look like ai and it's finding that balance of the two because i think we're living in a world right now where at least through um tiktok if i'm looking at the feed every single comment on any type of video people just are paranoid about whether this is this ai I feel like this is AI and people were always questioning that so so yeah so I think that's going to be there's I've already seen cases of pretty strong backlash in two campaigns I can't and I'm telling a story without the exact story in my head which I've done before on the podcast it's never great but there was a example of a heritage British brand recently that did a a brand campaign on Instagram and they clearly used AI for it and the comments were just destroying them i've also started to see on someone sort of posted on twitter like films at the end of credits have started to input input put like generative ai was not used in the creation of this this media which i think is like a sign of like interesting sign of where where we're maybe going and the sort of narrative that's going to be brought into more of those like traditional creative formats where people are probably going to be a little bit more definitely more like negative about ai taking over like creative industries film television um but yeah i think it's going to be really interesting to see how that sentiment plays out across the whole because we operate in a bit of a vacuum through meta through tiktok through social channels which are always going to be the first and they're probably going to have the least negative reception but it'd be really interesting to see how that plays out in these traditional mediums yeah definitely like i feel like speaking to friends and people that aren't in our industry at all um it feels like it's it's very villainized in many ways in other industries obviously not so much in in uh in the the world the land of cowboys if it works we'll do it that's the narrative i guess that's a good segue into like where we think this is maybe going into 2026 we've touched on that production department piece which i think i think we're quite far along ai adoption as an agency um and we're doing a lot of work operationally like onboarding sales which i mentioned um just giving for giving our team the tools to do better work through leveraging ai with better context inputted into it and bringing in the production team but where do you where do you think we're going to head into like early 2026 and beyond yeah i think in starting with something that's really boring in many ways but is gonna i think the people that are going to win with this new creative diversity are the ones that actually build systems of organizing assets and that also comes through ai so i think going into next year i really see us for example building agents that when you say shoot some net new content or a net new piece of b-roll or even a static with someone holding a product or you know a product flat lay through the use of an agent taking that and creating 10 20 30 more versions of that that creates the visual differentiation within the creative because i think what is the you know if we're going to want to keep continue to iterate at scale through visual diversity we're not only going to have to find ways to multiply how we diversify the visuals but also how that creative is organized into folders into things like frame 4.0 where it's easily accessible because otherwise if you know as volume continues to increase those that have proper organization in place and you know through how we are almost modulating the assets that we upload whether that's hooks whether that's b-roll whether that's product fat lays to be able to easily find these at scale i think are the ones that are going to win yeah i dread to think in some brands how much of like a graveyard of asset production that some some brands have like the way they're not doing that well and things just sit there they get used once and never get touched again yeah i also think it's interesting to think about if you do that well that becomes a great context library in itself when you start to be able to upload things to sort of vo3 i think that's why data warehousing is so important because the the the data from the ad account becomes the context that then if you if you have trained the model on the right theory of okay this is this is what it's got the data of what works in the account it's got data of how to iterate based on certain metrics it can then get to a point where it can actually prompt really well and produce really good assets through or at least image or video from the asset library things that could be used exactly yeah and that's what i'd look into frame 4.0 if people aren't using that because it's frame 4.0 is great for that you can search uh it's it's it does ai labeling based on reading the visuals of the assets and it also um transcribes every asset so you can find things just by typing in a word or or do something so it's definitely one to look at i would say another thing just on the creator side is and i think jake paul was it was really interesting what he um obviously there's there was loads of the sort of two content that of jake paul just like put applying makeup for like funny hilarious actually and i think he's just so smart i think in he obviously there was so many of these videos coming out about him that it was almost like a little bit fishy and i knew he was involved in some way but he was the first person to like license his face on the platform he's also an investor of open ai and he then came out about that and obviously everyone's using his face now to create this funny content because he's he was the first person to i don't know if license if they use i think it's called something else on sora 2 but essentially licensing a face to be used and reproduced on the platform i think we're going to be in a world where next year where we're seeing more and more ugc creators doing that through a platform like sora so we pay to be able to use the license of their face to create assets or ads with that i think that makes sense the interesting counter to that is like why not just generate them from scratch when it's someone that doesn't have the sort of publicity to to his level because it's it's creative diversity and volume right so if you've got some if you look at like all of the stuff that jake paul that people have made jake paul do they've made him like go into mcdonald's and in a dress screaming at someone at the counter like applying makeup and doing like a get ready with me you could take that person's face and at a lower cost be able to produce hundreds if not thousands of assets yeah it makes sense i get it in his context it's like from a brand context it's like do you buy someone's license or do you just generate someone who doesn't exist i don't think i think there's a i don't think it's possible to do that yet in that quality yeah exactly i think it's only good to latch on it's really good at latching on someone's likeness after about a five second video upload from what i've seen which is nuts but yeah it's interesting um i i also think we may be interesting to think about how experts then maybe play out in that in that whole role like will we see like credible experts then license their likeness to brands that they're involved with it make that whole workflow easier yeah it's a it's a dangerous game isn't it like with yeah i just i'd be interested i mean where do you think it's going to go in terms of like legislation in the next 12 months there has to be some well it's it's it's illegal to do fully generated ai ads in certain categories already i believe in america i've definitely seen some some some some like publications and uh read through some some lengthy twitter threads on that but it will i don't know what it'll be what it'll look like but i just i i feel like there'll be a significant shift in advertising legislation yeah there has to be i mean i feel like if we we in the uk will get hit by that more than the uk states usually that's the order right yeah yeah because we've spoken about this before but it's it's like the race the race against china with ai isn't it and it's like if you if you introduce legislation and say the us it's going to slow down that and reduce their chances of winning that race yeah yeah it's interesting anything on your side in terms of predictions for next year yeah in general just in terms of ai god i i'm probably the episode coming out after this or a couple of episodes after this of doing a lengthy episode with around s ai seo and the rise of of like commerce through ai i think that's something that's going to be a massive opportunity into 2026 i think it's only really it's just starting to get going as a as a channel for for marketing and uh getting in distribution of products um which just adds more complexity to brands it's like more places to meet customers which is great but there's more things to think about um so i think we're going to see that that continue and accelerate pretty aggressively i think we're going to see more of these i just feel like most things on the internet are advertising channels whether that's whether it starts as an advertising channel or it gets to being an advertising channel at some point like you look at amazon that is an ads channel as well as a commerce channel like netflix like most of the things you most of the places you go on on the on online serve ads in some format open ai will start to do that soon i imagine they'll start to do it in every app it's just crazy like what that would do to their market cap yeah if you think about like the value of that company already without advertising i would go and look if you for those watching who haven't have have a look on twitter or linkedin and look for the retention curve on chat gpt usage it's absolutely bonkers so like a cohort graph on how churn rate it's it's crazy it's like basically doesn't doesn't exist basically doesn't exist yeah um so yeah yeah hard to hard to predict so hard to predict but um just got to keep yeah i think my philosophy has what i've learned this year it's like i didn't think it i didn't think good ai video gen would would be here right now i thought it would likely hit next year whereas it's it's you know it's it's come way quicker than than we thought it would and i think that's got to be the philosophy moving forward if you think something's coming short on your time horizons by 50 and it's likely gonna gonna hit probably sooner to to that point as we like to ask internally how do we make it happen in half the time hypothetically how do we how do we make this happen in 30 days not 60 yeah um side note that's a great question to ask great challenge your challenge your expectations of timelines um yeah even if you look at that alex cooper episode i think we we were discussing percentage of assets from ai this time that time next year and i think we definitely made some lower bound assumptions then where i'd be shocked if it isn't like eight percent we'll see yeah agreed but i think you still need to it's still I think the beauty of it is it reduces the barrier to entry but it increases you know the the creativity in which you have to deliver that yeah because you know it's a tool at the end of the day and it's how you use that tool 100 % 100 % it's about it's the wizard not the one on that note should we wrap up yeah let's do it probably got 90 % drop off as soon as I said that line.

54:31Oh yeah. Thank you all for watching as always. Yeah. Another, another banger. If you enjoyed these solo apps, let us know in the comments. Appreciate if you could subscribe and see you on the next episode.